Verified
JPMorgan Chase
Scaling- Banking
- 318,512 employees
- $19.8B tech budget (2026)
- LLM Suite, built in-house
- ~150,000 weekly users
- Evident AI Index #1 (2026)
JPMorgan Chase deploys AI through LLM Suite, an internal portal it built in 2024 that puts OpenAI and Anthropic models behind one interface, alongside coding assistants for its 65,000 technologists and almost 1,000 AI use cases across its businesses. About 150,000 employees used LLM Suite each week in Feb 2026, and the AI program moved into Global Technology in Jul 2026 under CTO Scot Baldry, who succeeds Teresa Heitsenrether as Chief Data & Analytics Officer when she retires at the end of 2026.
jpmorganchase.comTheir approach to AI
- The platform
- LLM Suite, a portal JPMorgan Chase built in 2024 to give staff secure access to outside models, now carries more than 200 internal applications (Lori Beer, Oct 2026).
- The models
- OpenAI and Anthropic models run behind LLM Suite, engineers use GitHub Copilot and Claude Code, and JPMorgan Chase also runs open-weight models in its own data centers (Oct 2026).
- The reach
- About 150,000 JPMorgan Chase employees used LLM Suite each week in Feb 2026; about 250,000, everyone except branch and call center staff, had access in Sep 2025.
- The catch
- JPMorgan Chase no longer puts a dollar figure on AI benefits, and CFO Jeremy Barnum said in Oct 2026 that the gains are not yet visible in quarterly earnings.
Snapshot
- Employees
- 318,512
- Dec 31, 2025, in 66 countries (2025 Form 10-K)
- Technology budget
- $19.8B (2026)
- About $18B in 2025 and $17B in 2024
- AI/ML headcount
- 2,000+
- AI/ML experts and data scientists, last disclosed Apr 2024; about 65,000 technologists in 2026
- Who owns AI
- Scot Baldry, CTO
- Taking over as firmwide CDAO from Teresa Heitsenrether (Jul 2026 memo); reports to Global CIO Lori Beer
- Platform
- LLM Suite, built in-house
- Released summer 2024; updated about every eight weeks
- Models
- OpenAI, Anthropic, open-weight
- Model-agnostic by policy (Jeremy Barnum, Oct 2026)
- Reach
- ~150,000 weekly users
- Feb 2026; about 250,000 had access in Sep 2025
- Use cases
- ~1,000
- Jul 2026; AI solutions in production doubled in 2025
Key findings
About 150,000 JPMorgan Chase employees used LLM Suite every week as of Feb 2026, out of about 250,000 with access in Sep 2025, and Lori Beer said in Oct 2026 that more than 200 internal applications are built on the platform.
JPMorgan Chase budgets about $19.8 billion for technology in 2026, up 10%, while CFO Jeremy Barnum said on Oct 13, 2026 that AI-specific expense ran at something like $700 million a year when Jamie Dimon cited $2 billion in Oct 2025, called that figure stale, and said AI spending is now too diffused across the bank to ring-fence.
Teresa Heitsenrether, who became JPMorgan Chase's first Chief Data & Analytics Officer, a new Operating Committee role, in Jun 2023, retires at the end of 2026, and CTO Scot Baldry is taking over her title and the AI program under a Jul 1, 2026 memo, reporting to Global CIO Lori Beer without an Operating Committee seat.
JPMorgan Chase keeps its models swappable: LLM Suite launched on OpenAI models in 2024, offered Anthropic models by Sep 2025, and Jeremy Barnum said in Oct 2026 that the bank also runs open-weight and open-source models, some in its own data centers, to limit lock-in and cost.
Jamie Dimon said on Jul 14, 2026 that JPMorgan Chase has almost 1,000 AI use cases, about 50 of them important, and that some discrete areas cut jobs by 30% to 40%, with most affected staff offered roles elsewhere.
Where JPMorgan sits
- Stage
- Scaling
JPMorgan Chase sits at Scaling: LLM Suite is open to about 250,000 employees, hundreds of AI use cases run in production (more than 400 by Apr 2024, and the count of AI solutions in production doubled in 2025), and a named executive owns the AI program. JPMorgan Chase falls short of Embedded on two tests: about 150,000 employees used LLM Suite each week in Feb 2026, under half of its 318,512 staff, and the bank stopped putting a dollar figure on AI's effect in Oct 2026.
How the stages are defined - Ownership
- CTO Scot Baldry, under Global CIO Lori Beer; becomes CDAO at the end of 2026
- Platform
- Hybrid: LLM Suite built in-house; coding tools bought
- Models
- OpenAI and Anthropic behind one portal, plus open-weight models
- Hosting
- Own data centers and cloud providers, AWS and Microsoft Azure included; bank data barred from model training
- Reach
- About 150,000 weekly users (Feb 2026)
How they deploy
Build the platform, buy the models
JPMorgan Chase built LLM Suite as a portal to outside large language models instead of training its own frontier model, and launched it with OpenAI's models in 2024. Teresa Heitsenrether said in Aug 2024 that the plan was to move between models by use case and not be beholden to any one provider. Jamie Dimon's Apr 2026 letter says there will be a wide variety of AI models, open and closed, large and small, and that no single tool will dominate.
Model-agnostic to control cost
Jeremy Barnum said on Oct 13, 2026 that JPMorgan Chase's strategy is to minimize lock-in and keep its options open, with no dependence on one model or one agent framework and the ability to run proprietary, open-weight and open-source models, some in its own data centers. Jamie Dimon added that model-routing tools will send each query to the best, quickest and cheapest model. Barnum said in Jul 2026 that token costs were trivial in the first half of 2026 and forecast to accelerate in the second half, and that summarizing an analyst report does not need the newest, most expensive model.
Shared platforms, delivery in the businesses
Jamie Dimon wrote in Apr 2026 that platforms for data, AI and coding must be companywide, while small dedicated teams attack specific problems. The Commercial & Investment Bank says each of its businesses has its own AI strategy tied to the client journey. A memo reported by Business Insider in Jul 2026 said that with the foundations in place and AI delivery shifting to businesses and functions, JPMorgan Chase would align its Chief Data & Analytics Office with Global Technology.
More building in-house
Derek Waldron said in Jun 2026 that JPMorgan Chase now looks harder at whether it can build software capabilities itself instead of buying from vendors, and that the moat around some software companies has diminished. Lori Beer said in Oct 2026 that many of the bank's AI agents are built by her own engineering teams.
Agents next, with guardrails
Derek Waldron told CNBC in Jun 2026 that JPMorgan Chase will deploy long-running autonomous agents in 2026 that work for an hour or two instead of a few minutes. Lori Beer said in Oct 2026 that the largest agent use case today is back-end technology work such as writing code, and that agents get guardrails at a very fine-grained level before they act on their own.
- Deployment model
- JPMorgan Chase runs AI as shared platforms with delivery in the lines of business. Global Technology, under Global CIO Lori Beer, builds and runs LLM Suite, the data and AI platform and the coding tools, and since Jul 2026 also houses the Chief Data & Analytics Office. Each business sets its own use cases with dedicated teams and KPIs; in the Commercial & Investment Bank, data and analytics officers report jointly to COO Guy Halamish and to business heads (Feb 2026). Engineers who use Claude Code are moving into Devspace, a sandbox hosted on AWS that keeps the AI away from staff credentials and internal systems (Sep 2026).
- Model providers
- JPMorgan Chase launched LLM Suite in 2024 with OpenAI models and offered models from both OpenAI and Anthropic by Sep 2025. Jeremy Barnum said in Oct 2026 that the bank also runs open-weight and open-source models, some in its own data centers, alongside proprietary ones. For software engineers, JPMorgan Chase built its own coding assistant (reported Mar 2025) and also uses GitHub Copilot and Anthropic's Claude Code; about 8,000 staff held Claude licenses in Aug 2026, according to an internal dashboard seen by Business Insider. JPMorgan Chase was the only bank among the launch partners of Anthropic's Project Glasswing in Apr 2026, testing Claude Mythos Preview for defensive security.
- Data and infrastructure
- JPMorgan Chase holds about an exabyte of data and runs more than 7,000 applications, according to American Banker in Oct 2026; Jamie Dimon cited 6,000 applications in Feb 2026. About 70% or more of the bank's applications ran their processing largely in public or private cloud by early 2026, up about 5 points in a year, and about 80% or more ran largely on modern infrastructure. JPMorgan Chase has worked with AWS since 2020 under a multi-cloud strategy, and Lori Beer said in Dec 2024 that AWS SageMaker is the foundation of the firmwide AI platform. Teresa Heitsenrether said in Aug 2024 that LLM Suite was built so bank data is not used to train outside models.
Platforms and internal tools
LLM SuiteSummer 2024
LLM Suite is JPMorgan Chase's in-house generative AI portal to outside large language models, used for writing, summarizing, idea generation and, increasingly, for building generative AI into business applications through internal APIs. JPMorgan Chase updates LLM Suite about every eight weeks with new data and connections (Sep 2025), and staff may use it to draft year-end performance reviews but not for pay decisions (Nov 2025). JPMorgan Chase dates the release to summer 2024, while CNBC reported in Sep 2025 that employees first reached OpenAI models through LLM Suite in 2023.
- Users
- About 150,000 weekly (Feb 2026); about 250,000 with access (Sep 2025)
- Models
- OpenAI and Anthropic
Employee AssistantBy Apr 2026
Employee Assistant is a personalized AI agent that gives JPMorgan Chase employees one place to get help and take action across the firm; COO Jennifer Piepszak described it as recently rolled out in the 2025 annual report published in Apr 2026.
AI coding assistants and DevspaceBy Mar 2025
JPMorgan Chase gives engineers an in-house coding assistant that Lori Beer said raised productivity 10% to 20% for tens of thousands of engineers (Mar 2025), plus GitHub Copilot and Anthropic's Claude Code. Claude Code users are moving into Devspace, a containerized sandbox on AWS, and some engineers have a $2,000 monthly Claude spending limit (Sep 2026).
- Users
- 40,000+ engineers with AI coding assistants (May 2025); 90%+ of CIB engineers (Apr 2026)
- Models
- In-house assistant, GitHub Copilot, Anthropic Claude Code
SpectrumIQ and Smart Monitor
SpectrumIQ is Asset & Wealth Management's proprietary suite of AI capabilities inside its Spectrum investment platform, tying together research, data and risk across about 90,000 securities and 22 million documents and taking in about 7,000 broker research reports a day. Smart Monitor, the SpectrumIQ investor assistant, pushes prioritized insights and cut the time from research to insight by 80% (Apr 2026).
Proxy IQJan 2026
Proxy IQ, which the 2025 annual report calls Proxy Voting powered by SpectrumIQ, is J.P. Morgan Asset Management's in-house AI stewardship tool that analyzes data from more than 3,000 annual U.S. company meetings. J.P. Morgan Asset Management began exclusive use for U.S. proxy votes in the first quarter of 2026, completed the switch on Apr 1, 2026, and stopped using external proxy advisers for U.S. voting.
Connect CoachEarly 2025
Connect Coach is an adviser platform with 25 specialized AI agents that helps Private Bank and U.S. Wealth Management advisers prepare for meetings, analyze portfolios and summarize calls, and it pushes about 1 million personalized AI-driven insights to front-office staff (Apr 2026). Reuters reported in May 2025 that private client advisers using JPMorgan's Coach AI tool found information up to 95% faster.
- Users
- 12,000 (Apr 2026)
EVEE Intelligent Q&ABy May 2025
EVEE Intelligent Q&A is a generative AI tool that lets Chase call center specialists ask questions and get concise answers drawn from Chase policies and documentation; Katie Hainsey, then head of AI/ML for digital, marketing and operations, told Tearsheet in May 2025 that EVEE improved call resolution times and employee and customer satisfaction.
IndexGPTMay 2024
IndexGPT builds thematic equity index baskets in which OpenAI's GPT-4 generates keywords for a theme and a separate natural language processing model scans news articles to find companies linked to it.
- Models
- OpenAI GPT-4
Data and AI platform on AWS SageMakerBy Dec 2024
JPMorgan Chase's firmwide data and AI platform for data scientists is built on AWS SageMaker, which Lori Beer said in Dec 2024 was used by about 5,000 employees, with AWS Bedrock under evaluation to give data scientists access to more models that can be fine-tuned on bank data.
- Users
- About 5,000 (Dec 2024)
Vendors and partners
OpenAI2024
Model provider behind LLM Suite at launch; GPT-4 in IndexGPT · OpenAI models powered LLM Suite when it launched in 2024, more than a year after JPMorgan Chase restricted staff use of ChatGPT.
AnthropicBy Sep 2025
Model provider in LLM Suite, Claude Code for engineers, and Project Glasswing launch partner · Anthropic models were in LLM Suite by Sep 2025; about 8,000 JPMorgan Chase staff held Claude licenses in Aug 2026 (Business Insider), and the bank began testing Claude Mythos Preview for defensive security in Apr 2026.
Amazon Web Services2020
Public cloud; SageMaker underpins the firmwide AI platform; hosts Devspace · JPMorgan Chase had nearly 1,000 applications running on AWS in 2023, according to CIO.com, and treats AWS as one part of a multi-cloud strategy.
Microsoft Azure
Cloud under the bank's private model ecosystem · Lori Beer said in Oct 2026 that JPMorgan Chase built its model ecosystem within its private cloud on top of Microsoft Azure (Cybersecurity Dive).
GitHub (Microsoft)
GitHub Copilot coding assistant · An internal JPMorgan Chase dashboard reported by Business Insider in Apr 2026 showed about 70,000 provisioned and 24,000 recently active Copilot users.
Leadership
AI, data and platform leaders
Executives
AI org map
Central data and AI office folded into Global Technology; delivery embedded in each line of business
JPMorgan Chase created a Chief Data & Analytics Officer seat on its Operating Committee in Jun 2023, filled by Teresa Heitsenrether, with matching data and analytics roles in each line of business. A Jul 1, 2026 memo from Jamie Dimon and Jennifer Piepszak said Heitsenrether will retire at the end of 2026 and CTO Scot Baldry will take on her role and the firm's overall AI program, and Business Insider reported that the data and analytics office is being aligned with Global Technology. Baldry reports to Global CIO Lori Beer, the Operating Committee member who oversees about 65,000 technologists.
JPMorgan Chase's AI work sits in three places: Global Technology, with about 65,000 technologists under Global CIO Lori Beer; a central data and analytics office now led by CTO Scot Baldry that includes Derek Waldron's AI/ML Centers of Excellence of about 150 applied and research scientists; and data and analytics teams inside each line of business. JPMorgan Chase last gave a firmwide count of more than 2,000 AI and machine learning experts and data scientists in Apr 2024, when its AI Research group had about 200 people, and Evident rated that research team the largest among the 50 banks it tracks as of Apr 2026.
Reporting lines
Chairman and CEO
Chief Operating Officer
Jennifer Piepszak
Manages and coordinates Technology, Operations and Data & Analytics (Jan 2025 announcement)
Global Technology
Lori Beer, Global CIO
About 65,000 technologists and a $19.8B budget; Beer sits on the Operating Committee; the line to the COO is inferred from the Jan 2025 announcement, not published
CTO and Chief Data & Analytics Office
AI and data strategy and governance, commercialization and value delivery, and policy engagement (Jul 2026 memo); Teresa Heitsenrether stays through the end of 2026
Firmwide Chief Data Office
Chief Analytics Office and AI/ML Centers of Excellence
About 150 applied and research scientists; reporting line after Jul 2026 not published, previously part of the CDAO organization
AI/ML and Data Platforms, Data and AI Product
Manoj Sindhwani
Cybersecurity and Technology Controls
Pat Opet, Global CISO
Commercial & Investment Bank
Doug Petno, Co-President and CEO
CIB Data and AI Strategy
Guy Halamish, CIB COO
Data and analytics officers in payments, banking, securities services and markets report jointly to Halamish and business heads
Payments Data and Analytics
Zachery Anderson, CDAO
Consumer & Community Banking
Troy Rohrbaugh, Co-President and CEO
Asset & Wealth Management
Mary Callahan Erdoes, CEO
JPMorgan Chase's leadership page, checked Oct 31, 2026, lists 11 Operating Committee members and no data and AI chief, while Teresa Heitsenrether's bio page still calls her Chief Data & Analytics Officer and a CVS Health filing dated Aug 17, 2026 described her as an Operating Committee member. JPMorgan Chase has not published Lori Beer's own reporting line; the Jan 2025 announcement gave COO Jennifer Piepszak the job of managing and coordinating Technology and Data & Analytics.
Where the AI workforce sits
Central AI Research and AI/ML Centers of Excellence
Estimate: about 200 AI Research staff (Apr 2024) and about 150 Centers of Excellence scientists (2026) against the 2,000+ AI/ML experts and data scientists last disclosed in Apr 2024; the groups may overlap and the total has likely grown.
~15-20%Line-of-business data, analytics and AI/ML teams
Estimate: the remainder of the 2,000+ figure. JPMorgan Chase gives each line of business its own data and analytics roles, and CIB data and analytics officers report jointly to the CIB COO and business heads.
~80-85%
Named people
Use cases
JPMorgan Chase had almost 1,000 AI use cases in Jul 2026, Jamie Dimon said, with about 50 rated most important across risk, fraud, marketing, hedging, prospecting, note-taking, idea generation and document reading. JPMorgan Chase has published a business return for only a few named use cases: fraud, payments screening and exceptions, KYC reviews, wealth management sales and coding. For AI as a whole, JPMorgan Chase put savings at nearly $1.5 billion in May 2025 and Jamie Dimon put them at about $2 billion a year in Oct 2025; in Oct 2026 Jeremy Barnum said he is moving away from a dollar figure and that the gains are not yet visible in quarterly earnings.
| Use case | Area | Status | ROI |
|---|---|---|---|
| Fraud and scam prevention Chase uses machine learning in fraud controls, including improved models that flag suspicious logins, alongside passkeys and checks on higher-risk payments. | Consumer & Community Banking | In production | Cost of fraud flat in basis points as attacks grew 12% a year, with AI a major part (May 2025)Source |
| Operations productivity Chase operations leads Consumer & Community Banking's adoption of new AI tools. | Consumer & Community Banking | In production | Target: more than 40% gross productivity efficiency by 2030, on track to beat (Apr 2026) |
| EVEE Intelligent Q&A EVEE lets Chase call center specialists in customer service, fraud and claims, home lending, wealth management and collections ask questions and get concise answers from bank policies and documentation. | Consumer & Community Banking | In production | |
| Agentic AI in marketing, pricing and personalization Chase is extending mature predictive AI in credit, operations and fraud to agentic capabilities in coding, marketing, sales optimization, pricing and personalization. | Consumer & Community Banking | Announced | |
| Payments transaction screening J.P. Morgan uses AI to screen payment transactions. | Commercial & Investment Bank | In production | More than double the screening volume with half the manual operator checks (Apr 2026) |
| Payment exceptions J.P. Morgan Payments uses AI models to resolve payment exceptions, so fewer of them need manual handling as volumes grow. | Commercial & Investment Bank | In production | Manual payment exceptions down more than 50% while volumes rose over 50% (May 2025) |
| KYC reviews JPMorgan Chase uses AI and other technology in the know-your-customer reviews it runs when onboarding Commercial & Investment Bank clients. | Commercial & Investment Bank | In production | KYC unit cost down 40% since 2022, credited to AI and technology (May 2025) |
| Cash-flow forecasting for treasury clients J.P. Morgan offers corporate treasury clients an AI cash-flow forecasting tool for liquidity management. | Commercial & Investment Bank | In production | |
| Prime Finance inventory and pricing J.P. Morgan's Prime Finance business applies AI to manage its securities inventory, sharpen pricing, strengthen risk management and use capital more efficiently. | Commercial & Investment Bank | In production | |
| Pitch decks and deal documents LLM Suite produced a five-page investment banking deck for a meeting with Nvidia's leaders in about 30 seconds in a CNBC demonstration, and JPMorgan Chase is training AI to draft other banking documents such as confidential M&A memos. | Commercial & Investment Bank | Pilot | |
| Proxy IQ Proxy IQ analyzes data from more than 3,000 annual U.S. company meetings and, from Apr 1, 2026, replaced external proxy advisers for J.P. Morgan Asset Management's U.S. votes. | Asset & Wealth Management | In production | |
| SpectrumIQ research and trading SpectrumIQ ties research, data and risk together inside the Spectrum platform, which automates nearly 75% of equity trading and almost 85% of FX trading and rebalances trades for more than 2,500 clients a month. J.P. Morgan says its Smart Monitor tool cuts the time from research to insight by 80%. | Asset & Wealth Management | In production | |
| Connect Coach for advisers Connect Coach uses 25 specialized AI agents to prepare advisers for meetings, analyze portfolios and summarize calls across the Private Bank and U.S. Wealth Management, and had 12,000 users and about 1 million AI-driven insights pushed by Apr 2026. | Asset & Wealth Management | In production | Asset & Wealth Management gross sales up 20% from 2023 to 2024, partly credited to its AI adviser tools (May 2025)Source |
| Overnight screening for private bankers J.P. Morgan's private bank runs AI overnight across market activity, client positions and research so bankers can spend their time with clients. | Asset & Wealth Management | In production | |
| AI-assisted controls reviews Asset & Wealth Management built an AI workflow with about 80 prompts for controls work that requires reading 50-plus pages and asking the right questions, then shared it across JPMorgan Chase, where it grew from 200 users to 3,000 with 3,000 to 5,000 more identified (Feb 2026). | Asset & Wealth Management | In production | |
| AI coding assistants JPMorgan Chase engineers use an in-house coding assistant, GitHub Copilot and Claude Code, and Lori Beer says most code is now assisted or written with agentic coding tools. | Global Technology | In production | Coding efficiency up more than 10% from AI in 2025Source |
| Employee Assistant Employee Assistant is a personalized AI agent that gives JPMorgan Chase staff one place to get help and take action across the firm. | Firmwide | In production | |
| Performance review drafting JPMorgan Chase staff can use LLM Suite to draft year-end performance reviews, keep responsibility for the final text, and may not use the tool for pay decisions. | Firmwide | In production | |
| Long-running autonomous agents JPMorgan Chase plans agents that work through multistep workflows across software for an hour or two at a time, which Derek Waldron said would arrive in 2026. | Firmwide | Announced | |
| Vulnerability hunting with Claude Mythos Preview JPMorgan Chase is evaluating Claude Mythos Preview for defensive security as a Project Glasswing launch partner, an effort Pat Opet said the bank would approach in a rigorous, independent way. | Cybersecurity | Pilot |
Adoption
JPMorgan Chase spread AI in layers: predictive machine learning in fraud, marketing, credit and risk for more than a decade, an employee generative AI portal (LLM Suite) from summer 2024, coding assistants for its technologists, and agents for multistep work from Sep 2025. LLM Suite went from more than 60,000 users in Aug 2024 to about 250,000 with access in Sep 2025 and about 150,000 weekly users in Feb 2026, and the Commercial & Investment Bank alone reported more than 65,000 active users in Apr 2026. Coding tools show a split picture: the CIB says more than 90% of its engineers use AI code assistants, while a firmwide GitHub Copilot dashboard reported by Business Insider in Apr 2026 showed about 24,000 recently active users out of 70,000 provisioned.
By the numbers
Rollout, first pilot to today
Jul 2018
JPMorgan Chase hired Carnegie Mellon professor Manuela Veloso to found its AI Research group.
Jun 2023
JPMorgan Chase created a Chief Data & Analytics Officer role on its Operating Committee and appointed Teresa Heitsenrether to lead data, analytics and AI adoption.
Apr 2024
Jamie Dimon reported more than 2,000 AI and machine learning experts and data scientists at JPMorgan Chase and more than 400 AI use cases in production, in areas such as marketing, fraud and risk.
May 2024
J.P. Morgan launched IndexGPT, which uses OpenAI's GPT-4 to build thematic index baskets.
Aug 2024
LLM Suite reached more than 60,000 JPMorgan Chase employees on OpenAI models, more than a year after the bank restricted staff use of ChatGPT.
Dec 2024
Lori Beer told AWS re:Invent that LLM Suite was deployed to more than 200,000 JPMorgan Chase employees and that AWS SageMaker underpins the firmwide AI platform.
Mar 2025
Lori Beer said tens of thousands of JPMorgan Chase engineers raised productivity 10% to 20% with a coding assistant the bank built.
May 2025
JPMorgan Chase's investor day reported more than 40,000 engineers with AI coding assistants, more than 200,000 LLM Suite users and about 100 generative AI solutions in production.
Sep 2025
Derek Waldron said about 250,000 JPMorgan Chase employees, everyone except branch and call center staff, had LLM Suite access, half used it roughly daily, and the bank had begun deploying agentic AI for multistep tasks.
Feb 2026
Jamie Dimon said about 150,000 JPMorgan Chase employees use LLM Suite every week, and the bank said its AI solutions in production doubled in 2025.
Apr 2026
JPMorgan Chase reported that more than 90% of CIB engineers use AI code assistants, more than 65,000 CIB staff actively use LLM Suite, and an Employee Assistant agent had rolled out to employees.
Jul 2026
Jamie Dimon said JPMorgan Chase has almost 1,000 AI use cases, with about 50 rated most important across risk, fraud, marketing, hedging, prospecting, note-taking, idea generation and document reading.
Sep 2026
Business Insider reported that about 8,000 JPMorgan Chase staff held Claude licenses in Aug 2026 as Claude Code users moved into the Devspace sandbox.
Oct 2026
Lori Beer said more than 200 applications run on LLM Suite and that most code at JPMorgan Chase is now assisted or written with agentic coding tools.
Economics & controls
Spend and ROI
Technology budget
JPMorgan Chase expects to spend about $19.8 billion on technology in 2026, up 10% or $1.9 billion from about $18 billion in 2025 and $17 billion in 2024. The 2026 figure splits about $8.8 billion for the Commercial & Investment Bank, $8.2 billion for Consumer & Community Banking, $2.2 billion for Asset & Wealth Management and $0.6 billion for Corporate, and includes $9.2 billion of investment spending that covers AI and machine learning. Jeremy Barnum said AI-related shortages are pushing up memory and hardware costs, and about a quarter of the roughly $600 million in identified technology efficiencies are AI-related (Feb 2026).
What AI itself costs
Jamie Dimon said in an Oct 7, 2025 Bloomberg TV interview that JPMorgan Chase spends about $2 billion a year on AI and saves about the same. Jeremy Barnum said on Oct 13, 2026 that the AI expense figure at the time was significantly lower, something like $700 million a year, called that number stale, and said AI spending is now too diffused across the bank to ring-fence. Barnum said in Jul 2026 that token costs were trivial in the first half of 2026 but forecast to accelerate in the second half, and Business Insider reported $2,000 monthly Claude Code caps for some engineers in Sep 2026.
Reported ROI
JPMorgan Chase said in May 2025 that AI initiatives had saved nearly $1.5 billion across fraud prevention, personalization, trading, operations and credit decisions, and Jamie Dimon put annual savings at about $2 billion in Oct 2025. The Feb 2026 company update said revenue is the largest source of AI benefits, and Consumer & Community Banking reported a nearly 60% increase in value from AI and machine learning in 2025. Jeremy Barnum said in Oct 2026 that he is moving away from a dollar figure and that the benefits are not yet observable in quarterly earnings.
Productivity markers
JPMorgan Chase reported coding efficiency improvements of more than 10% from AI in 2025 (Feb 2026), and the Commercial & Investment Bank said AI lets it review more than double the payment screening volume with half the manual operator checks (Apr 2026). Jamie Dimon said the roughly 150,000 weekly LLM Suite users believe they save about four hours a week, but the bank does not count those hours in project returns or treat them as headcount reductions. Derek Waldron credited AI with a 20% increase in private banking gross sales and said bankers could eventually cover up to 50% more clients (Jun 2026).
Headcount
JPMorgan Chase had 318,512 employees at Dec 31, 2025, roughly flat on the year, with operations down 4% and support functions down 2% while front-office roles grew 4% and technology grew 1%. Marianne Lake said in May 2025 that consumer operations headcount would trend down about 10% over five years as the business grows, and Jamie Dimon said at Davos in Jan 2026 that the bank will likely employ fewer people in five years. Dimon said in Jul 2026 that some discrete areas cut jobs by 30% to 40% with most affected staff offered roles elsewhere, while Jennifer Piepszak wrote in Apr 2026 that lower headcount is not the goal, even if it is the likely outcome in certain jobs.
Governance and risk
Model risk review
JPMorgan Chase governs AI and machine learning models through Model Risk Governance and Review, an independent function that must approve new models and material changes before use, tiers models by complexity, exposure and reliance, and can require mitigation when a model is used before review. Model risk stays with the business users who own each model, according to the 2025 Form 10-K.
Data boundary
JPMorgan Chase restricted staff use of ChatGPT in 2023 and built LLM Suite so that bank data is not used to train outside models, Teresa Heitsenrether said in Aug 2024. Derek Waldron said in Jul 2025 that an enterprise should decide deliberately which channels bring models in, from model providers and cloud platforms to software vendors and open-source models, keep data from being stored where it should not be or used for training, and shut down every other channel.
Coding agents in a sandbox
JPMorgan Chase is moving Claude Code users into Devspace, a containerized sandbox hosted on AWS that limits the AI's access to employee credentials and internal systems (Business Insider, Sep 2026). Pat Opet said in Apr 2026 that agents should run with an identity but no entitlements, so the bank controls each request for access. Lori Beer said in Oct 2026 that agents get guardrails at a very fine-grained level before they act on their own.
Outside agents and customers
Jamie Dimon said on Oct 13, 2026 that JPMorgan Chase will not let outside AI agents into its systems, citing the risk of rogue agents and the liability that shifts to customers who hand over credentials. A JPMorgan spokesperson told Fortune the bank blocks third-party agents that try to reach accounts with customers' banking credentials and supports customer-authorized data sharing through open banking providers.
Disclosed risks and regulation
JPMorgan Chase's 2025 Form 10-K lists risks from generative and agentic AI, including biased or inaccurate output, agentic systems reaching sensitive data, AI-enabled cyberattacks, conflicting regulation and the loss of direct customer relationships if AI agents manage financial decisions. Jamie Dimon's Apr 2026 letter warned against both overregulating after a first serious incident and underreacting, and Scot Baldry's remit includes AI governance and engagement with policymakers and regulators (Jul 2026 memo).
Cyber defense
JPMorgan Chase joined Anthropic's Project Glasswing as a launch partner in Apr 2026 to evaluate Claude Mythos Preview for defensive security, with Pat Opet saying the bank would take a rigorous, independent approach. Jamie Dimon called cyber one of the bank's biggest risks in his Apr 2026 letter and said on Oct 6, 2026 that AI-related cyber risk went up tenfold after Mythos.
Measuring use
JPMorgan Chase tracks engineers' use of AI coding tools on internal dashboards that classify non, light and heavy users, which the bank says measures the effectiveness of its technology investments (Business Insider, Apr 2026). A JPMorgan spokesperson said in Sep 2026 that the bank ties AI costs to measurable business value across adoption, productivity, quality, speed, capacity, risk reduction and business impact.
Recent signals
Oct 22, 2026
Lori Beer said at the Evident AI Symposium that LLM Suite is actively used by 200,000 JPMorgan Chase employees, that the bank built its model ecosystem within its private cloud on Microsoft Azure, and that coding assistants account for most of its token spend.
Oct 13, 2026
Jeremy Barnum said on the Q3 2026 call that JPMorgan Chase's AI-specific expense was something like $700 million a year when $2 billion was cited in 2025, a figure he called stale, that he is moving away from a dollar value for AI benefits, and that the benefits are not observable in quarterly earnings; Jamie Dimon said the bank will not let outside AI agents into its systems.
Oct 6, 2026
Evident's 2026 AI Index ranked JPMorgan Chase the most advanced bank in AI for a fifth straight year, ahead of Capital One and RBC.
Oct 6, 2026
Jamie Dimon told Bloomberg TV that AI-related cyber risk went up tenfold after Anthropic's Mythos model and that AI created vulnerabilities JPMorgan Chase did not know about.
Oct 1, 2026
Lori Beer told American Banker that JPMorgan Chase has built more than 200 applications on LLM Suite and that most of its code is now assisted or written with agentic coding tools.
Sep 17, 2026
Business Insider reported that JPMorgan Chase set $2,000 monthly Claude Code spending limits for some engineers and is moving Claude users into Devspace, an AWS-hosted sandbox; about 8,000 staff held Claude licenses in Aug 2026.
Jul 27, 2026
Business Insider reported that JPMorgan Chase is aligning its Chief Data & Analytics Office with Global Technology, that Manoj Sindhwani is adding the data and AI product team, and that Gerard Francis, chief product officer for data and AI, is leaving.
Jul 14, 2026
Jamie Dimon said JPMorgan Chase has almost 1,000 AI use cases and has cut jobs by 30% to 40% in some discrete areas, while Jeremy Barnum forecast a meaningful rise in token costs in the second half of 2026.
Jul 1, 2026
Jamie Dimon and Jennifer Piepszak told JPMorgan Chase staff that Chief Data & Analytics Officer Teresa Heitsenrether will retire at the end of 2026 and CTO Scot Baldry will take on her role without joining the Operating Committee.
Jun 9, 2026
Derek Waldron told CNBC that JPMorgan Chase will deploy long-running autonomous agents in 2026 and that AI tools helped lift private banking gross sales by 20%.
Apr 24, 2026
eFinancialCareers, citing Business Insider, reported that JPMorgan Chase tracks engineers' GitHub Copilot use on dashboards, with about 70,000 provisioned and 24,000 recently active users.
Apr 7, 2026
JPMorgan Chase joined Anthropic's Project Glasswing as a launch partner, the only bank on the list, to test Claude Mythos Preview for defensive security.
Apr 6, 2026
Jamie Dimon's annual letter said AI will affect virtually every function at JPMorgan Chase and will definitely eliminate some jobs, and the annual report restated the $19.8 billion 2026 technology budget and described a new Employee Assistant agent.
Feb 27, 2026
Risk.net reported that Guy Halamish was named COO of JPMorgan Chase's Commercial & Investment Bank with responsibility for its data and AI strategy, and that head of AI Research Manuela Veloso had left after almost eight years.
Feb 23, 2026
JPMorgan Chase told investors that AI solutions in production doubled in 2025, technology spending would rise 10% to about $19.8 billion in 2026, and about 150,000 employees use LLM Suite every week.
Jan 7, 2026
J.P. Morgan Asset Management cut ties with proxy advisers for U.S. votes in favor of Proxy IQ, an in-house AI platform covering more than 3,000 annual company meetings, with the switch completed on Apr 1, 2026.
Quick answers
- How does JPMorgan Chase deploy AI?
- JPMorgan Chase deploys AI through LLM Suite, an internal portal it built in 2024 that puts OpenAI and Anthropic models behind one interface, alongside coding assistants for its 65,000 technologists and almost 1,000 AI use cases across its businesses. About 150,000 employees used LLM Suite each week in Feb 2026, and the AI program moved into Global Technology in Jul 2026 under CTO Scot Baldry, who succeeds Teresa Heitsenrether as Chief Data & Analytics Officer when she retires at the end of 2026.
- What does JPMorgan Chase use AI for?
- JPMorgan Chase had almost 1,000 AI use cases in Jul 2026, Jamie Dimon said, with about 50 rated most important across risk, fraud, marketing, hedging, prospecting, note-taking, idea generation and document reading. JPMorgan Chase has published a business return for only a few named use cases: fraud, payments screening and exceptions, KYC reviews, wealth management sales and coding. For AI as a whole, JPMorgan Chase put savings at nearly $1.5 billion in May 2025 and Jamie Dimon put them at about $2 billion a year in Oct 2025; in Oct 2026 Jeremy Barnum said he is moving away from a dollar figure and that the gains are not yet visible in quarterly earnings.
- Who leads AI at JPMorgan Chase?
- AI at JPMorgan Chase is led by Lori Beer (Global Chief Information Officer); Scot Baldry (Chief Technology Officer; taking over as firmwide Chief Data & Analytics Officer); Teresa Heitsenrether (Chief Data & Analytics Officer, retiring at the end of 2026). Ownership of AI on the org chart: CTO Scot Baldry, under Global CIO Lori Beer; becomes CDAO at the end of 2026.
- What AI platform does JPMorgan Chase use?
- LLM Suite, a portal JPMorgan Chase built in 2024 to give staff secure access to outside models, now carries more than 200 internal applications (Lori Beer, Oct 2026).
- Which AI models does JPMorgan Chase use?
- OpenAI and Anthropic models run behind LLM Suite, engineers use GitHub Copilot and Claude Code, and JPMorgan Chase also runs open-weight models in its own data centers (Oct 2026).
- How many JPMorgan employees use AI?
- About 150,000 JPMorgan Chase employees used LLM Suite each week in Feb 2026; about 250,000, everyone except branch and call center staff, had access in Sep 2025.
- How far has AI spread at JPMorgan Chase?
- JPMorgan Chase sits at Scaling: LLM Suite is open to about 250,000 employees, hundreds of AI use cases run in production (more than 400 by Apr 2024, and the count of AI solutions in production doubled in 2025), and a named executive owns the AI program. JPMorgan Chase falls short of Embedded on two tests: about 150,000 employees used LLM Suite each week in Feb 2026, under half of its 318,512 staff, and the bank stopped putting a dollar figure on AI's effect in Oct 2026.
Open questions
- Who will speak for AI on JPMorgan Chase's Operating Committee after Teresa Heitsenrether leaves, given that Scot Baldry is not joining it and Lori Beer is the remaining technology member?
- Where do Derek Waldron's Chief Analytics Office and Sumitra Ganesh's AI Research group report after the Jul 2026 reorganization?
- How many AI and machine learning specialists does JPMorgan Chase employ today, given that the last firmwide figure, more than 2,000, dates from Apr 2024?
- What is JPMorgan Chase's AI-specific expense today, given that Jeremy Barnum called the roughly $700 million figure for 2025 stale, what did it include, and how fast did token costs grow in the second half of 2026?
- How much value does AI deliver now that JPMorgan Chase has stopped publishing a dollar estimate, and how do the earlier figures of about $1.5 billion (May 2025) and about $2 billion (Oct 2025) break down?
- How many of JPMorgan Chase's almost 1,000 AI use cases are in production, and how many are still pilots?
- What business return do JPMorgan Chase's deployed tools with no published figures deliver, such as EVEE, Proxy IQ and Employee Assistant?
- What share of LLM Suite traffic goes to OpenAI, Anthropic and open-weight models, and is any model fine-tuned on JPMorgan Chase data?
- Which discrete areas cut jobs by 30% to 40% because of AI, how many roles were affected, and how many people were redeployed?
- When, if ever, will JPMorgan Chase let customer-authorized outside AI agents act on accounts, given that it set no timeline in Oct 2026?
- What did JPMorgan Chase find when it tested Claude Mythos Preview through Project Glasswing?
- How far will Claude Code and Devspace roll out beyond about 8,000 Claude license holders, and do AI usage scores feed into engineers' performance reviews?
Sources
JPMorgan Chase
- Leadership and Operating Committee
- Lori Beer, Global CIO, biography
- Teresa Heitsenrether, Chief Data & Analytics Officer, biography
- Jennifer Piepszak, COO, biography
- LLM Suite named 2025 Innovation of the Year by American Banker (Jun 2025)
- Lori Beer showcases advancements at AWS re:Invent (Dec 2024)
- Doug Petno and Troy Rohrbaugh named Co-Presidents (Jun 25, 2026)
- New responsibilities for senior leaders (Jan 14, 2025)
Filings, letters and investor materials
- Jamie Dimon, letter to shareholders, 2025 Annual Report (Apr 6, 2026)
- Jennifer Piepszak, COO letter, 2025 Annual Report (Apr 2026)
- Marianne Lake, Consumer & Community Banking letter, 2025 Annual Report (Apr 2026)
- Mary Callahan Erdoes, Asset & Wealth Management letter, 2025 Annual Report (Apr 2026)
- Doug Petno and Troy Rohrbaugh, Commercial & Investment Bank letter, 2025 Annual Report (Apr 2026)
- 2026 Company Update, full event transcript (Feb 23, 2026)
- 2026 Company Update, firm overview presentation (Feb 23, 2026)
- 2025 Investor Day, firm overview presentation (May 19, 2025)
- 2025 Investor Day, full transcript (May 19, 2025)
- Jamie Dimon, letter to shareholders for 2023 (Apr 2024)
- JPMorgan Chase 2025 Form 10-K (filed Feb 2026)
- CVS Health 8-K exhibit naming Teresa Heitsenrether to its board (Aug 17, 2026)
- Q3 2026 earnings call transcript, Benzinga via Webull (Oct 13, 2026)
- Q2 2026 earnings call transcript, The Motley Fool via Webull (Jul 14, 2026)
News
- Cybersecurity Dive: JPMorgan Chase and Goldman Sachs on AI security, Lori Beer at the Evident AI Symposium (Oct 2026)
- Bloomberg Law: JPMorgan AI chief, longtime trader exits after four decades (Jul 2026)
- PYMNTS: JPMorgan AI chief Teresa Heitsenrether plans year-end retirement (Jul 1, 2026)
- Business Insider: JPMorgan's AI leadership is getting a reset (Jul 27, 2026)
- Business Insider: JPMorgan rolls out Claude changes, $2,000 spending limits and extra security (Sep 17, 2026)
- eFinancialCareers: JPMorgan's AI usage dashboards for engineers (Apr 24, 2026)
- eFinancialCareers: JPMorgan's CTO change, Scot Baldry promoted (Feb 24, 2025)
- eFinancialCareers: JPMorgan's AI jobs now number 2,000 people (Apr 9, 2024)
- CNBC: JPMorgan Chase plans to deploy more powerful AI agents this year (Jun 9, 2026)
- CNBC: JPMorgan Chase's blueprint to become a fully AI-connected megabank (Sep 30, 2025)
- CNBC via NBC News: JPMorgan Chase gives employees an AI assistant powered by OpenAI (Aug 9, 2024)
- Fortune: Jamie Dimon bars outside AI agents from JPMorgan accounts (Oct 15, 2026)
- Fortune: Jamie Dimon at Davos on AI and layoffs (Jan 22, 2026)
- PYMNTS: JPMorgan CEO warns AI risks jumped tenfold after Mythos (Oct 6, 2026)
- PYMNTS: JPMorgan Chase unveils AI tool for thematic investing (May 2024)
- Bloomberg via BGOV: JPMorgan tops global AI banking index for fifth straight year (Oct 6, 2026)
- Bloomberg Law: JPMorgan replaces proxy advisers with AI for voting US shares (Jan 7, 2026)
- Bloomberg via BGOV: Dimon says AI cost savings now match money spent (Oct 7, 2025)
- Risk.net: JP Morgan's data and AI move (Feb 27, 2026)
- American Banker: The Most Powerful Women in Banking, Lori Beer (Oct 1, 2026)
- American Banker: Innovation of the Year, gen AI at JPMorganChase (Jul 2025)
- Reuters via Yahoo Finance: JPMorgan engineers' efficiency jumps as much as 20% (Mar 2025)
- Reuters via AOL: JPMorgan says AI tools lifted wealth sales during market turmoil (May 2025)
- Tearsheet: JPMorgan Chase's gen AI implementation (May 27, 2025)
- CIO.com: JPMorgan Chase builds AI foundation on AWS (Dec 2024)
- CeFPro: JPMorgan launches internal AI tool for writing performance reviews (Nov 2025)
- Forrester: Derek Waldron speaker profile (Nov 2026 event)
- FinTech Futures: January 2026 movers and shakers, Zachery Anderson to J.P. Morgan Payments
Data confidence
- Confirmed
- Stated in a primary or reputable public source.
- Estimated
- Derived or approximated from public reporting.
- Unconfirmed
- Reported, but not yet corroborated.
JPMorgan Chase profile · Enterprise AI diffusion report by Imagine · facts checked against their sources on Oct 31, 2026
