Comprehensive Compilation of Real-World LLM & AI Agent Use Cases in Financial Services
140
85 commits
updated Aug 25, 2026
README
Compilation of Real-World LLM & AI Agent Use Cases in Financial Services
Context
Large language models (LLMs) and AI agents are transforming financial services: from streamlining regulatory compliance and accelerating deal analysis to personalizing customer experiences and automating complex workflows across banking, trading, and asset management.
This repository curates real-world LLM and agentic AI implementations in finance, showcasing enterprise platforms, specialized models, and integrated solutions deployed by leading financial institutions worldwide.
Have a great example of LLMs in the financial services sector? Contribute by opening a pull request or submitting an issue.
๐ Table of Contents
- Retail & Commercial Banking
- Covers personal and business banking servicesโchecking and savings accounts, mortgages, SME financing, and credit solutions. Also addresses central banking operations, digital transformation in traditional banks, and AI-driven advisory for both retail and corporate clients.
- Wealth & Asset Management
- Encompasses asset and wealth advisory, mutual funds, private equity, institutional asset management, robo-advisory, and portfolio management for high-net-worth individuals and institutional clients.
- Capital Markets & Trading
- Covers investment banking, securities trading, market infrastructure, equity research, credit analysis, and financial market analytics. Includes broker-dealers, exchanges, and financial data platforms.
- Payments & Transaction Services
- Focuses on digital payments, card networks, cross-border transactions, BNPL, and payment infrastructure. Also highlights AI-based fraud detection, risk assessment, and automated payment processing innovations.
- Insurance & Risk Management
- Addresses life, health, property, and casualty insurance, along with reinsurance, underwriting automation, and actuarial modeling. Features AI-driven claims processing, fraud detection, personalized policy recommendations, and broader risk assessment tools.
Each example can be classified into one of 3 categories based on type of use case:
- Enterprise Platform: Large-scale LLM deployments spanning multiple lines of business or the entire organization. These implementations typically address a broad range of tasksโsuch as customer support, internal operations, and complianceโwithin a unified, centrally managed framework.
- Fine-Tuned Model: Models either pre-trained or fine-tuned on focused datasets (e.g., regulatory filings, financial news) to achieve deep domain expertise. By narrowing the training data to specific tasks and terminology, they deliver higher accuracy and more relevant insights than general-purpose models.
- Integrated Solution: Vendor-provided or off-the-shelf AI solutions deployed for specific workflows and use cases. Financial institutions leverage ready-made integrations, cloud services, and prompt engineering to streamline tasksโsuch as AI-powered chatbots or document processingโwithout extensive internal development.
Retail & Commercial Banking
| Name | Type | Date | Description | Site | Paper |
|---|---|---|---|---|---|
| DBS Agentic AI for Corporate Bankers | Enterprise Platform | Aug 2026 | DBS scaled an agentic AI solution that streamlines credit memo preparation for corporate bankers serving large and mid-sized clients, expanding from a 150-participant pilot to around 1,500 relationship managers and credit risk managers globally. Specialised agents tackle more than 70 tasks; with credit memos and related credit activities taking up to 40% of relationship managers' time, DBS targets a reduction of at least 30%. No model or vendor named. | ๐ | - |
| Westpac Lending AI Agents | Integrated Solution | Aug 2026 | Westpac put five specialist AI agents built on Amazon Bedrock AgentCore into production home loan and credit card assessment. An account scrutiny agent classifies transactions from manual bank statements; others extract payslip data and verify it against lending policy, with bankers retaining oversight. They process 32,000+ payslips and 1.5M+ transactions weekly, saving 12,500 banker hours in their first month โ over 150,000 annualised. | ๐ | - |
| DBS Joy Agentic AI | Enterprise Platform | Jul 2026 | DBS made DBS Joy, its corporate virtual assistant (launched 2018, given GenAI capabilities in 2025), fully agentic for 350,000 corporate and SME clients in Singapore, letting them retrieve and analyse account and transaction data within a single conversation rather than searching the platform manually. No model or vendor named. In H1 2026 chats rose 22% and active users 61%, cutting calls/emails to customer service by 7%. Another 100,000 Hong Kong users follow in September, with DBS Joy and the retail DBS digibot together expected to handle 1M+ chats monthly across 10M customers. | ๐ | - |
| HSBC & Google Cloud | Enterprise Platform | Jun 2026 | HSBC and Google Cloud announced a multi-year AI partnership using Gemini models, the Gemini Enterprise Agent Platform and Google DeepMind. Applications include financial-crime monitoring across the near one billion transactions HSBC screens monthly (intervening twice as fast), a relationship-manager decision assistant cutting prep from hours to minutes, and hyper-personalised wealth advisory. The deal targets more than 200 new AI use cases over two years. | ๐ | - |
| Santander AI-First | Enterprise Platform | Jun 2026 | Santander extended AI access to all 185,000 employees under its "AI-first" strategy, using a multi-provider stack (Microsoft Copilot, OpenAI ChatGPT, Anthropic Claude, Google Gemini, G42). Openbank's models process ~100,000 AML alerts/year (hours to minutes); UK self-service handles ~240,000 calls/year, freeing ~45,000 service-team hours. AI delivered โฌ35M in Q1 2026, on track to exceed โฌ200M in 2026, targeting โฌ1B over 2026-2028. | ๐ | - |
| Revolut AIR | Integrated Solution | Apr 2026 | Revolut launched AIR (AI by Revolut), an in-app AI assistant that lets customers manage money through natural conversation โ answering questions on spending, budgeting and investing, pausing recurring payments, managing card controls, checking live exchange rates, and helping with trip budgets and in-app eSIM purchases. Sensitive actions require biometric approval, and Revolut says personal data is never stored by third-party AI partners or used for model training. Rolling out from April 2026 to its ~13 million UK customers first, with other markets to follow. | ๐ | - |
| PRAGMA (Revolut) | Fine-Tuned Model | Apr 2026 | Revolut's family of encoder-only Transformer foundation models (10M, 100M and 1B parameters) pre-trained with a masked-language-modelling objective on 24 billion banking events (207 billion tokens) from 26 million pseudonymised user records across 111 countries over 25 months of transaction, app, trading and communication data. Frozen embeddings with a lightweight head โ or LoRA fine-tuning โ outperform internal task-specific baselines on credit scoring, external fraud, product recommendation and lifetime-value prediction. | - | ๐ |
| Starling Assistant | Integrated Solution | Mar 2026 | Starling Bank launched Starling Assistant, billed as the UK's first agentic AI financial assistant, built on Google Gemini running inside Starling's own Google Cloud environment. Rolled out to personal current account holders from 20 March 2026 (business and joint accounts to follow), it uses voice and natural language to set up savings transfers, organise bill payments, create budget spaces, analyse spending and guide accessibility support. | ๐ | - |
| NatWest & OpenAI | Enterprise Platform | Feb 2026 | NatWest became the first UK bank to sign a strategic collaboration with OpenAI, embedding OpenAI models (version unspecified) into its Cora customer chatbot and AskArchie employee assistant. Customers can ask natural-language spending questions and report fraud in real time; around 25,000 customers gain the new agentic financial assistant by end of Q1 2026, with voice-to-voice capability piloting later in 2026. | ๐ | - |
| HSBC & Harvey AI | Enterprise Platform | Jan 2026 | HSBC partnered with Harvey AI to deploy AI across its global in-house legal function, streamlining workflows in contract analysis, due diligence, compliance, and litigation. | ๐ | - |
| BNY Mellon - Eliza | Enterprise Platform | Dec 2025 | BNY developed Eliza, an enterprise AI platform originally powered by OpenAI models, enabling 20,000+ employees to build agents across 125+ live use cases in legal, risk, and client management โ cutting legal contract review time 75% with 99% workforce adoption. In December 2025, BNY partnered with Google Cloud to integrate Gemini Enterprise (plus Gemini 3 and Veo 3) into Eliza for agentic market research. | ๐ ๐ | - |
| HSBC & Mistral AI | Enterprise Platform | Dec 2025 | HSBC partnered with Mistral AI in a multi-year strategic collaboration to deploy self-hosted GenAI models across the bank, powering productivity tools for tailored client communications, financial analysis, multilingual translation, and faster development cycles. | ๐ | - |
| Lloyds AI Financial Assistant | Enterprise Platform | Nov 2025 | Lloyds Banking Group unveiled what it calls the UK's first large-scale, multi-feature agentic AI financial assistant, built on its proprietary Generative AI and agentic framework combining curated bank data with LLMs (no external model or vendor named). Launching in the mobile app in early 2026 for 21M+ customers, it offers 24/7 financial coaching with memory, expanding to mortgages, car finance and protection. | ๐ | - |
| Macquarie Bank & Gemini Enterprise | Enterprise Platform | Oct 2025 | Macquarie Bank is rolling out Google Cloud's Gemini Enterprise to all employees โ not just technical teams โ for personal agents (document summarization, research, content drafting) and enterprise agents spanning system-design documentation, code generation, conversation summaries, and customer support. 99% of employees completed generative-AI training, with a target of every employee integrating AI into daily workflows within six months. | ๐ | - |
| Bank of America's AskGPS | Enterprise Platform | Sep 2025 | BofA built AskGPS (Ask Global Payments Solutions), an in-house GenAI assistant trained on 3,200+ internal documents to help employees serving 40,000+ business clients answer complex inquiries. It reduces response times from hours to seconds, saving tens of thousands of employee hours annually. | ๐ | - |
| Wells Fargo & Google Cloud AI | Enterprise Platform | Aug 2025 | Wells Fargo uses Google Agentspace (powered by Gemini models) to deploy agentic AI across its workforce, enabling employees to find and synthesize information faster, automate tasks and workflows, and increase organizational agility. | ๐ | - |
| U.S Bank & Amazon | Integrated Solution | Aug 2025 | U.S. Bank deployed a real-time agent-assist system using Amazon Q in Connect together with Amazon Bedrock and Claude Haiku model to transcribe calls, detect customer intent, and surface relevant knowledge-base recommendations during voice interactions. | ๐ | - |
| Brex Assistant | Integrated Solution | Aug 2025 | Brex built an AI-powered financial assistant using Amazon Bedrock and Claude models to automate expense management, achieving 75% workflow automation and raising compliance rates from 70% to the mid-90s. | ๐ | - |
| PayU | Integrated Solution | Jul 2025 | PayU implemented a secure enterprise-AI assistant using Amazon Bedrock, deploying multi-agent workflows (via RAG and text-to-SQL) to enable employees to query business metrics, HR policies and operations dataโall within a private VPC to satisfy regulatory data-residency and compliance needs. | ๐ | - |
| Intuit TurboTax | Integrated Solution | Jul 2025 | Intuit built Intuit Assist, an LLM-powered AI assistant within TurboTax that helps millions of users understand their tax situations, deductions, and refunds using a hybrid ClaudeโGPT system with RAG, fine-tuning, and strict regulatory compliance | ๐ | - |
| Nubank | Integrated Solution | May 2025 | Nubank built an AI โprivate bankerโ using GPT-4 and agentic frameworks like LangChain and LangGraph to automate customer support and money-transfer workflows. The system handles millions of user queries and transactions autonomouslyโreducing transfer times and improving accuracyโthrough coordinated LLM agents and a production-grade LLMOps platform. | ๐ | - |
| Apoidea Group | Fine-Tuned Model | May 2025 | Apoidea Group, in collaboration with Amazon SageMaker HyperPod and using the LLaMAโFactory framework, built a solution called SuperAcc that is based on a fine-tuned multimodal model Qwen2โVLโ7BโInstruct to enhance visual information extraction from banking documents, boosting table-structure recognition accuracy substantially and reducing processing times from hours to minutes. | ๐ | - |
| Lloyds & Google Cloud | Enterprise Platform | Apr 2025 | Lloyds Banking Group migrated 15 modelling systems (hundreds of models) from on-premise infrastructure to Google Cloud's Vertex AI (now Gemini Enterprise Agent Platform), where 300+ data scientists and AI developers have launched 80+ ML use cases and 18+ GenAI systems into production. An income-verification model cut mortgage-application processing from days to seconds, with zero unplanned platform downtime. | ๐ | - |
| Capital One Chat Concierge | Fine-Tuned Model | Feb 2025 | Chat Concierge, built on a fine-tuned Llama LLM, helps car buyers compare vehicles, explore financing, estimate trade-in values, and schedule test drives. Fine-tuned on Capital One's proprietary data, it also utilized an agentic approach to understand preferences, tailor recommendations, ensure policy compliance, and engage in human-like interactions. | ๐ | - |
| Deutsche Bank & Google | Enterprise Platform | Feb 2025 | Deutsche Bank leverages Google Cloudโs Vertex AI and Gemini LLMs to streamline banking operation. It involves automating document processing for regulatory compliance, enhancing customer support with AI-powered assistants, and accelerating software development for financial services. | ๐ | - |
| CommBiz Gen AI | Integrated Solution | Jan 2025 | Together with AWS, the Commonwealth Bank of Australia (CBA) rolled out a Gen-AI powered messaging service to assist tens of thousands of business customers with inquiries, facilitating quicker payments and efficient transactions. They leveraged Amazon Bedrock Knowledge Bases, Claude 3 and Cohere LLMs, and Amazon OpenSearch as the vector database. | ๐ | - |
| North for Banking | Enterprise Platform | Jan 2025 | RBC and Cohere co-developed and securely deployed an enterprise generative AI (genAI) solution optimized for financial services, building upon Cohere's proprietary foundation models | ๐ | - |
| Amazon Finance | Enterprise Platform | Dec 2024 | Amazon Finance Automation built a generative-AI Q&A chat assistant on Amazon Bedrock to empower analysts to quickly answer customer queries by retrieving and synthesising policy knowledge, accelerating response times and reducing manual research | ๐ | - |
| Banestas & Google | Enterprise Platform | Dec 2024 | Banestes, a Brazilian bank, used Gemini in Google Workspace to streamline work dynamics, such as accelerating credit analysis by simplifying balance sheet reviews and boosting productivity in marketing and legal departments. | ๐ | - |
| Commerzbank & Google | Integrated Solution | Nov 2024 | Commerzbank utilizes Google's Gemini 1.5 Pro, a multimodal large language model, to automate financial advisory workflows by enabling efficient documentation of client interactions and analysis of complex financial data. Its long context allows it to handle lengthy financial documents and multimedia content seamlessly. | ๐ | - |
| OCBC ChatGPT | Integrated Solution | Nov 2024 | OCBC Bank became the first Singapore bank to deploy a generative AI chatbot, OCBC ChatGPT, to all 30,000 employees across 19 countries, aiming to assist with tasks such as writing, research, and ideation. Developed with Microsoft Azure and following a successful six-month trial, the chatbot operates securely on the bank's private cloud to ensure data security. | ๐ | - |
| BBVA & OpenAI | Enterprise Platform | Nov 2024 | Global financial institution BBVA signed an agreement with OpenAI for 3,000 ChatGPT Enterprises licenses, leading to increased productivity and creativity. Staff across various departments have developed over 2.9k specialized GPTs that boost efficiency, spark creativity, and share expert knowledge across their organization of 125,000 e.g., tasks like translating risk-specific terminology and drafting responses to client inquiries. | ๐ | - |
| PennyMac & Google | Enterprise Platform | Oct 2024 | Pennymac Financial Serives, a national home loan lender and servicer, integrates Google's Gemini LLMs across various departments to enhance efficiency and reduce costs while ensuring security and compliance. HR uses it for job descriptions and policy drafting, while the underwriting team uses Gemini to analyze proprietary data, improving regulatory understanding and best practices. | ๐ | - |
| Belfius & Mistral AI | Enterprise Platform | Sep 2024 | Belfius, a leading Northern European digital bank, deployed Mistral Large to power its customer-facing chatbot in Dutch, French, and English, migrating from a scripted bot platform to an AI-powered solution that delivers personalized banking experiences and reduces manual operator workload. | ๐ | - |
| BNP Paribas & Mistral AI | Enterprise Platform | Jul 2024 | BNP Paribas has partnered with Mistral AI to integrate LLMs across various sectors, including customer support, sales, and IT. This collaboration enables the bank to deploy advanced AI models on-premises, ensuring compliance with regulatory standards. | ๐ | - |
| DBS GPT | Enterprise Platform | Jul 2024 | DBS Bank developed an in-house GenAI co-pilot, the CSO Assistant, to support 500 customer service officers in Singapore. It delivers real-time call transcription, solutioning, and auto-filling of service requestsโachieving near 100% accuracy and reducing call handling time by up to 20%. | ๐ | - |
| Bitext Mistral-7b-Banking | Fine-Tuned Model | Jun 2024 | Fine-tuned version of the Mistral-7B-Instruct-v0.2, specifically tailored for the banking domain. It is optimized to answer questions and assist users with various banking transactions | ๐ | ๐ |
| Scotiabank & Google | Integrated Solution | May 2024 | Scotiabank uses LLMs via Google Cloud to enhance customer service and automate document processing. It features LLM-powered chatbot summarization for seamless handoffs and faster resolutions, plus enhanced Q&A and search for streamlined employee workflows. | ๐ | - |
| ING Bank | Integrated Solution | Feb 2024 | ING implemented a generative AI-powered customer-facing chatbot to enhance customer service efficiency. This chatbot utilizes a multi-step process that retrieves information from data stores, ranks potential answers by relevance, and applies strict guardrails to ensure accurate and appropriate responses. | ๐ | - |
| Citi & GitHub Copilot | Integrated Solution | Feb 2024 | Citi has focused on improving developer productivity by embracing LLM-based coding assistants by rolling out GitHub Copilot (powered by OpenAIโs Codex) to all of its 40,000 software developers enterprise-wideโ. It serves as an AI pair-programmer, suggesting code snippets, functions, or fixes inside developersโ code editors. | ๐ | - |
| Akbank & Azure OpenAI | Integrated Solution | Jan 2024 | Akbank, one of Tรผrkiye's largest banks, uses Azure OpenAI Service to power an AI chatbot that automates customer support, improving accuracy to 90% and cutting response times by three minutes per interaction. This enhances efficiency and allows agents to focus on proactive service. | ๐ | - |
| Ally Financial - Ally.ai | Integrated Solution | Dec 2023 | Ally Financial, the largest digital-only bank in the US and a leading auto lender, launched Azure OpenAI LLM-powered Ally.ai to more than 700 customer care associates in summarizing conversations between them and Ally customers. This automation of post-call documentation for customer service associates is done through LLM summarization of customer call transcripts. | ๐ | - |
| Westpac & KAI-GPT | Fine-Tuned Model | Jun 2023 | Australian lender Westpac is using KAI-GPT, a banking industry-specific LLM, to help bankers locate, interpret and understand information from policies, regulatory filings, procedures, and complex financial products. KAI-GPT is based on Pythia-Chat-Base-7B, fine-tuned on banking-related dataset comprising 24k question-answer pairs from Common Crawl, 18k questions from Kasisto's own conversational data, and 245 million words from 44k banking-related documents | ๐ | ๐ |
| XuanYuan 2.0 | Fine-Tuned Model | May 2023 | Chat model (built upon the BLOOM-176B architecture) trained by combining general-domain with domain-specific knowledge and integrating the stages of pre-training and fine-tuning, It is capable of providing accurate and contextually appropriate responses in the Chinese financial domain. | - | ๐ |
| BBT-FinT5 | Fine-Tuned Model | Feb 2023 | Chinese financial pre-training language model (1B parameters) based on the T5 model, and pre-trained on the 300Gb financial corpus called FinCorpus | - | ๐ |
Wealth & Asset Management
| Name | Type | Date | Description | Site | Paper |
|---|---|---|---|---|---|
| OCBC WoW | Integrated Solution | Jul 2026 | OCBC launched WoW (Whole-of-Wealth), billed as the region's first AI-native banking app, using two GenAI avatars (Wendy and Wayne) to deliver hyper-personalised, 24/7 wealth adviceโportfolio analysis, market news, and investment ideasโvia voice or text to Premier Private Client customers. A related agentic-AI source-of-wealth (SOW) tool cut SOW report preparation from 10 days to one hour. | ๐ | - |
| Morgan Stanley at Work (MCP) | Enterprise Platform | Jun 2026 | Morgan Stanley at Work is opening its Shareworks and Equity Edge stock-plan administration platforms to external AI agents via the Model Context Protocol (MCP), letting clients' agents pull equity-compensation data and insights directly in an agent-to-system manner. The platforms hold roughly $7.35 trillion in assets across about 3,400 administration clients, including nearly half the S&P 500; full rollout is targeted for 2027. | ๐ | - |
| Charles Schwab Portfolio Insights | Integrated Solution | May 2026 | Schwab launched Portfolio Insights, its first generative-AI capability for retail investors (no LLM vendor named). It produces narrative daily performance summaries for the up-to-five holdings with the largest moves, recaps market and industry news on clients' S&P 500 positions, and adds Schwab Center for Financial Research commentary. Rollout to eligible self-directed U.S. clients began May 2026. | ๐ | - |
| Citi Sky | Enterprise Platform | Apr 2026 | Citi Wealth unveiled Citi Sky, an always-on conversational AI "team member" built on Google Cloud's Gemini Enterprise Agent platform, Google DeepMind's real-time avatar technology, and the Gemini Live API. It delivers Chief Investment Office market insights and financial guidance (e.g., CD-maturity alerts) via audio/video in English and Spanish, with a phased rollout to U.S. Citigold clients from summer 2026. | ๐ | - |
| Vanguard Expert Insights | Integrated Solution | Apr 2026 | Vanguard launched Expert Insights, a generative-AI portfolio-analysis tool (no model or vendor disclosed) that lets financial advisors generate instant, methodology-aligned, client-ready guidance and pressure-test portfolios. Currently piloting with select advisors, it will be embedded on-demand in Vanguard's Portfolio Analytics Tool later in 2026. Vanguard's annual advisor portfolio-analysis engagements have roughly quadrupled over the past six years. | ๐ | - |
| Morningstar Advisory AI Assistant | Integrated Solution | Mar 2026 | Morningstar embedded a new AI assistant in its Direct Advisory Suite (distinct from 2023's "Mo" chatbot), grounded in Morningstar's independent data and research. Advisors use natural-language requests to surface meaningful rating changes, prepare meeting briefs, and turn client statements into actionable proposals; it is also reachable through other AI tools via Morningstar's MCP connections. Initial U.S. rollout, with broader U.S./Canada availability through 2026. | ๐ | - |
| Franklin Templeton & Wand AI | Enterprise Platform | Nov 2025 | Franklin Templeton (~$1.66 trillion AUM) formed a multi-year partnership with Wand AI to deploy agentic AI in live production across investment research, operations, and transformation initiatives, moving beyond pilots to enterprise-wide automation. Underlying LLMs were not disclosed; wide-scale rollout is targeted for 2026. | ๐ | - |
| Ardian GAIA | Enterprise Platform | Sep 2025 | Ardian developed GAIA, an internal GenAI platform (powered by Azure OpenAI and Mistral AI) that transforms how investment teams analyze and extract insights from complex documents, enabling 500+ weekly active users to analyze due diligence documents, financial reports, and legal filings. The platform has processed 18,000+ documents and generated 280,000+ AI-powered queries, achieving 2X growth in documents processed while accelerating deal execution and freeing teams to focus on strategic analysis. | ๐ | - |
| Fidelity International Freya | Integrated Solution | Sep 2025 | Fidelity International launched Freya, a conversational LLM-based assistant on its Personal Investing platform (500,000 customers), answering natural-language queries with reasoning-based responses and built-in guardrails that stop short of regulated advice. The underlying model/vendor is unnamed. A phased rollout will expand Freya to Workplace Investing and Fidelity Adviser Solutions. | ๐ | - |
| AlphaAgents (BlackRock) | - | Aug 2025 | BlackRock developed a role-based multi-agent system powered by GPT-4o to enhance wealth management in equity portfolio constructions, enabling LLM agents to collaborate autonomously on equity research, stock selection, and portfolio optimization. | - | ๐ |
| Nippon India Mutual Fund | Integrated Solution | Jul 2025 | Nippon Life India Asset Management Ltd. (Nippon India Mutual Fund) used Amazon Bedrock Knowledge Bases to deploy a RAG AI assistant, leveraging Claude 3 Sonnet on Amazon Bedrock to reformulate queries and rerank responses. The solution increased answer accuracy by over 95% and reduced hallucinations by >90%, cutting report generation time from 2 days to about 10 minutes. | ๐ | - |
| Alpian & Google Cloud | Integrated Solution | Jun 2025 | Alpian, Switzerland's first fully cloud-native private bank, built a Gemini-powered platform engineering tool on Google Cloud (Vertex AI, BigQuery) that lets developers deploy services and interact with infrastructure through a conversational interface within a deliberately bounded scope that preserves security and compliance guardrails. It cut deployment times from days to hours, sped up feature deployment by 25%, and automates compliance workflows aligned with Swiss FINMA standards. | ๐ | - |
| BlackRock Aladdin | Integrated Solution | Jun 2025 | BlackRockโs Aladdin Copilot uses generative AI to let investors query portfolio and risk data in plain language, surfacing precise, contextual insights. It enhances productivity, decision-making, and personalization across the Aladdin platform while meeting strict compliance and accuracy standards. | ๐ | - |
| FinE5 | Fine-Tuned Model | Feb 2025 | Fin-E5, the finance-adapted embedding model in FinMTEB, is built on the E5 model and trained using a persona-based data synthesis method to enhance performance across financial embedding tasks. | - | ๐ |
| PIMCO & Azure | Enterprise Platform | Feb 2025 | PIMCO developed ChatGWM, an AI-powered search tool built on Azure AI, to enhance client service by streamlining information retrieval for its associates. ChatGWM utilizes retrieval-augmented generation (RAG) to search across approved structured and unstructured data sources, and then process them with Azure OpenAI LLM to provide accurate and up-to-date information swiftly. | ๐ | - |
| Aditya Birla Capital | Integrated Solution | Jan 2025 | Financial services provider Aditya Birla Capital implemented SimpliFi, a generative AI chatbot built on Azure OpenAI Service, to enhance customer engagement and streamline financial services. SimpliFi assists users in navigating financial solutions independently, aligning with the preferences of their 25 to 35-year-old demographic. | ๐ | - |
| Touchstone-GPT | Fine-Tuned Model | Nov 2024 | Open-source financial LLM trained through continual pre-training and financial instruction tuning, which demonstrates strong performance on the financial bilingual (English and Chinese) Golden Touchstone benchmark. | ๐ | ๐ |
| Banca Investis | Integrated Solution | Nov 2024 | Banca Investis launched NIWA, a GenAI-powered investment advisory platform designed to enhance customer engagement through hyper-personalized financial services. Serving as a "digital junior banker," NIWA analyzes over 500 pieces of information dailyโincluding clients' financial assets, preferences, and market researchโto provide tailored financial insights and real-time responses to investment-related inquiries. | ๐ | - |
| FinTral | Fine-Tuned Model | Aug 2024 | Suite of multimodal LLMs built upon the Mistral-7b model and tailored for financial analysis. FinTral integrates textual, numerical, tabular, and image data, and is pretrained on a 20 billion token, high quality dataset | - | ๐ |
| Goldman Sachs & Meta | Enterprise Platform | Aug 2024 | Goldman Sachs introduced the GS AI Platform, a GenAI tool designed to enhance employee productivity across various use cases. Built on Meta's LLaMa models, the assistant aids in tasks such as summarizing and proofreading emails, extracting information from documents, as well as translating code between programming languages. | ๐ | - |
| Nomura & Meta | Enterprise Platform | Aug 2024 | Leading Japanese financial institution Nomura uses Meta's LLaMa models on Amazon Bedrock to democratize generative AI by driving faster innovation, transparency, bias guardrails, and robust performance across text summarization, code generation, log analysis, and document processing. | ๐ | - |
| JPMorgan Chase IndexGPT | Enterprise Platform | Jul 2024 | JPMorgan Chase launched a generative AI-based tool (via AWS Bedrock) called IndexGPT, designed to serve as a 'research analyst' for over 50,000 employees, aiding in various tasks that enhance productivity and decision-making within the firm. It is able to generate and refine written documents, provide creative solutions and summarize extensive documents. | ๐ | - |
| IDEA-FinQA | Fine-Tuned Model | Jun 2024 | Financial question-answering system based on Qwen1.5-14B-Chat, utilizing real-time knowledge injection and supporting various data collection and querying methodologies, and comprises three main modules: the data collector, the data querying module, and LLM-based agents tasked with specific functions. | ๐ | ๐ |
| BCI & Azure | Enterprise Platform | Mar 2024 | British Columbia Investment Management Corporation (BCI) integrated Microsoft 365 Copilot and Azure OpenAI Service to enhance productivity and streamline operations. This implementation has led to a 10%-20% productivity boost for 84% of initial users, saving over 2,300 person-hours through automation. Notably, the time required to write internal audit reports decreased by 30%. | ๐ | - |
| RAVEN | Fine-Tuned Model | Jan 2024 | Fine-tuned LLaMA-2 13B Chat model designed to enhance financial data analysis by integrating external tools. Used supervised fine-tuning with parameter-efficient techniques, utilizing a diverse set of financial question-answering datasets, including TAT-QA, Financial PhraseBank, WikiSQL, and OTT-QA | - | ๐ |
| InvestLM | Fine-Tuned Model | Sep 2023 | Financial domain LLM tuned on LLaMA-65B, using a carefully curated instruction dataset related to financial investment. The small yet diverse instruction dataset covers a wide range of financial related topics, from Chartered Financial Analyst (CFA) exam questions to SEC filings to Stackexchange quantitative finance discussions. | ๐ | ๐ |
| CFGPT | Fine-Tuned Model | Sep 2023 | Financial LLM based on InternLM-7B that is designed to handle financial texts effectively. It was pre-trained on 584 million documents (141 billion tokens) from Chinese financial sources like announcements, research reports, social media content, and financial news, and then fine-tuned on 1.5 million instruction pairs (1.5 billion tokens) tailored for specific tasks of financial analysis and decision-making. | ๐ | ๐ |
| FinMA | Fine-Tuned Model | Jun 2023 | Comprehensive framework that introduces FinMA (Financial Multi-task Assistant), an open-source financial LLM fine-tuned (7B and 30B versions) from LLaMA using a diverse, multi-task instruction dataset of 136,000 samples. The dataset encompasses various financial tasks, document types, and data modalities. | ๐ | ๐ |
| Fin-Llama | Fine-Tuned Model | Jun 2023 | Specialized version of LLaMA 33B, fine-tuned (with QLoRA and 4-bit quantization) for financial applications using a 16.9k instruction dataset. | ๐ | - |
| Morningstar - Mo chatbot | Integrated Solution | May 2023 | Morningstar introduced Mo, an AI chatbot powered by the Morningstar Intelligence Engine, which combines Morningstar's extensive investment research library with Microsoft's Azure OpenAI Service. Mo is designed to provide investors and financial professionals with concise, conversational insights by processing natural language queries and summarizing relevant information from over 750,000 investment options. | ๐ | - |
| Cornucopia-LLaMA-Fin-Chinese | Fine-Tuned Model | Apr 2023 | Open-source LLaMA-based model fine-tuned for Chinese financial applications. It uses instruction tuning with Chinese financial Q&A datasets to enhance domain-specific performance. | ๐ | - |
| Morgan Stanley & OpenAI | Pre-trained | Mar 2023 | Morgan Stanley Wealth Management announced a partnership with OpenAI to develop an internal-facing GPT-powered assistant (AI @ Morgan Stanley Assistant), allowing financial advisors to query the bankโs vast research repository and internal knowledge base in natural language | ๐ | ๐ |
| FLANG-ELECTRA | Fine-Tuned Model | Oct 2022 | Domain specific Financial LANGuage model (FLANG) which uses financial keywords and phrases for better masking, and built on the ELECTRA-base architecture. Note: Considered a smaller LM as it has fewer than 1B params | ๐ | ๐ |
| FinBERT-21 | Fine-Tuned Model | Jul 2020 | FinBERT (BERT for Financial Text Mining) is a domain specific language model pre-trained on large-scale financial corpora, allowing it to capture language knowledge and semantic information from the finance domain. Note: Considered a smaller LM as it has fewer than 1B params | - | ๐ |
Capital Markets & Trading
| Name | Type | Date | Description | Site | Paper |
|---|---|---|---|---|---|
| S&P Global Energy HorizonsAgents | Enterprise Platform | May 2026 | S&P Global Energy launched HorizonsAgents, a suite of AI agents converting complex energy and sustainability data into decision-ready, auditable insights for banks, investors, project developers, and corporate finance teams. Four agents โ Transition Finance (beta since November 2025), Data Center Intelligence, Sustainability Benchmarking, and Net Zero Investment โ deliver comparable outputs in minutes, aligned with ISSB frameworks. No underlying models named. | ๐ | - |
| Robinhood Agentic Trading | Integrated Solution | May 2026 | Robinhood launched Agentic Trading (equities in beta), letting customers connect their own third-party AI agents via Robinhood's AI-native MCP servers to a dedicated, separately funded account. Connected agents can analyze concentration/sector exposure, build and backtest thematic strategies, and autonomously execute trades, with safeguards including trade previews, push notifications, fraud review, and instant disconnect. The announcement names no specific LLM. | ๐ | - |
| Obin AI & Google Cloud | Enterprise Platform | Apr 2026 | Obin AI, an enterprise agentic-AI company founded by Google AI, JPMorgan, and Silver Lake veterans, partnered with Google Cloud, integrating Gemini and the Agent Development Kit and listing on Google Cloud Marketplace. Its Obin Financial Agent helps asset management, commercial lending, and insurance teams accelerate complex financial analyses; clients include institutions managing over $1 trillion in assets. | ๐ | - |
| QRAFTI | - | Apr 2026 | A multi-agent agentic framework for empirical quantitative-finance research that exposes a panel-data toolkit plus data access and factor construction via MCP servers as callable tools. Uses chained tool calls with reflection-based planning to replicate established factors, formulate and test new signals, and generate standardized research reports with narrative analysis. | - | ๐ |
| FactSet AI for Banking | Enterprise Platform | Mar 2026 | FactSet launched FactSet AI for Banking (alpha), an AI-native workflow-automation platform built with (and backed by an investment in) Finster AI for investment banking and sell-side teams. It automates creation of transaction artifacts across the deal cycle, from origination through execution, within a single governed environment with deep Microsoft Office integration, drawing on leading AI models (specific vendors not named). | ๐ | - |
| Bloomberg ASKB | Enterprise Platform | Feb 2026 | Bloomberg launched ASKB (beta), an agentic conversational interface for the Terminal running a coordinated network of parallel agents built on a mix of commercial and open-weight LLMs. It answers natural-language queries, auto-generates Bloomberg Query Language, and runs multi-step workflows (earnings prep, post-event analysis), synthesizing Bloomberg data, news, filings, and analytics plus research from 800+ providers with source attribution. Distinct from BloombergGPT. | ๐ | - |
| AlphaSense Generative Search | Enterprise Platform | Jan 2026 | AlphaSense rebuilt Generative Search as a multi-agent research system with dynamic planning and reasoning over hundreds of millions of premium documents plus structured financial data, with pre-built workflow agents that automate earnings analysis, market scans, and deliverable creation (slides and reports) for banking, PE/VC, and IR teams. A July 2026 update added native PowerPoint and Excel assistants; Blue Deer Capital reports building board-ready decks 60-70% faster. | ๐ | - |
| S&P Global & Google Cloud | Enterprise Platform | Dec 2025 | S&P Global and Google Cloud struck a multi-year partnership headlined by the S&P Global Data Retrieval Agent, built by S&P's Kensho AI unit, which lets customers query S&P Global data directly inside Gemini Enterprise with citation-backed answers. The deal also unifies S&P's proprietary datasets on BigQuery for AI-ready distribution and deploys Gemini Enterprise internally for workforce productivity. | ๐ | - |
| FinRpt | Fine-Tuned Model | Nov 2025 | First comprehensive framework for automating Equity Research Report generation using LLMs, featuring a dataset integrating 7 financial data types, an 11-metric evaluation system, and a multi-agent framework with SFT and RL training. AAAI 2026. | ๐ | ๐ |
| LLM Pro Finance Suite | Fine-Tuned Model | Nov 2025 | A suite of five instruction-tuned multilingual finance LLMs (8Bโ70B) built on Llama, Qwen and Gemma bases via continual pre-training plus SFT on ~2M samples (54% finance; English, French, German). Targets financial reporting analysis, risk, compliance, sentiment, translation and RAG, preserving general capability. Open 8B checkpoints published on Hugging Face under the OVHaiLLM org. | ๐ | ๐ |
| S&P Capital IQ Pro ChatIQ | Enterprise Platform | Oct 2025 | S&P Global upgraded Capital IQ Pro with Multi-Document ChatIQ and Document Intelligence 2.0 (ChatIQ built with S&P's Kensho AI unit), letting banking and buyside analysts query and compare across filings, transcripts, investor presentations, news, and research spanning multiple companies or industries, returning generative-AI insights with precise citations for full auditability. No adoption metrics disclosed. | ๐ | - |
| Coinvisor | Fine-Tuned Model | Oct 2025 | An RL-enhanced chatbot agent for interactive cryptocurrency investment analysis, using a reinforcement-learning-based tool-selection mechanism for multi-step planning and real-time integration of diverse crypto data sources. Improves tool-orchestration recall by 40.7% and F1 by 26.6% over baselines, with a 20-investor user study rating average satisfaction 4.64/5. | - | ๐ |
| Deutsche Bank DB Lumina | Enterprise Platform | Sep 2025 | Deutsche Bank Research built DB Lumina, an AI-powered research assistant developed with Google Cloud Consulting, combining Gemini models with a RAG architecture (Gemini Embeddings, Vertex AI Vector Search, pgvector) that grounds responses in internal research and sources like SEC filings with inline citations. In production since September 2024 with ~5,000 users (targeting 10,000+), it saves analysts 30โ45 minutes on earnings note templates and up to 2 hours on research reports. | ๐ | - |
| Trading-R1 | Fine-Tuned Model | Sep 2025 | Tauric Research's 4B financial trading model built on a Qwen3-4B backbone, trained with a multi-stage easy-to-hard SFT-plus-RL curriculum on Tauric-TR1-DB (100k samples over 18 months, 14 equities/ETFs, five data sources). Generates structured, evidence-grounded investment theses with volatility-adjusted decisions, delivering improved risk-adjusted returns and lower drawdowns versus open and proprietary baselines. | ๐ | ๐ |
| Moody's AI Studio | Integrated Solution | Aug 2025 | Moody's built AI Studio, a multi-agent AI platform that automates financial workflows like credit memo generation, cutting analyst time from 40 hours to 2โ3 minutes by orchestrating specialized agents using proprietary and third-party data. The platform has been commercialized for financial clients and adopted internally by 40,000 employees, driving large-scale efficiency and competitive advantage. | ๐ | - |
| Goldman Sachs & Cognition Devin | Integrated Solution | Jul 2025 | Goldman Sachs began piloting Cognition's Devin, an autonomous AI software engineer, across its 12,000-plus developersโinitially hundreds of instances, potentially scaling to thousandsโfor tasks such as coding, testing, and debugging under a human-supervised "hybrid workforce" model. CIO Marco Argenti estimates three-to-four-times developer productivity gains. Distinct from Goldman's Meta-based GS AI Platform. | ๐ | - |
| IG Group & Anthropic | Integrated Solution | Apr 2025 | IG Group partnered with Anthropic to deploy Claude for Work, doubling productivity in financial content creation, cutting over 70 hours of manual work weekly, and improving compliance and analytics workflowsโall with ROI achieved in under three months. | ๐ | - |
| Hebbia & OpenAI | Integrated Solution | Mar 2025 | Hebbia and OpenAI have partnered to enhance financial institutions' efficiency by integrating OpenAI's advanced language modelsโo1, GPT-4o, and o3-miniโinto Hebbia's Matrix platform. This multi-agent system automates complex tasks such as drafting investment committee memos, extracting key terms from credit agreements, and identifying red flags in legal documents, achieving up to 92% accuracy and reducing research time by up to 85%. | ๐ | - |
| Endex & OpenAI | Integrated Solution | Feb 2025 | Endex, an AI platform for financial firms, uses OpenAIโs advanced modelsโGPT-4o, o1-mini, and o3-miniโto power autonomous agents that retrieve, synthesize, and analyze complex financial data, helping firms automate tasks like earnings summaries, investment memos, and due diligence with greater speed and accuracy. | ๐ | - |
| FinBLOOM | Fine-Tuned Model | Feb 2025 | FinBloom 7B, built on Bloom 7B, was trained on 14M financial news articles and 12M SEC filings, then fine-tuned with 50K financial queries for enhanced real-time data retrieval. This process ensures strong contextual understanding for financial decision-making. | - | ๐ |
| MUFG Bank | Integrated Solution | Feb 2025 | MUFG Bank leveraged LLMs to automate data extraction and summarization from corporate reports, enabling faster financial analysis for FX & Derivative Sales. The system, leveraging retrieval-augmented generation (RAG) and fine-tuned prompts, reduces client presentation preparation from hours to minutes, thereby improving client advisory efficiency. | ๐ | - |
| Rogo & OpenAI | Fine-Tuned Model | Feb 2025 | Rogo's fine-tuned OpenAI models and integration of extensive financial datasets (including S&P Global, Crunchbase, and FactSet) allows it to scale financial analysis and deliver real-time financial intelligence to >5k financial professionals, shifting their focus from manual work to high-value decision making. It uses GPT-4o for in-depth financial analysis, o1-mini for contextualizing financial data, and o1 for evaluations and advanced reasoning. | ๐ | - |
| TigerGPT | Integrated Solution | Feb 2025 | Tiger Brokers integrated DeepSeek's AI model, DeepSeek-R1, into its AI-powered chatbot, TigerGPT. This adoption aims to enhance market analysis and trading capabilities for its customers through the improved logical reasoning capabilities. | ๐ | - |
| London Stock Exchange Group & Amazon Q | Integrated Solution | Jan 2025 | London Stock Exchange Groupโs post-trade services team deployed Amazon Q Business (running on Amazon Bedrock) with a front-end custom UI, document crawling of FAQs and rulebooks, and retrieval-augmented generation (RAG) plus validation via Claude v2 to deliver accurate answers in seconds for complex member queries. | ๐ | - |
| FinGPT | Fine-Tuned Model | Jun 2023 | Open-source financial LLM (FinLLM) using a data-centric approach (based on Llama 2) for automated data curation and efficient adaptation, aiming to democratize AI in finance with applications in robo-advising, algorithmic trading, and low-code development. | ๐ | ๐ |
| Ask FT | Integrated Solution | Mar 2024 | LLM tool by Financial Times (FT) that enables subscribers to query and receive responses derived from two decades of published FT content. | ๐ | - |
| BloombergGPT | Fine-Tuned Model | Mar 2023 | 50-billion-parameter LLM specifically designed for financial applications and the industry's unique terminology, trained on a 363-billion-token dataset sourced from Bloombergโs proprietary data, complemented with 345 billion tokens from general-purpose datasets | ๐ | ๐ |
Payments & Transaction Services
| Name | Type | Date | Description | Site | Paper |
|---|---|---|---|---|---|
| Airwallex AgentOS | Integrated Solution | Jul 2026 | Airwallex launched AgentOS (early access), letting finance teams operate their Airwallex accounts through AI agents like Claude and Cursor via the Airwallex CLI or MCP server, collapsing multi-step, multi-platform workflows into single natural-language requests using pre-packaged Skills. Part of an AI-native push alongside a $320M Series H at an $11B valuation. | ๐ | - |
| Visa & OpenAI | Enterprise Platform | Jun 2026 | Visa partnered with OpenAI to embed secure Visa payments into ChatGPT and OpenAI agent experiences, letting AI agents shop and pay on a user's behalf at Visa-accepting merchants. Built on Visa Intelligent Commerce with tokenized credentials, real-time authorization, fraud monitoring and user spend controls. Announced at the Visa Payments Forum in San Francisco. | ๐ | - |
| Mastercard Agent Pay for Machines | Enterprise Platform | Jun 2026 | Mastercard launched Agent Pay for Machines (AP4M), extending Agent Pay so AI agents transact autonomously for high-frequency, always-on machine-to-machine payments. Provides agent credentialing, programmatically enforced spend limits and guaranteed multi-rail settlement across cards, accounts and stablecoins, with a cryptographic "Verifiable Intent" trust layer. Launched with 30+ partners including Adyen, Stripe, Checkout.com and Coinbase. | ๐ | - |
| Klarna Shopping Search (ChatGPT) | Integrated Solution | May 2026 | Klarna launched an AI-powered Shopping Search app inside ChatGPT, powered by its Product Search MCP server, surfacing real-time prices, availability and offers from over 100 million products and 400 million merchant listings across 13 markets within a single conversation before redirecting users to the merchant's site to check out. | ๐ | - |
| Affirm & Google | Integrated Solution | May 2026 | Affirm partnered with Google to embed its buy-now-pay-later options into Google Search, AI Mode, the Gemini app and Google Pay, giving shoppers transparent pay-over-time financingโwith real-time eligibility checks and full cost/schedule disclosureโinside agentic-commerce checkouts. The integration follows the Google- and Shopify-developed Universal Commerce Protocol. Announced May 12, 2026. | ๐ | - |
| Ant International AMP | Enterprise Platform | Apr 2026 | Ant International launched the Agentic Mobile Protocol (AMP), an open-sourced payment framework for AI agents transacting through mobile interfaces, letting LLMs, platforms and merchants connect to 4.4 billion digital wallet users via the Alipay+ ecosystem (40+ wallet partners, 1.8 billion accounts, 150 million merchants). It halves integration steps versus traditional card binding and adds a "Know Your Agent" identity framework, money-back protection against account takeovers, cross-device support (smartphones, smartwatches, AR glasses, in-car systems) and real-time agent-to-agent settlement of payments as small as US$0.000001. | ๐ | - |
| Stripe Minions | Integrated Solution | Feb 2026 | Stripe is deploying โMinions,โ LLM-powered one-shot end-to-end coding agents that turn a Slack prompt into a fully formed, CI-passing pull request autonomously. The agents now generate and merge over 1,300 PRs per week, materially expanding engineering throughput | ๐ | - |
| PayPal & Google Cloud | Integrated Solution | Oct 2025 | PayPal and Google Cloud introduced a joint agentic commerce solution that lets merchants deploy AI-powered shopping experiences on their own digital surfaces, combining Google Cloud's Conversational Commerce Agent (or custom agents built with the Agent Development Kit) with PayPal's payment capabilities while merchants keep control of the agent's tone, look, and customer relationships. Transactions run on open protocols โ Agent2Agent (A2A) and the Agent Payments Protocol (AP2, built on A2A and MCP) โ using cryptographically signed cart and payment mandates for trust and fraud controls. Announced the same day as PayPal's Agent Ready and Store Sync services. | ๐ | - |
| PayPal Agentic Commerce | Integrated Solution | Oct 2025 | PayPal launched agentic commerce services: "Agent Ready" lets PayPal merchants accept payments on AI shopping surfaces such as Perplexity with built-in fraud detection, buyer protection and dispute resolution, and "Store Sync" makes merchant catalogs discoverable to AI agents while keeping them the merchant of record. Agent Ready rolls out in early 2026. | ๐ | - |
| Modern Treasury AI | Integrated Solution | Jun 2025 | Modern Treasury built Modern Treasury AI (chat, task management, monitoring, and an AI agent) for payment operations, using LangChain's LangGraph to create controllable, auditable agents with traceable logic, granular permissioning, and human-in-the-loop approvals. Its first production agent automates complex payments, reconciliation, and ledger workflows, accelerating time-to-market versus traditional process automation. | ๐ | - |
| Mercado Libre & OpenAI | Integrated Solution | Sep 2024 | Mercado Libre uses GPT-4o to power Verdi, an internal development platform that enables 17,000 engineers to rapidly build AI toolsโincluding for its fintech arm, Mercado Pago. Verdi accelerates application development and automates tasks like customer service mediation, supporting financial operations at scale while improving efficiency and consistency across its digital payments ecosystem. | ๐ | - |
| FinQuery & Google | Integrated Solution | Jun 2024 | FinQuery, a fintech company, is using Gemini as a productivity and collaboration tool to help in brainstorming sessions, draft emails 20% faster, manage complex cross-organizational project plans, and aid engineering teams with debugging code and evaluating new monitoring tools. | ๐ | - |
| Discover Financial Services & Google | Enterprise Platform | Apr 2024 | Discover Financial partnered with Google Cloud to utilize LLMs to helps its 10,000 contact center representatives to search and synthesize information across detailed policies and procedures during calls. | ๐ | - |
| Adyen | Integrated Solution | Nov 2023 | Adyen, a publicly-traded financial technology platform, uses LLMs to improve support operations, and enhance efficiency and response times, with smart ticket routing (which assigns tickets based on sentiment and content) and support agent copilot (which help agents answer tickets faster and more accurately) | ๐ | - |
| Stripe & OpenAI | Integrated Solution | Mar 2023 | Stripe integrates OpenAIโs GPT-4 to enhance its payment platform by analyzing business websites for better support, improving developer assistance through technical documentation processing, and detecting fraud by analyzing community interactions. | ๐ | - |
Insurance & Risk Management
| Name | Type | Date | Description | Site | Paper |
|---|---|---|---|---|---|
| Prudential Just Ask | Integrated Solution | May 2026 | Prudential's Individual Life Insurance business introduced Just Ask, a generative AI tool for third-party advisors and distributors that surfaces product information, generates preliminary underwriting quick quotes and locates new-business forms, with book-of-business insights planned. It generates certain underwriting rates in minutes versus several business days; about 85% of first-time users reported it was helpful. | ๐ | - |
| INS-S1 | Fine-Tuned Model | Mar 2026 | Ant Group's industrial-scale insurance LLM family, built via a Verifiable Data Synthesis System and a Progressive SFT-RL curriculum blending verified-reasoning RL (RLVR) and AI feedback (RLAIF) for actuarial reasoning and regulatory compliance. Achieves a record-low 0.6% hallucination rate (HHEM) while outperforming DeepSeek-R1 and Gemini-2.5-Pro. Released with the 39k-sample INSEva benchmark. | - | ๐ |
| Travelers AI Claim Assistant | Integrated Solution | Feb 2026 | Travelers launched an agentic AI Claim Assistant developed with OpenAI, using OpenAI's models and Realtime API for a fully agentic voice service that handles inbound auto damage claim calls end-to-end โ consulting, advising and supporting customers, and escalating to live specialists when needed. Initially for auto damage claims, with expansion to additional lines of business planned. | ๐ | - |
| Humana Agent Assist | Integrated Solution | Feb 2026 | Humana launched Agent Assist, built with Google Cloud's Vertex AI, Gemini and Gemini Enterprise for CX, to support member-service advocates with real-time conversation summaries, benefit details, automated call summaries and compliance guidance under a human-in-the-loop model. It supports 20,000+ advocates handling up to 80 million calls a year; piloted October 2025, with full rollout in 2026. | ๐ | - |
| LLM Claim Automation | Fine-Tuned Model | Feb 2026 | A locally deployed, domain-adapted LLM that generates structured corrective-action recommendations from unstructured claim narratives, fine-tuned from pretrained models with LoRA on millions of historical warranty-claim records. About 80% of outputs are near-identical matches to ground-truth corrective actions, outperforming general-purpose and prompt-based LLMs; positioned as a viable approach for insurance claim processing. | - | ๐ |
| John Hancock Quick Quote | Integrated Solution | Jan 2026 | John Hancock (Manulife's US life insurer) launched Quick Quote, a reasoning-based generative AI underwriting-support tool it bills as among the first quote-generating GenAI engines in US life insurance. It returns non-binding preliminary risk assessments for less-complex cases in ~15 minutes versus about a day (up to age 75, $10M face), designed for 4,000+ requests/month, with 20,000+ cases supported since a late-2024 pilot. No specific model or vendor named. | ๐ | - |
| Oscar Health Oswell | Integrated Solution | Oct 2025 | Oscar Health unveiled Oswell, a member-facing health AI agent powered by OpenAI that explains lab results and medications, checks drug interactions, requests prescription refills, reviews symptoms and helps members prepare questions for doctors. Rolling out in 2026 across 573 counties in 20 states, reaching Oscar's ~2 million members. | ๐ | - |
| SIGNAL IDUNA & Gemini Enterprise | Enterprise Platform | Oct 2025 | German insurer SIGNAL IDUNA is rolling out Google Cloud's Gemini Enterprise to over 10,000 employees and sales partners under its MOMENTUM 2030 strategy, letting staff build autonomous AI agents for their own job functions integrated with CRM, HR, and marketing systems. Its earlier Co SI Health Assistant made information searches 30% faster and cut escalations from 27% to 3%. | ๐ | - |
| Manulife & Azure | Integrated Solution | Apr 2025 | Manulife built a RAG-based call center assistant on Microsoft Azure using Azure AI Search with GPT-3.5 for responses and Llama 3/GPT-4 for validation, reducing query time from 21.9 to 7.3 seconds and boosting service accuracy and efficiency. | - | ๐ |
| Allianz Insurance Copilot | Integrated Solution | Feb 2025 | An internal generative AI system to assist with claims management. Launched in 2024 for auto claims, it leverages cutting-edge LLMs to streamline workflows and automate key tasks. | ๐ | - |
| AllianzGPT | Enterprise Platform | Feb 2025 | Internal chatbot that leverages capabilities from different LLM providers (e.g., OpenAI, DeepSeek) on Azure to support general productivity tasks, as well as specific functional areas like audit, Risk Consulting actuarial and other business areas. | ๐ | - |
| Newfront & Anthropic | Integrated Solution | Dec 2024 | Newfront, an insurance platform serving 20% of U.S. startups with unicorn status, integrated Anthropic's Claude AI to enhance efficiency and client service. Claude automates complex insurance tasks, such as answering detailed benefits questions, reviewing contracts, and processing loss run documents. | ๐ | - |
| Open-Insurance-LLM-Llama3-8B | Fine-Tuned Model | Nov 2024 | Llama 3 (8B model) fine-tuned with LoRA on the InsuranceQA dataset โ a corpus of insurance domain Q&A pairs โ to specialize it for insurance queries and conversations | ๐ | - |
| Zurich Insurance & OpenAI | Integrated Solution | Nov 2024 | Zurich Insurance Group turned to Microsoft Azure OpenAI Service to develop AI applications that lead to more accurate and efficient risk management evaluations, accelerating the underwriting process, reducing turnaround times, and increasing customer satisfaction. | ๐ | - |
| Zurich Insurance & Azure | Integrated Solution | Nov 2024 | Zurich Insurance Group leverages the LLM capabilities in Azure OpenAI Service to enhance its underwriting process by converting unstructured customer dataโsuch as images, emails, and reports in various languagesโinto structured, actionable insights. | ๐ | - |
| EXL Insurance LLM | Fine-Tuned Model | Sep 2024 | Industry-specific LLM that supports critical claims and underwriting-related tasks, such as claims reconciliation, data extraction and interpretation, question-answering, anomaly detection and chronology summarization. EXL utilized NVIDIA NeMo end-to-end platform for the fine-tuning process on 2 billion tokens of private insurance data. | ๐ | - |
| Swiss Re & mea | Integrated Solution | Sep 2024 | Swiss Re partnered with mea (a generative AI-powered global platform) for insurance process automation, such as extracting unstructured data from submission-related documents (e.g., Schedules of Value) and convert it into structured, analysable data. | ๐ | - |
| Bitext Mistral-7B-Insurance | Fine-Tuned Model | Jul 2024 | Fine-tuned version of Mistral-7B-Instruct-v0.2, specifically tailored for the insurance domain. It is optimized to answer questions and assist users with various insurance-related procedures | ๐ | ๐ |
| Five Sigma - Clive | Integrated Solution | Jul 2024 | Insurance tech startup Five Sigma leveraged Google's Gemini models to create an AI claims adjuster engine (Clive) which frees up human claims handlers to focus on areas where a human touch is valuable, like complex decision-making and empathic customer service. | ๐ | - |
| New York Life | Enterprise Platform | Jul 2024 | New York Life, the largest mutual life insurance company in the US, is actively integrating GenAI across various business functions to enhance efficiency e.g., underwriting, customer service, and hiring. | ๐ | - |
| Trumble Insurance Agency & Google | Integrated Solution | May 2024 | Trumble Insurance Agency is using Gemini for Google Workspace to significantly improve its creativity and the value that it delivers to its clients with enhanced efficiency, productivity, and creativity. | ๐ | - |
| AXA Secure GPT | Enterprise Platform | Apr 2024 | AXA developed AXA Secure GPT, a generative AI platform powered by Azure OpenAI Services, to enhance employee productivity while ensuring data security. This platform enables AXA's 140k employees to utilize AI tools within a secure environment, facilitating tasks such as drafting reports, summarizing documents, and generating content. | ๐ | - |
| Allstate Insurance | Integrated Solution | Mar 2024 | Allstate implemented generative AI models, specifically OpenAI's GPT, customized with company-specific terminology, to enhance customer communications and experience. This system generates approximately 50k daily emails for claims representatives, ensuring messages are clear, empathetic, and free from industry jargon. | ๐ | - |
| Groupama & Azure OpenAI | Integrated Solution | Feb 2024 | Groupama, a leading French mutual insurance group, uses Azure OpenAI Service to power a virtual assistant that helps customer managers respond to policyholder inquiries with 80% accuracy, improving efficiency and response quality. It is done through providing pre-written responses to policyholder inquiries, drawing from a comprehensive and secure documentation corpus. | ๐ | - |
| LAQO insurance & Azure OpenAI | Integrated Solution | Nov 2023 | LAQO, Croatia's first fully digital insurer, partnered with Infobip to develop Pavle, a 24/7 AI assistant powered by Azure OpenAI Service. Pavle resolves 30% of customer queries, allowing human agents to focus on complex cases and customer acquisition. | ๐ | - |
| Roots Automation - InsurGPT | Fine-Tuned Model | May 2023 | Roots Automation released an LLM fine-tuned on insurance-specific documents (ACORD forms, First Notice of Loss (FNOL), loss runs) to automate and cut claims processing time significantly | ๐ | - |
agent
agentic
agentic-ai
agentic-workflow
agents
agent-skill
artificial-intelligence
banking
domain-specific-models
finance
finance-llm
finance-llms
financial-services
fine-tuning
generative-ai
insurance
large-language-model
large-language-models
llm
machine-learning
Contributors
kennethleungty
85 commits
Similar repos
Comprehensive Compilation of Real-World LLM & AI Agent Use Cases in Financial Services
140
85 commits
updated Aug 25, 2026
README
Compilation of Real-World LLM & AI Agent Use Cases in Financial Services
Context
Large language models (LLMs) and AI agents are transforming financial services: from streamlining regulatory compliance and accelerating deal analysis to personalizing customer experiences and automating complex workflows across banking, trading, and asset management.
This repository curates real-world LLM and agentic AI implementations in finance, showcasing enterprise platforms, specialized models, and integrated solutions deployed by leading financial institutions worldwide.
Have a great example of LLMs in the financial services sector? Contribute by opening a pull request or submitting an issue.
๐ Table of Contents
- Retail & Commercial Banking
- Covers personal and business banking servicesโchecking and savings accounts, mortgages, SME financing, and credit solutions. Also addresses central banking operations, digital transformation in traditional banks, and AI-driven advisory for both retail and corporate clients.
- Wealth & Asset Management
- Encompasses asset and wealth advisory, mutual funds, private equity, institutional asset management, robo-advisory, and portfolio management for high-net-worth individuals and institutional clients.
- Capital Markets & Trading
- Covers investment banking, securities trading, market infrastructure, equity research, credit analysis, and financial market analytics. Includes broker-dealers, exchanges, and financial data platforms.
- Payments & Transaction Services
- Focuses on digital payments, card networks, cross-border transactions, BNPL, and payment infrastructure. Also highlights AI-based fraud detection, risk assessment, and automated payment processing innovations.
- Insurance & Risk Management
- Addresses life, health, property, and casualty insurance, along with reinsurance, underwriting automation, and actuarial modeling. Features AI-driven claims processing, fraud detection, personalized policy recommendations, and broader risk assessment tools.
Each example can be classified into one of 3 categories based on type of use case:
- Enterprise Platform: Large-scale LLM deployments spanning multiple lines of business or the entire organization. These implementations typically address a broad range of tasksโsuch as customer support, internal operations, and complianceโwithin a unified, centrally managed framework.
- Fine-Tuned Model: Models either pre-trained or fine-tuned on focused datasets (e.g., regulatory filings, financial news) to achieve deep domain expertise. By narrowing the training data to specific tasks and terminology, they deliver higher accuracy and more relevant insights than general-purpose models.
- Integrated Solution: Vendor-provided or off-the-shelf AI solutions deployed for specific workflows and use cases. Financial institutions leverage ready-made integrations, cloud services, and prompt engineering to streamline tasksโsuch as AI-powered chatbots or document processingโwithout extensive internal development.
Retail & Commercial Banking
| Name | Type | Date | Description | Site | Paper |
|---|---|---|---|---|---|
| DBS Agentic AI for Corporate Bankers | Enterprise Platform | Aug 2026 | DBS scaled an agentic AI solution that streamlines credit memo preparation for corporate bankers serving large and mid-sized clients, expanding from a 150-participant pilot to around 1,500 relationship managers and credit risk managers globally. Specialised agents tackle more than 70 tasks; with credit memos and related credit activities taking up to 40% of relationship managers' time, DBS targets a reduction of at least 30%. No model or vendor named. | ๐ | - |
| Westpac Lending AI Agents | Integrated Solution | Aug 2026 | Westpac put five specialist AI agents built on Amazon Bedrock AgentCore into production home loan and credit card assessment. An account scrutiny agent classifies transactions from manual bank statements; others extract payslip data and verify it against lending policy, with bankers retaining oversight. They process 32,000+ payslips and 1.5M+ transactions weekly, saving 12,500 banker hours in their first month โ over 150,000 annualised. | ๐ | - |
| DBS Joy Agentic AI | Enterprise Platform | Jul 2026 | DBS made DBS Joy, its corporate virtual assistant (launched 2018, given GenAI capabilities in 2025), fully agentic for 350,000 corporate and SME clients in Singapore, letting them retrieve and analyse account and transaction data within a single conversation rather than searching the platform manually. No model or vendor named. In H1 2026 chats rose 22% and active users 61%, cutting calls/emails to customer service by 7%. Another 100,000 Hong Kong users follow in September, with DBS Joy and the retail DBS digibot together expected to handle 1M+ chats monthly across 10M customers. | ๐ | - |
| HSBC & Google Cloud | Enterprise Platform | Jun 2026 | HSBC and Google Cloud announced a multi-year AI partnership using Gemini models, the Gemini Enterprise Agent Platform and Google DeepMind. Applications include financial-crime monitoring across the near one billion transactions HSBC screens monthly (intervening twice as fast), a relationship-manager decision assistant cutting prep from hours to minutes, and hyper-personalised wealth advisory. The deal targets more than 200 new AI use cases over two years. | ๐ | - |
| Santander AI-First | Enterprise Platform | Jun 2026 | Santander extended AI access to all 185,000 employees under its "AI-first" strategy, using a multi-provider stack (Microsoft Copilot, OpenAI ChatGPT, Anthropic Claude, Google Gemini, G42). Openbank's models process ~100,000 AML alerts/year (hours to minutes); UK self-service handles ~240,000 calls/year, freeing ~45,000 service-team hours. AI delivered โฌ35M in Q1 2026, on track to exceed โฌ200M in 2026, targeting โฌ1B over 2026-2028. | ๐ | - |
| Revolut AIR | Integrated Solution | Apr 2026 | Revolut launched AIR (AI by Revolut), an in-app AI assistant that lets customers manage money through natural conversation โ answering questions on spending, budgeting and investing, pausing recurring payments, managing card controls, checking live exchange rates, and helping with trip budgets and in-app eSIM purchases. Sensitive actions require biometric approval, and Revolut says personal data is never stored by third-party AI partners or used for model training. Rolling out from April 2026 to its ~13 million UK customers first, with other markets to follow. | ๐ | - |
| PRAGMA (Revolut) | Fine-Tuned Model | Apr 2026 | Revolut's family of encoder-only Transformer foundation models (10M, 100M and 1B parameters) pre-trained with a masked-language-modelling objective on 24 billion banking events (207 billion tokens) from 26 million pseudonymised user records across 111 countries over 25 months of transaction, app, trading and communication data. Frozen embeddings with a lightweight head โ or LoRA fine-tuning โ outperform internal task-specific baselines on credit scoring, external fraud, product recommendation and lifetime-value prediction. | - | ๐ |
| Starling Assistant | Integrated Solution | Mar 2026 | Starling Bank launched Starling Assistant, billed as the UK's first agentic AI financial assistant, built on Google Gemini running inside Starling's own Google Cloud environment. Rolled out to personal current account holders from 20 March 2026 (business and joint accounts to follow), it uses voice and natural language to set up savings transfers, organise bill payments, create budget spaces, analyse spending and guide accessibility support. | ๐ | - |
| NatWest & OpenAI | Enterprise Platform | Feb 2026 | NatWest became the first UK bank to sign a strategic collaboration with OpenAI, embedding OpenAI models (version unspecified) into its Cora customer chatbot and AskArchie employee assistant. Customers can ask natural-language spending questions and report fraud in real time; around 25,000 customers gain the new agentic financial assistant by end of Q1 2026, with voice-to-voice capability piloting later in 2026. | ๐ | - |
| HSBC & Harvey AI | Enterprise Platform | Jan 2026 | HSBC partnered with Harvey AI to deploy AI across its global in-house legal function, streamlining workflows in contract analysis, due diligence, compliance, and litigation. | ๐ | - |
| BNY Mellon - Eliza | Enterprise Platform | Dec 2025 | BNY developed Eliza, an enterprise AI platform originally powered by OpenAI models, enabling 20,000+ employees to build agents across 125+ live use cases in legal, risk, and client management โ cutting legal contract review time 75% with 99% workforce adoption. In December 2025, BNY partnered with Google Cloud to integrate Gemini Enterprise (plus Gemini 3 and Veo 3) into Eliza for agentic market research. | ๐ ๐ | - |
| HSBC & Mistral AI | Enterprise Platform | Dec 2025 | HSBC partnered with Mistral AI in a multi-year strategic collaboration to deploy self-hosted GenAI models across the bank, powering productivity tools for tailored client communications, financial analysis, multilingual translation, and faster development cycles. | ๐ | - |
| Lloyds AI Financial Assistant | Enterprise Platform | Nov 2025 | Lloyds Banking Group unveiled what it calls the UK's first large-scale, multi-feature agentic AI financial assistant, built on its proprietary Generative AI and agentic framework combining curated bank data with LLMs (no external model or vendor named). Launching in the mobile app in early 2026 for 21M+ customers, it offers 24/7 financial coaching with memory, expanding to mortgages, car finance and protection. | ๐ | - |
| Macquarie Bank & Gemini Enterprise | Enterprise Platform | Oct 2025 | Macquarie Bank is rolling out Google Cloud's Gemini Enterprise to all employees โ not just technical teams โ for personal agents (document summarization, research, content drafting) and enterprise agents spanning system-design documentation, code generation, conversation summaries, and customer support. 99% of employees completed generative-AI training, with a target of every employee integrating AI into daily workflows within six months. | ๐ | - |
| Bank of America's AskGPS | Enterprise Platform | Sep 2025 | BofA built AskGPS (Ask Global Payments Solutions), an in-house GenAI assistant trained on 3,200+ internal documents to help employees serving 40,000+ business clients answer complex inquiries. It reduces response times from hours to seconds, saving tens of thousands of employee hours annually. | ๐ | - |
| Wells Fargo & Google Cloud AI | Enterprise Platform | Aug 2025 | Wells Fargo uses Google Agentspace (powered by Gemini models) to deploy agentic AI across its workforce, enabling employees to find and synthesize information faster, automate tasks and workflows, and increase organizational agility. | ๐ | - |
| U.S Bank & Amazon | Integrated Solution | Aug 2025 | U.S. Bank deployed a real-time agent-assist system using Amazon Q in Connect together with Amazon Bedrock and Claude Haiku model to transcribe calls, detect customer intent, and surface relevant knowledge-base recommendations during voice interactions. | ๐ | - |
| Brex Assistant | Integrated Solution | Aug 2025 | Brex built an AI-powered financial assistant using Amazon Bedrock and Claude models to automate expense management, achieving 75% workflow automation and raising compliance rates from 70% to the mid-90s. | ๐ | - |
| PayU | Integrated Solution | Jul 2025 | PayU implemented a secure enterprise-AI assistant using Amazon Bedrock, deploying multi-agent workflows (via RAG and text-to-SQL) to enable employees to query business metrics, HR policies and operations dataโall within a private VPC to satisfy regulatory data-residency and compliance needs. | ๐ | - |
| Intuit TurboTax | Integrated Solution | Jul 2025 | Intuit built Intuit Assist, an LLM-powered AI assistant within TurboTax that helps millions of users understand their tax situations, deductions, and refunds using a hybrid ClaudeโGPT system with RAG, fine-tuning, and strict regulatory compliance | ๐ | - |
| Nubank | Integrated Solution | May 2025 | Nubank built an AI โprivate bankerโ using GPT-4 and agentic frameworks like LangChain and LangGraph to automate customer support and money-transfer workflows. The system handles millions of user queries and transactions autonomouslyโreducing transfer times and improving accuracyโthrough coordinated LLM agents and a production-grade LLMOps platform. | ๐ | - |
| Apoidea Group | Fine-Tuned Model | May 2025 | Apoidea Group, in collaboration with Amazon SageMaker HyperPod and using the LLaMAโFactory framework, built a solution called SuperAcc that is based on a fine-tuned multimodal model Qwen2โVLโ7BโInstruct to enhance visual information extraction from banking documents, boosting table-structure recognition accuracy substantially and reducing processing times from hours to minutes. | ๐ | - |
| Lloyds & Google Cloud | Enterprise Platform | Apr 2025 | Lloyds Banking Group migrated 15 modelling systems (hundreds of models) from on-premise infrastructure to Google Cloud's Vertex AI (now Gemini Enterprise Agent Platform), where 300+ data scientists and AI developers have launched 80+ ML use cases and 18+ GenAI systems into production. An income-verification model cut mortgage-application processing from days to seconds, with zero unplanned platform downtime. | ๐ | - |
| Capital One Chat Concierge | Fine-Tuned Model | Feb 2025 | Chat Concierge, built on a fine-tuned Llama LLM, helps car buyers compare vehicles, explore financing, estimate trade-in values, and schedule test drives. Fine-tuned on Capital One's proprietary data, it also utilized an agentic approach to understand preferences, tailor recommendations, ensure policy compliance, and engage in human-like interactions. | ๐ | - |
| Deutsche Bank & Google | Enterprise Platform | Feb 2025 | Deutsche Bank leverages Google Cloudโs Vertex AI and Gemini LLMs to streamline banking operation. It involves automating document processing for regulatory compliance, enhancing customer support with AI-powered assistants, and accelerating software development for financial services. | ๐ | - |
| CommBiz Gen AI | Integrated Solution | Jan 2025 | Together with AWS, the Commonwealth Bank of Australia (CBA) rolled out a Gen-AI powered messaging service to assist tens of thousands of business customers with inquiries, facilitating quicker payments and efficient transactions. They leveraged Amazon Bedrock Knowledge Bases, Claude 3 and Cohere LLMs, and Amazon OpenSearch as the vector database. | ๐ | - |
| North for Banking | Enterprise Platform | Jan 2025 | RBC and Cohere co-developed and securely deployed an enterprise generative AI (genAI) solution optimized for financial services, building upon Cohere's proprietary foundation models | ๐ | - |
| Amazon Finance | Enterprise Platform | Dec 2024 | Amazon Finance Automation built a generative-AI Q&A chat assistant on Amazon Bedrock to empower analysts to quickly answer customer queries by retrieving and synthesising policy knowledge, accelerating response times and reducing manual research | ๐ | - |
| Banestas & Google | Enterprise Platform | Dec 2024 | Banestes, a Brazilian bank, used Gemini in Google Workspace to streamline work dynamics, such as accelerating credit analysis by simplifying balance sheet reviews and boosting productivity in marketing and legal departments. | ๐ | - |
| Commerzbank & Google | Integrated Solution | Nov 2024 | Commerzbank utilizes Google's Gemini 1.5 Pro, a multimodal large language model, to automate financial advisory workflows by enabling efficient documentation of client interactions and analysis of complex financial data. Its long context allows it to handle lengthy financial documents and multimedia content seamlessly. | ๐ | - |
| OCBC ChatGPT | Integrated Solution | Nov 2024 | OCBC Bank became the first Singapore bank to deploy a generative AI chatbot, OCBC ChatGPT, to all 30,000 employees across 19 countries, aiming to assist with tasks such as writing, research, and ideation. Developed with Microsoft Azure and following a successful six-month trial, the chatbot operates securely on the bank's private cloud to ensure data security. | ๐ | - |
| BBVA & OpenAI | Enterprise Platform | Nov 2024 | Global financial institution BBVA signed an agreement with OpenAI for 3,000 ChatGPT Enterprises licenses, leading to increased productivity and creativity. Staff across various departments have developed over 2.9k specialized GPTs that boost efficiency, spark creativity, and share expert knowledge across their organization of 125,000 e.g., tasks like translating risk-specific terminology and drafting responses to client inquiries. | ๐ | - |
| PennyMac & Google | Enterprise Platform | Oct 2024 | Pennymac Financial Serives, a national home loan lender and servicer, integrates Google's Gemini LLMs across various departments to enhance efficiency and reduce costs while ensuring security and compliance. HR uses it for job descriptions and policy drafting, while the underwriting team uses Gemini to analyze proprietary data, improving regulatory understanding and best practices. | ๐ | - |
| Belfius & Mistral AI | Enterprise Platform | Sep 2024 | Belfius, a leading Northern European digital bank, deployed Mistral Large to power its customer-facing chatbot in Dutch, French, and English, migrating from a scripted bot platform to an AI-powered solution that delivers personalized banking experiences and reduces manual operator workload. | ๐ | - |
| BNP Paribas & Mistral AI | Enterprise Platform | Jul 2024 | BNP Paribas has partnered with Mistral AI to integrate LLMs across various sectors, including customer support, sales, and IT. This collaboration enables the bank to deploy advanced AI models on-premises, ensuring compliance with regulatory standards. | ๐ | - |
| DBS GPT | Enterprise Platform | Jul 2024 | DBS Bank developed an in-house GenAI co-pilot, the CSO Assistant, to support 500 customer service officers in Singapore. It delivers real-time call transcription, solutioning, and auto-filling of service requestsโachieving near 100% accuracy and reducing call handling time by up to 20%. | ๐ | - |
| Bitext Mistral-7b-Banking | Fine-Tuned Model | Jun 2024 | Fine-tuned version of the Mistral-7B-Instruct-v0.2, specifically tailored for the banking domain. It is optimized to answer questions and assist users with various banking transactions | ๐ | ๐ |
| Scotiabank & Google | Integrated Solution | May 2024 | Scotiabank uses LLMs via Google Cloud to enhance customer service and automate document processing. It features LLM-powered chatbot summarization for seamless handoffs and faster resolutions, plus enhanced Q&A and search for streamlined employee workflows. | ๐ | - |
| ING Bank | Integrated Solution | Feb 2024 | ING implemented a generative AI-powered customer-facing chatbot to enhance customer service efficiency. This chatbot utilizes a multi-step process that retrieves information from data stores, ranks potential answers by relevance, and applies strict guardrails to ensure accurate and appropriate responses. | ๐ | - |
| Citi & GitHub Copilot | Integrated Solution | Feb 2024 | Citi has focused on improving developer productivity by embracing LLM-based coding assistants by rolling out GitHub Copilot (powered by OpenAIโs Codex) to all of its 40,000 software developers enterprise-wideโ. It serves as an AI pair-programmer, suggesting code snippets, functions, or fixes inside developersโ code editors. | ๐ | - |
| Akbank & Azure OpenAI | Integrated Solution | Jan 2024 | Akbank, one of Tรผrkiye's largest banks, uses Azure OpenAI Service to power an AI chatbot that automates customer support, improving accuracy to 90% and cutting response times by three minutes per interaction. This enhances efficiency and allows agents to focus on proactive service. | ๐ | - |
| Ally Financial - Ally.ai | Integrated Solution | Dec 2023 | Ally Financial, the largest digital-only bank in the US and a leading auto lender, launched Azure OpenAI LLM-powered Ally.ai to more than 700 customer care associates in summarizing conversations between them and Ally customers. This automation of post-call documentation for customer service associates is done through LLM summarization of customer call transcripts. | ๐ | - |
| Westpac & KAI-GPT | Fine-Tuned Model | Jun 2023 | Australian lender Westpac is using KAI-GPT, a banking industry-specific LLM, to help bankers locate, interpret and understand information from policies, regulatory filings, procedures, and complex financial products. KAI-GPT is based on Pythia-Chat-Base-7B, fine-tuned on banking-related dataset comprising 24k question-answer pairs from Common Crawl, 18k questions from Kasisto's own conversational data, and 245 million words from 44k banking-related documents | ๐ | ๐ |
| XuanYuan 2.0 | Fine-Tuned Model | May 2023 | Chat model (built upon the BLOOM-176B architecture) trained by combining general-domain with domain-specific knowledge and integrating the stages of pre-training and fine-tuning, It is capable of providing accurate and contextually appropriate responses in the Chinese financial domain. | - | ๐ |
| BBT-FinT5 | Fine-Tuned Model | Feb 2023 | Chinese financial pre-training language model (1B parameters) based on the T5 model, and pre-trained on the 300Gb financial corpus called FinCorpus | - | ๐ |
Wealth & Asset Management
| Name | Type | Date | Description | Site | Paper |
|---|---|---|---|---|---|
| OCBC WoW | Integrated Solution | Jul 2026 | OCBC launched WoW (Whole-of-Wealth), billed as the region's first AI-native banking app, using two GenAI avatars (Wendy and Wayne) to deliver hyper-personalised, 24/7 wealth adviceโportfolio analysis, market news, and investment ideasโvia voice or text to Premier Private Client customers. A related agentic-AI source-of-wealth (SOW) tool cut SOW report preparation from 10 days to one hour. | ๐ | - |
| Morgan Stanley at Work (MCP) | Enterprise Platform | Jun 2026 | Morgan Stanley at Work is opening its Shareworks and Equity Edge stock-plan administration platforms to external AI agents via the Model Context Protocol (MCP), letting clients' agents pull equity-compensation data and insights directly in an agent-to-system manner. The platforms hold roughly $7.35 trillion in assets across about 3,400 administration clients, including nearly half the S&P 500; full rollout is targeted for 2027. | ๐ | - |
| Charles Schwab Portfolio Insights | Integrated Solution | May 2026 | Schwab launched Portfolio Insights, its first generative-AI capability for retail investors (no LLM vendor named). It produces narrative daily performance summaries for the up-to-five holdings with the largest moves, recaps market and industry news on clients' S&P 500 positions, and adds Schwab Center for Financial Research commentary. Rollout to eligible self-directed U.S. clients began May 2026. | ๐ | - |
| Citi Sky | Enterprise Platform | Apr 2026 | Citi Wealth unveiled Citi Sky, an always-on conversational AI "team member" built on Google Cloud's Gemini Enterprise Agent platform, Google DeepMind's real-time avatar technology, and the Gemini Live API. It delivers Chief Investment Office market insights and financial guidance (e.g., CD-maturity alerts) via audio/video in English and Spanish, with a phased rollout to U.S. Citigold clients from summer 2026. | ๐ | - |
| Vanguard Expert Insights | Integrated Solution | Apr 2026 | Vanguard launched Expert Insights, a generative-AI portfolio-analysis tool (no model or vendor disclosed) that lets financial advisors generate instant, methodology-aligned, client-ready guidance and pressure-test portfolios. Currently piloting with select advisors, it will be embedded on-demand in Vanguard's Portfolio Analytics Tool later in 2026. Vanguard's annual advisor portfolio-analysis engagements have roughly quadrupled over the past six years. | ๐ | - |
| Morningstar Advisory AI Assistant | Integrated Solution | Mar 2026 | Morningstar embedded a new AI assistant in its Direct Advisory Suite (distinct from 2023's "Mo" chatbot), grounded in Morningstar's independent data and research. Advisors use natural-language requests to surface meaningful rating changes, prepare meeting briefs, and turn client statements into actionable proposals; it is also reachable through other AI tools via Morningstar's MCP connections. Initial U.S. rollout, with broader U.S./Canada availability through 2026. | ๐ | - |
| Franklin Templeton & Wand AI | Enterprise Platform | Nov 2025 | Franklin Templeton (~$1.66 trillion AUM) formed a multi-year partnership with Wand AI to deploy agentic AI in live production across investment research, operations, and transformation initiatives, moving beyond pilots to enterprise-wide automation. Underlying LLMs were not disclosed; wide-scale rollout is targeted for 2026. | ๐ | - |
| Ardian GAIA | Enterprise Platform | Sep 2025 | Ardian developed GAIA, an internal GenAI platform (powered by Azure OpenAI and Mistral AI) that transforms how investment teams analyze and extract insights from complex documents, enabling 500+ weekly active users to analyze due diligence documents, financial reports, and legal filings. The platform has processed 18,000+ documents and generated 280,000+ AI-powered queries, achieving 2X growth in documents processed while accelerating deal execution and freeing teams to focus on strategic analysis. | ๐ | - |
| Fidelity International Freya | Integrated Solution | Sep 2025 | Fidelity International launched Freya, a conversational LLM-based assistant on its Personal Investing platform (500,000 customers), answering natural-language queries with reasoning-based responses and built-in guardrails that stop short of regulated advice. The underlying model/vendor is unnamed. A phased rollout will expand Freya to Workplace Investing and Fidelity Adviser Solutions. | ๐ | - |
| AlphaAgents (BlackRock) | - | Aug 2025 | BlackRock developed a role-based multi-agent system powered by GPT-4o to enhance wealth management in equity portfolio constructions, enabling LLM agents to collaborate autonomously on equity research, stock selection, and portfolio optimization. | - | ๐ |
| Nippon India Mutual Fund | Integrated Solution | Jul 2025 | Nippon Life India Asset Management Ltd. (Nippon India Mutual Fund) used Amazon Bedrock Knowledge Bases to deploy a RAG AI assistant, leveraging Claude 3 Sonnet on Amazon Bedrock to reformulate queries and rerank responses. The solution increased answer accuracy by over 95% and reduced hallucinations by >90%, cutting report generation time from 2 days to about 10 minutes. | ๐ | - |
| Alpian & Google Cloud | Integrated Solution | Jun 2025 | Alpian, Switzerland's first fully cloud-native private bank, built a Gemini-powered platform engineering tool on Google Cloud (Vertex AI, BigQuery) that lets developers deploy services and interact with infrastructure through a conversational interface within a deliberately bounded scope that preserves security and compliance guardrails. It cut deployment times from days to hours, sped up feature deployment by 25%, and automates compliance workflows aligned with Swiss FINMA standards. | ๐ | - |
| BlackRock Aladdin | Integrated Solution | Jun 2025 | BlackRockโs Aladdin Copilot uses generative AI to let investors query portfolio and risk data in plain language, surfacing precise, contextual insights. It enhances productivity, decision-making, and personalization across the Aladdin platform while meeting strict compliance and accuracy standards. | ๐ | - |
| FinE5 | Fine-Tuned Model | Feb 2025 | Fin-E5, the finance-adapted embedding model in FinMTEB, is built on the E5 model and trained using a persona-based data synthesis method to enhance performance across financial embedding tasks. | - | ๐ |
| PIMCO & Azure | Enterprise Platform | Feb 2025 | PIMCO developed ChatGWM, an AI-powered search tool built on Azure AI, to enhance client service by streamlining information retrieval for its associates. ChatGWM utilizes retrieval-augmented generation (RAG) to search across approved structured and unstructured data sources, and then process them with Azure OpenAI LLM to provide accurate and up-to-date information swiftly. | ๐ | - |
| Aditya Birla Capital | Integrated Solution | Jan 2025 | Financial services provider Aditya Birla Capital implemented SimpliFi, a generative AI chatbot built on Azure OpenAI Service, to enhance customer engagement and streamline financial services. SimpliFi assists users in navigating financial solutions independently, aligning with the preferences of their 25 to 35-year-old demographic. | ๐ | - |
| Touchstone-GPT | Fine-Tuned Model | Nov 2024 | Open-source financial LLM trained through continual pre-training and financial instruction tuning, which demonstrates strong performance on the financial bilingual (English and Chinese) Golden Touchstone benchmark. | ๐ | ๐ |
| Banca Investis | Integrated Solution | Nov 2024 | Banca Investis launched NIWA, a GenAI-powered investment advisory platform designed to enhance customer engagement through hyper-personalized financial services. Serving as a "digital junior banker," NIWA analyzes over 500 pieces of information dailyโincluding clients' financial assets, preferences, and market researchโto provide tailored financial insights and real-time responses to investment-related inquiries. | ๐ | - |
| FinTral | Fine-Tuned Model | Aug 2024 | Suite of multimodal LLMs built upon the Mistral-7b model and tailored for financial analysis. FinTral integrates textual, numerical, tabular, and image data, and is pretrained on a 20 billion token, high quality dataset | - | ๐ |
| Goldman Sachs & Meta | Enterprise Platform | Aug 2024 | Goldman Sachs introduced the GS AI Platform, a GenAI tool designed to enhance employee productivity across various use cases. Built on Meta's LLaMa models, the assistant aids in tasks such as summarizing and proofreading emails, extracting information from documents, as well as translating code between programming languages. | ๐ | - |
| Nomura & Meta | Enterprise Platform | Aug 2024 | Leading Japanese financial institution Nomura uses Meta's LLaMa models on Amazon Bedrock to democratize generative AI by driving faster innovation, transparency, bias guardrails, and robust performance across text summarization, code generation, log analysis, and document processing. | ๐ | - |
| JPMorgan Chase IndexGPT | Enterprise Platform | Jul 2024 | JPMorgan Chase launched a generative AI-based tool (via AWS Bedrock) called IndexGPT, designed to serve as a 'research analyst' for over 50,000 employees, aiding in various tasks that enhance productivity and decision-making within the firm. It is able to generate and refine written documents, provide creative solutions and summarize extensive documents. | ๐ | - |
| IDEA-FinQA | Fine-Tuned Model | Jun 2024 | Financial question-answering system based on Qwen1.5-14B-Chat, utilizing real-time knowledge injection and supporting various data collection and querying methodologies, and comprises three main modules: the data collector, the data querying module, and LLM-based agents tasked with specific functions. | ๐ | ๐ |
| BCI & Azure | Enterprise Platform | Mar 2024 | British Columbia Investment Management Corporation (BCI) integrated Microsoft 365 Copilot and Azure OpenAI Service to enhance productivity and streamline operations. This implementation has led to a 10%-20% productivity boost for 84% of initial users, saving over 2,300 person-hours through automation. Notably, the time required to write internal audit reports decreased by 30%. | ๐ | - |
| RAVEN | Fine-Tuned Model | Jan 2024 | Fine-tuned LLaMA-2 13B Chat model designed to enhance financial data analysis by integrating external tools. Used supervised fine-tuning with parameter-efficient techniques, utilizing a diverse set of financial question-answering datasets, including TAT-QA, Financial PhraseBank, WikiSQL, and OTT-QA | - | ๐ |
| InvestLM | Fine-Tuned Model | Sep 2023 | Financial domain LLM tuned on LLaMA-65B, using a carefully curated instruction dataset related to financial investment. The small yet diverse instruction dataset covers a wide range of financial related topics, from Chartered Financial Analyst (CFA) exam questions to SEC filings to Stackexchange quantitative finance discussions. | ๐ | ๐ |
| CFGPT | Fine-Tuned Model | Sep 2023 | Financial LLM based on InternLM-7B that is designed to handle financial texts effectively. It was pre-trained on 584 million documents (141 billion tokens) from Chinese financial sources like announcements, research reports, social media content, and financial news, and then fine-tuned on 1.5 million instruction pairs (1.5 billion tokens) tailored for specific tasks of financial analysis and decision-making. | ๐ | ๐ |
| FinMA | Fine-Tuned Model | Jun 2023 | Comprehensive framework that introduces FinMA (Financial Multi-task Assistant), an open-source financial LLM fine-tuned (7B and 30B versions) from LLaMA using a diverse, multi-task instruction dataset of 136,000 samples. The dataset encompasses various financial tasks, document types, and data modalities. | ๐ | ๐ |
| Fin-Llama | Fine-Tuned Model | Jun 2023 | Specialized version of LLaMA 33B, fine-tuned (with QLoRA and 4-bit quantization) for financial applications using a 16.9k instruction dataset. | ๐ | - |
| Morningstar - Mo chatbot | Integrated Solution | May 2023 | Morningstar introduced Mo, an AI chatbot powered by the Morningstar Intelligence Engine, which combines Morningstar's extensive investment research library with Microsoft's Azure OpenAI Service. Mo is designed to provide investors and financial professionals with concise, conversational insights by processing natural language queries and summarizing relevant information from over 750,000 investment options. | ๐ | - |
| Cornucopia-LLaMA-Fin-Chinese | Fine-Tuned Model | Apr 2023 | Open-source LLaMA-based model fine-tuned for Chinese financial applications. It uses instruction tuning with Chinese financial Q&A datasets to enhance domain-specific performance. | ๐ | - |
| Morgan Stanley & OpenAI | Pre-trained | Mar 2023 | Morgan Stanley Wealth Management announced a partnership with OpenAI to develop an internal-facing GPT-powered assistant (AI @ Morgan Stanley Assistant), allowing financial advisors to query the bankโs vast research repository and internal knowledge base in natural language | ๐ | ๐ |
| FLANG-ELECTRA | Fine-Tuned Model | Oct 2022 | Domain specific Financial LANGuage model (FLANG) which uses financial keywords and phrases for better masking, and built on the ELECTRA-base architecture. Note: Considered a smaller LM as it has fewer than 1B params | ๐ | ๐ |
| FinBERT-21 | Fine-Tuned Model | Jul 2020 | FinBERT (BERT for Financial Text Mining) is a domain specific language model pre-trained on large-scale financial corpora, allowing it to capture language knowledge and semantic information from the finance domain. Note: Considered a smaller LM as it has fewer than 1B params | - | ๐ |
Capital Markets & Trading
| Name | Type | Date | Description | Site | Paper |
|---|---|---|---|---|---|
| S&P Global Energy HorizonsAgents | Enterprise Platform | May 2026 | S&P Global Energy launched HorizonsAgents, a suite of AI agents converting complex energy and sustainability data into decision-ready, auditable insights for banks, investors, project developers, and corporate finance teams. Four agents โ Transition Finance (beta since November 2025), Data Center Intelligence, Sustainability Benchmarking, and Net Zero Investment โ deliver comparable outputs in minutes, aligned with ISSB frameworks. No underlying models named. | ๐ | - |
| Robinhood Agentic Trading | Integrated Solution | May 2026 | Robinhood launched Agentic Trading (equities in beta), letting customers connect their own third-party AI agents via Robinhood's AI-native MCP servers to a dedicated, separately funded account. Connected agents can analyze concentration/sector exposure, build and backtest thematic strategies, and autonomously execute trades, with safeguards including trade previews, push notifications, fraud review, and instant disconnect. The announcement names no specific LLM. | ๐ | - |
| Obin AI & Google Cloud | Enterprise Platform | Apr 2026 | Obin AI, an enterprise agentic-AI company founded by Google AI, JPMorgan, and Silver Lake veterans, partnered with Google Cloud, integrating Gemini and the Agent Development Kit and listing on Google Cloud Marketplace. Its Obin Financial Agent helps asset management, commercial lending, and insurance teams accelerate complex financial analyses; clients include institutions managing over $1 trillion in assets. | ๐ | - |
| QRAFTI | - | Apr 2026 | A multi-agent agentic framework for empirical quantitative-finance research that exposes a panel-data toolkit plus data access and factor construction via MCP servers as callable tools. Uses chained tool calls with reflection-based planning to replicate established factors, formulate and test new signals, and generate standardized research reports with narrative analysis. | - | ๐ |
| FactSet AI for Banking | Enterprise Platform | Mar 2026 | FactSet launched FactSet AI for Banking (alpha), an AI-native workflow-automation platform built with (and backed by an investment in) Finster AI for investment banking and sell-side teams. It automates creation of transaction artifacts across the deal cycle, from origination through execution, within a single governed environment with deep Microsoft Office integration, drawing on leading AI models (specific vendors not named). | ๐ | - |
| Bloomberg ASKB | Enterprise Platform | Feb 2026 | Bloomberg launched ASKB (beta), an agentic conversational interface for the Terminal running a coordinated network of parallel agents built on a mix of commercial and open-weight LLMs. It answers natural-language queries, auto-generates Bloomberg Query Language, and runs multi-step workflows (earnings prep, post-event analysis), synthesizing Bloomberg data, news, filings, and analytics plus research from 800+ providers with source attribution. Distinct from BloombergGPT. | ๐ | - |
| AlphaSense Generative Search | Enterprise Platform | Jan 2026 | AlphaSense rebuilt Generative Search as a multi-agent research system with dynamic planning and reasoning over hundreds of millions of premium documents plus structured financial data, with pre-built workflow agents that automate earnings analysis, market scans, and deliverable creation (slides and reports) for banking, PE/VC, and IR teams. A July 2026 update added native PowerPoint and Excel assistants; Blue Deer Capital reports building board-ready decks 60-70% faster. | ๐ | - |
| S&P Global & Google Cloud | Enterprise Platform | Dec 2025 | S&P Global and Google Cloud struck a multi-year partnership headlined by the S&P Global Data Retrieval Agent, built by S&P's Kensho AI unit, which lets customers query S&P Global data directly inside Gemini Enterprise with citation-backed answers. The deal also unifies S&P's proprietary datasets on BigQuery for AI-ready distribution and deploys Gemini Enterprise internally for workforce productivity. | ๐ | - |
| FinRpt | Fine-Tuned Model | Nov 2025 | First comprehensive framework for automating Equity Research Report generation using LLMs, featuring a dataset integrating 7 financial data types, an 11-metric evaluation system, and a multi-agent framework with SFT and RL training. AAAI 2026. | ๐ | ๐ |
| LLM Pro Finance Suite | Fine-Tuned Model | Nov 2025 | A suite of five instruction-tuned multilingual finance LLMs (8Bโ70B) built on Llama, Qwen and Gemma bases via continual pre-training plus SFT on ~2M samples (54% finance; English, French, German). Targets financial reporting analysis, risk, compliance, sentiment, translation and RAG, preserving general capability. Open 8B checkpoints published on Hugging Face under the OVHaiLLM org. | ๐ | ๐ |
| S&P Capital IQ Pro ChatIQ | Enterprise Platform | Oct 2025 | S&P Global upgraded Capital IQ Pro with Multi-Document ChatIQ and Document Intelligence 2.0 (ChatIQ built with S&P's Kensho AI unit), letting banking and buyside analysts query and compare across filings, transcripts, investor presentations, news, and research spanning multiple companies or industries, returning generative-AI insights with precise citations for full auditability. No adoption metrics disclosed. | ๐ | - |
| Coinvisor | Fine-Tuned Model | Oct 2025 | An RL-enhanced chatbot agent for interactive cryptocurrency investment analysis, using a reinforcement-learning-based tool-selection mechanism for multi-step planning and real-time integration of diverse crypto data sources. Improves tool-orchestration recall by 40.7% and F1 by 26.6% over baselines, with a 20-investor user study rating average satisfaction 4.64/5. | - | ๐ |
| Deutsche Bank DB Lumina | Enterprise Platform | Sep 2025 | Deutsche Bank Research built DB Lumina, an AI-powered research assistant developed with Google Cloud Consulting, combining Gemini models with a RAG architecture (Gemini Embeddings, Vertex AI Vector Search, pgvector) that grounds responses in internal research and sources like SEC filings with inline citations. In production since September 2024 with ~5,000 users (targeting 10,000+), it saves analysts 30โ45 minutes on earnings note templates and up to 2 hours on research reports. | ๐ | - |
| Trading-R1 | Fine-Tuned Model | Sep 2025 | Tauric Research's 4B financial trading model built on a Qwen3-4B backbone, trained with a multi-stage easy-to-hard SFT-plus-RL curriculum on Tauric-TR1-DB (100k samples over 18 months, 14 equities/ETFs, five data sources). Generates structured, evidence-grounded investment theses with volatility-adjusted decisions, delivering improved risk-adjusted returns and lower drawdowns versus open and proprietary baselines. | ๐ | ๐ |
| Moody's AI Studio | Integrated Solution | Aug 2025 | Moody's built AI Studio, a multi-agent AI platform that automates financial workflows like credit memo generation, cutting analyst time from 40 hours to 2โ3 minutes by orchestrating specialized agents using proprietary and third-party data. The platform has been commercialized for financial clients and adopted internally by 40,000 employees, driving large-scale efficiency and competitive advantage. | ๐ | - |
| Goldman Sachs & Cognition Devin | Integrated Solution | Jul 2025 | Goldman Sachs began piloting Cognition's Devin, an autonomous AI software engineer, across its 12,000-plus developersโinitially hundreds of instances, potentially scaling to thousandsโfor tasks such as coding, testing, and debugging under a human-supervised "hybrid workforce" model. CIO Marco Argenti estimates three-to-four-times developer productivity gains. Distinct from Goldman's Meta-based GS AI Platform. | ๐ | - |
| IG Group & Anthropic | Integrated Solution | Apr 2025 | IG Group partnered with Anthropic to deploy Claude for Work, doubling productivity in financial content creation, cutting over 70 hours of manual work weekly, and improving compliance and analytics workflowsโall with ROI achieved in under three months. | ๐ | - |
| Hebbia & OpenAI | Integrated Solution | Mar 2025 | Hebbia and OpenAI have partnered to enhance financial institutions' efficiency by integrating OpenAI's advanced language modelsโo1, GPT-4o, and o3-miniโinto Hebbia's Matrix platform. This multi-agent system automates complex tasks such as drafting investment committee memos, extracting key terms from credit agreements, and identifying red flags in legal documents, achieving up to 92% accuracy and reducing research time by up to 85%. | ๐ | - |
| Endex & OpenAI | Integrated Solution | Feb 2025 | Endex, an AI platform for financial firms, uses OpenAIโs advanced modelsโGPT-4o, o1-mini, and o3-miniโto power autonomous agents that retrieve, synthesize, and analyze complex financial data, helping firms automate tasks like earnings summaries, investment memos, and due diligence with greater speed and accuracy. | ๐ | - |
| FinBLOOM | Fine-Tuned Model | Feb 2025 | FinBloom 7B, built on Bloom 7B, was trained on 14M financial news articles and 12M SEC filings, then fine-tuned with 50K financial queries for enhanced real-time data retrieval. This process ensures strong contextual understanding for financial decision-making. | - | ๐ |
| MUFG Bank | Integrated Solution | Feb 2025 | MUFG Bank leveraged LLMs to automate data extraction and summarization from corporate reports, enabling faster financial analysis for FX & Derivative Sales. The system, leveraging retrieval-augmented generation (RAG) and fine-tuned prompts, reduces client presentation preparation from hours to minutes, thereby improving client advisory efficiency. | ๐ | - |
| Rogo & OpenAI | Fine-Tuned Model | Feb 2025 | Rogo's fine-tuned OpenAI models and integration of extensive financial datasets (including S&P Global, Crunchbase, and FactSet) allows it to scale financial analysis and deliver real-time financial intelligence to >5k financial professionals, shifting their focus from manual work to high-value decision making. It uses GPT-4o for in-depth financial analysis, o1-mini for contextualizing financial data, and o1 for evaluations and advanced reasoning. | ๐ | - |
| TigerGPT | Integrated Solution | Feb 2025 | Tiger Brokers integrated DeepSeek's AI model, DeepSeek-R1, into its AI-powered chatbot, TigerGPT. This adoption aims to enhance market analysis and trading capabilities for its customers through the improved logical reasoning capabilities. | ๐ | - |
| London Stock Exchange Group & Amazon Q | Integrated Solution | Jan 2025 | London Stock Exchange Groupโs post-trade services team deployed Amazon Q Business (running on Amazon Bedrock) with a front-end custom UI, document crawling of FAQs and rulebooks, and retrieval-augmented generation (RAG) plus validation via Claude v2 to deliver accurate answers in seconds for complex member queries. | ๐ | - |
| FinGPT | Fine-Tuned Model | Jun 2023 | Open-source financial LLM (FinLLM) using a data-centric approach (based on Llama 2) for automated data curation and efficient adaptation, aiming to democratize AI in finance with applications in robo-advising, algorithmic trading, and low-code development. | ๐ | ๐ |
| Ask FT | Integrated Solution | Mar 2024 | LLM tool by Financial Times (FT) that enables subscribers to query and receive responses derived from two decades of published FT content. | ๐ | - |
| BloombergGPT | Fine-Tuned Model | Mar 2023 | 50-billion-parameter LLM specifically designed for financial applications and the industry's unique terminology, trained on a 363-billion-token dataset sourced from Bloombergโs proprietary data, complemented with 345 billion tokens from general-purpose datasets | ๐ | ๐ |
Payments & Transaction Services
| Name | Type | Date | Description | Site | Paper |
|---|---|---|---|---|---|
| Airwallex AgentOS | Integrated Solution | Jul 2026 | Airwallex launched AgentOS (early access), letting finance teams operate their Airwallex accounts through AI agents like Claude and Cursor via the Airwallex CLI or MCP server, collapsing multi-step, multi-platform workflows into single natural-language requests using pre-packaged Skills. Part of an AI-native push alongside a $320M Series H at an $11B valuation. | ๐ | - |
| Visa & OpenAI | Enterprise Platform | Jun 2026 | Visa partnered with OpenAI to embed secure Visa payments into ChatGPT and OpenAI agent experiences, letting AI agents shop and pay on a user's behalf at Visa-accepting merchants. Built on Visa Intelligent Commerce with tokenized credentials, real-time authorization, fraud monitoring and user spend controls. Announced at the Visa Payments Forum in San Francisco. | ๐ | - |
| Mastercard Agent Pay for Machines | Enterprise Platform | Jun 2026 | Mastercard launched Agent Pay for Machines (AP4M), extending Agent Pay so AI agents transact autonomously for high-frequency, always-on machine-to-machine payments. Provides agent credentialing, programmatically enforced spend limits and guaranteed multi-rail settlement across cards, accounts and stablecoins, with a cryptographic "Verifiable Intent" trust layer. Launched with 30+ partners including Adyen, Stripe, Checkout.com and Coinbase. | ๐ | - |
| Klarna Shopping Search (ChatGPT) | Integrated Solution | May 2026 | Klarna launched an AI-powered Shopping Search app inside ChatGPT, powered by its Product Search MCP server, surfacing real-time prices, availability and offers from over 100 million products and 400 million merchant listings across 13 markets within a single conversation before redirecting users to the merchant's site to check out. | ๐ | - |
| Affirm & Google | Integrated Solution | May 2026 | Affirm partnered with Google to embed its buy-now-pay-later options into Google Search, AI Mode, the Gemini app and Google Pay, giving shoppers transparent pay-over-time financingโwith real-time eligibility checks and full cost/schedule disclosureโinside agentic-commerce checkouts. The integration follows the Google- and Shopify-developed Universal Commerce Protocol. Announced May 12, 2026. | ๐ | - |
| Ant International AMP | Enterprise Platform | Apr 2026 | Ant International launched the Agentic Mobile Protocol (AMP), an open-sourced payment framework for AI agents transacting through mobile interfaces, letting LLMs, platforms and merchants connect to 4.4 billion digital wallet users via the Alipay+ ecosystem (40+ wallet partners, 1.8 billion accounts, 150 million merchants). It halves integration steps versus traditional card binding and adds a "Know Your Agent" identity framework, money-back protection against account takeovers, cross-device support (smartphones, smartwatches, AR glasses, in-car systems) and real-time agent-to-agent settlement of payments as small as US$0.000001. | ๐ | - |
| Stripe Minions | Integrated Solution | Feb 2026 | Stripe is deploying โMinions,โ LLM-powered one-shot end-to-end coding agents that turn a Slack prompt into a fully formed, CI-passing pull request autonomously. The agents now generate and merge over 1,300 PRs per week, materially expanding engineering throughput | ๐ | - |
| PayPal & Google Cloud | Integrated Solution | Oct 2025 | PayPal and Google Cloud introduced a joint agentic commerce solution that lets merchants deploy AI-powered shopping experiences on their own digital surfaces, combining Google Cloud's Conversational Commerce Agent (or custom agents built with the Agent Development Kit) with PayPal's payment capabilities while merchants keep control of the agent's tone, look, and customer relationships. Transactions run on open protocols โ Agent2Agent (A2A) and the Agent Payments Protocol (AP2, built on A2A and MCP) โ using cryptographically signed cart and payment mandates for trust and fraud controls. Announced the same day as PayPal's Agent Ready and Store Sync services. | ๐ | - |
| PayPal Agentic Commerce | Integrated Solution | Oct 2025 | PayPal launched agentic commerce services: "Agent Ready" lets PayPal merchants accept payments on AI shopping surfaces such as Perplexity with built-in fraud detection, buyer protection and dispute resolution, and "Store Sync" makes merchant catalogs discoverable to AI agents while keeping them the merchant of record. Agent Ready rolls out in early 2026. | ๐ | - |
| Modern Treasury AI | Integrated Solution | Jun 2025 | Modern Treasury built Modern Treasury AI (chat, task management, monitoring, and an AI agent) for payment operations, using LangChain's LangGraph to create controllable, auditable agents with traceable logic, granular permissioning, and human-in-the-loop approvals. Its first production agent automates complex payments, reconciliation, and ledger workflows, accelerating time-to-market versus traditional process automation. | ๐ | - |
| Mercado Libre & OpenAI | Integrated Solution | Sep 2024 | Mercado Libre uses GPT-4o to power Verdi, an internal development platform that enables 17,000 engineers to rapidly build AI toolsโincluding for its fintech arm, Mercado Pago. Verdi accelerates application development and automates tasks like customer service mediation, supporting financial operations at scale while improving efficiency and consistency across its digital payments ecosystem. | ๐ | - |
| FinQuery & Google | Integrated Solution | Jun 2024 | FinQuery, a fintech company, is using Gemini as a productivity and collaboration tool to help in brainstorming sessions, draft emails 20% faster, manage complex cross-organizational project plans, and aid engineering teams with debugging code and evaluating new monitoring tools. | ๐ | - |
| Discover Financial Services & Google | Enterprise Platform | Apr 2024 | Discover Financial partnered with Google Cloud to utilize LLMs to helps its 10,000 contact center representatives to search and synthesize information across detailed policies and procedures during calls. | ๐ | - |
| Adyen | Integrated Solution | Nov 2023 | Adyen, a publicly-traded financial technology platform, uses LLMs to improve support operations, and enhance efficiency and response times, with smart ticket routing (which assigns tickets based on sentiment and content) and support agent copilot (which help agents answer tickets faster and more accurately) | ๐ | - |
| Stripe & OpenAI | Integrated Solution | Mar 2023 | Stripe integrates OpenAIโs GPT-4 to enhance its payment platform by analyzing business websites for better support, improving developer assistance through technical documentation processing, and detecting fraud by analyzing community interactions. | ๐ | - |
Insurance & Risk Management
| Name | Type | Date | Description | Site | Paper |
|---|---|---|---|---|---|
| Prudential Just Ask | Integrated Solution | May 2026 | Prudential's Individual Life Insurance business introduced Just Ask, a generative AI tool for third-party advisors and distributors that surfaces product information, generates preliminary underwriting quick quotes and locates new-business forms, with book-of-business insights planned. It generates certain underwriting rates in minutes versus several business days; about 85% of first-time users reported it was helpful. | ๐ | - |
| INS-S1 | Fine-Tuned Model | Mar 2026 | Ant Group's industrial-scale insurance LLM family, built via a Verifiable Data Synthesis System and a Progressive SFT-RL curriculum blending verified-reasoning RL (RLVR) and AI feedback (RLAIF) for actuarial reasoning and regulatory compliance. Achieves a record-low 0.6% hallucination rate (HHEM) while outperforming DeepSeek-R1 and Gemini-2.5-Pro. Released with the 39k-sample INSEva benchmark. | - | ๐ |
| Travelers AI Claim Assistant | Integrated Solution | Feb 2026 | Travelers launched an agentic AI Claim Assistant developed with OpenAI, using OpenAI's models and Realtime API for a fully agentic voice service that handles inbound auto damage claim calls end-to-end โ consulting, advising and supporting customers, and escalating to live specialists when needed. Initially for auto damage claims, with expansion to additional lines of business planned. | ๐ | - |
| Humana Agent Assist | Integrated Solution | Feb 2026 | Humana launched Agent Assist, built with Google Cloud's Vertex AI, Gemini and Gemini Enterprise for CX, to support member-service advocates with real-time conversation summaries, benefit details, automated call summaries and compliance guidance under a human-in-the-loop model. It supports 20,000+ advocates handling up to 80 million calls a year; piloted October 2025, with full rollout in 2026. | ๐ | - |
| LLM Claim Automation | Fine-Tuned Model | Feb 2026 | A locally deployed, domain-adapted LLM that generates structured corrective-action recommendations from unstructured claim narratives, fine-tuned from pretrained models with LoRA on millions of historical warranty-claim records. About 80% of outputs are near-identical matches to ground-truth corrective actions, outperforming general-purpose and prompt-based LLMs; positioned as a viable approach for insurance claim processing. | - | ๐ |
| John Hancock Quick Quote | Integrated Solution | Jan 2026 | John Hancock (Manulife's US life insurer) launched Quick Quote, a reasoning-based generative AI underwriting-support tool it bills as among the first quote-generating GenAI engines in US life insurance. It returns non-binding preliminary risk assessments for less-complex cases in ~15 minutes versus about a day (up to age 75, $10M face), designed for 4,000+ requests/month, with 20,000+ cases supported since a late-2024 pilot. No specific model or vendor named. | ๐ | - |
| Oscar Health Oswell | Integrated Solution | Oct 2025 | Oscar Health unveiled Oswell, a member-facing health AI agent powered by OpenAI that explains lab results and medications, checks drug interactions, requests prescription refills, reviews symptoms and helps members prepare questions for doctors. Rolling out in 2026 across 573 counties in 20 states, reaching Oscar's ~2 million members. | ๐ | - |
| SIGNAL IDUNA & Gemini Enterprise | Enterprise Platform | Oct 2025 | German insurer SIGNAL IDUNA is rolling out Google Cloud's Gemini Enterprise to over 10,000 employees and sales partners under its MOMENTUM 2030 strategy, letting staff build autonomous AI agents for their own job functions integrated with CRM, HR, and marketing systems. Its earlier Co SI Health Assistant made information searches 30% faster and cut escalations from 27% to 3%. | ๐ | - |
| Manulife & Azure | Integrated Solution | Apr 2025 | Manulife built a RAG-based call center assistant on Microsoft Azure using Azure AI Search with GPT-3.5 for responses and Llama 3/GPT-4 for validation, reducing query time from 21.9 to 7.3 seconds and boosting service accuracy and efficiency. | - | ๐ |
| Allianz Insurance Copilot | Integrated Solution | Feb 2025 | An internal generative AI system to assist with claims management. Launched in 2024 for auto claims, it leverages cutting-edge LLMs to streamline workflows and automate key tasks. | ๐ | - |
| AllianzGPT | Enterprise Platform | Feb 2025 | Internal chatbot that leverages capabilities from different LLM providers (e.g., OpenAI, DeepSeek) on Azure to support general productivity tasks, as well as specific functional areas like audit, Risk Consulting actuarial and other business areas. | ๐ | - |
| Newfront & Anthropic | Integrated Solution | Dec 2024 | Newfront, an insurance platform serving 20% of U.S. startups with unicorn status, integrated Anthropic's Claude AI to enhance efficiency and client service. Claude automates complex insurance tasks, such as answering detailed benefits questions, reviewing contracts, and processing loss run documents. | ๐ | - |
| Open-Insurance-LLM-Llama3-8B | Fine-Tuned Model | Nov 2024 | Llama 3 (8B model) fine-tuned with LoRA on the InsuranceQA dataset โ a corpus of insurance domain Q&A pairs โ to specialize it for insurance queries and conversations | ๐ | - |
| Zurich Insurance & OpenAI | Integrated Solution | Nov 2024 | Zurich Insurance Group turned to Microsoft Azure OpenAI Service to develop AI applications that lead to more accurate and efficient risk management evaluations, accelerating the underwriting process, reducing turnaround times, and increasing customer satisfaction. | ๐ | - |
| Zurich Insurance & Azure | Integrated Solution | Nov 2024 | Zurich Insurance Group leverages the LLM capabilities in Azure OpenAI Service to enhance its underwriting process by converting unstructured customer dataโsuch as images, emails, and reports in various languagesโinto structured, actionable insights. | ๐ | - |
| EXL Insurance LLM | Fine-Tuned Model | Sep 2024 | Industry-specific LLM that supports critical claims and underwriting-related tasks, such as claims reconciliation, data extraction and interpretation, question-answering, anomaly detection and chronology summarization. EXL utilized NVIDIA NeMo end-to-end platform for the fine-tuning process on 2 billion tokens of private insurance data. | ๐ | - |
| Swiss Re & mea | Integrated Solution | Sep 2024 | Swiss Re partnered with mea (a generative AI-powered global platform) for insurance process automation, such as extracting unstructured data from submission-related documents (e.g., Schedules of Value) and convert it into structured, analysable data. | ๐ | - |
| Bitext Mistral-7B-Insurance | Fine-Tuned Model | Jul 2024 | Fine-tuned version of Mistral-7B-Instruct-v0.2, specifically tailored for the insurance domain. It is optimized to answer questions and assist users with various insurance-related procedures | ๐ | ๐ |
| Five Sigma - Clive | Integrated Solution | Jul 2024 | Insurance tech startup Five Sigma leveraged Google's Gemini models to create an AI claims adjuster engine (Clive) which frees up human claims handlers to focus on areas where a human touch is valuable, like complex decision-making and empathic customer service. | ๐ | - |
| New York Life | Enterprise Platform | Jul 2024 | New York Life, the largest mutual life insurance company in the US, is actively integrating GenAI across various business functions to enhance efficiency e.g., underwriting, customer service, and hiring. | ๐ | - |
| Trumble Insurance Agency & Google | Integrated Solution | May 2024 | Trumble Insurance Agency is using Gemini for Google Workspace to significantly improve its creativity and the value that it delivers to its clients with enhanced efficiency, productivity, and creativity. | ๐ | - |
| AXA Secure GPT | Enterprise Platform | Apr 2024 | AXA developed AXA Secure GPT, a generative AI platform powered by Azure OpenAI Services, to enhance employee productivity while ensuring data security. This platform enables AXA's 140k employees to utilize AI tools within a secure environment, facilitating tasks such as drafting reports, summarizing documents, and generating content. | ๐ | - |
| Allstate Insurance | Integrated Solution | Mar 2024 | Allstate implemented generative AI models, specifically OpenAI's GPT, customized with company-specific terminology, to enhance customer communications and experience. This system generates approximately 50k daily emails for claims representatives, ensuring messages are clear, empathetic, and free from industry jargon. | ๐ | - |
| Groupama & Azure OpenAI | Integrated Solution | Feb 2024 | Groupama, a leading French mutual insurance group, uses Azure OpenAI Service to power a virtual assistant that helps customer managers respond to policyholder inquiries with 80% accuracy, improving efficiency and response quality. It is done through providing pre-written responses to policyholder inquiries, drawing from a comprehensive and secure documentation corpus. | ๐ | - |
| LAQO insurance & Azure OpenAI | Integrated Solution | Nov 2023 | LAQO, Croatia's first fully digital insurer, partnered with Infobip to develop Pavle, a 24/7 AI assistant powered by Azure OpenAI Service. Pavle resolves 30% of customer queries, allowing human agents to focus on complex cases and customer acquisition. | ๐ | - |
| Roots Automation - InsurGPT | Fine-Tuned Model | May 2023 | Roots Automation released an LLM fine-tuned on insurance-specific documents (ACORD forms, First Notice of Loss (FNOL), loss runs) to automate and cut claims processing time significantly | ๐ | - |
agent
agentic
agentic-ai
agentic-workflow
agents
agent-skill
artificial-intelligence
banking
domain-specific-models
finance
finance-llm
finance-llms
financial-services
fine-tuning
generative-ai
insurance
large-language-model
large-language-models
llm
machine-learning
Similar repos
Contributors
kennethleungty
85 commits