JPMorganChase’s LLM Suite tops 200,000 users as engineer AI adoption passes 90%

JPMorganChase’s LLM Suite is an internal generative AI platform that gives more than 200,000 employees access to large language models from providers including OpenAI and Anthropic through a secure abstraction layer, supporting research, document summarization, contract analysis, presentation drafting and coding assistance.

Executive Summary

JPMorganChase, the largest U.S. bank by assets, deployed LLM Suite beginning in summer 2024 and reached 200,000 onboarded employees within eight months, according to a JPMorganChase technology blog post. The platform functions as a secure abstraction layer that connects models from providers such as OpenAI and Anthropic to the bank’s internal systems, enabling employees to perform research, document summarization, contract analysis, presentation drafting and coding assistance. American Banker named LLM Suite its 2025 “Innovation of the Year,” and the publication reported that JPMorganChase was the first big bank to roll out generative AI to nearly all employees through a single portal.

The quantified outcome sits in the engineering organization. More than 90% of JPMorganChase engineers now use AI code assistants, and more than 65,000 colleagues in the Commercial and Investment Bank actively use LLM Suite, according to the bank’s Commercial and Investment Bank letter. Employees report average productivity gains of one to two hours per week from using the platform, JPMorganChase Chief Data and Analytics Officer Teresa Heitsenrether told American Banker.

For C-suite readers, the significance is the operating model around the model. A voluntary rollout, driven in part by employee demand, became a cultural shift across a workforce of more than 200,000 people. The bank frames LLM Suite not as a point tool but as an AI hub, with Chief Analytics Officer Derek Waldron saying, “The North Star for LLM Suite is to position it as an AI hub for employees.”

What is LLM Suite and why does it matter to a bank the size of JPMorganChase?

LLM Suite is not a single vendor product. It is an internal platform with an abstraction layer that routes requests to large language models from providers including OpenAI and Anthropic while keeping data inside a secure environment controlled by the bank. That architecture lets JPMorganChase swap or combine models without re-engineering employee workflows, and it gives the firm a single point for applying controls around access, data handling and usage.

The tasks the platform supports are the daily work of a global bank: research, document summarization, contract analysis, presentation drafting, and coding assistance for transaction screening, pricing and risk tasks. These use cases are high-volume, language-heavy and measurable, which made them suitable for broad internal rollout rather than isolated pilot projects.

American Banker, which gave the platform its 2025 “Innovation of the Year” award, described the deployment as the first time a major U.S. bank made generative AI available to nearly all employees through a single portal. The award recognized both the technical architecture and the organizational achievement of moving from zero to 200,000 onboarded users in eight months.

How did a voluntary rollout reach 200,000 employees in eight months?

The rollout was not mandatory. According to JPMorganChase’s technology blog, employee demand drove adoption, with the platform going from zero to 200,000 onboarded users within eight months. The bank’s 2025 Investor Day transcript separately says more than 200,000 employees globally have access to LLM Suite, and on average, certain key subsets of users report gaining several hours of productivity each week.

The distinction between onboarded users and access matters for enterprise readers. Onboarding signals active enrollment and use of a tool, while access is a measure of entitlement. JPMorganChase reported both: 200,000 onboarded users within eight months, and more than 200,000 employees globally with access as of the 2025 Investor Day.

1. Summer 2024: JPMorganChase releases LLM Suite internally.
2. Eight months after release: the platform reaches 200,000 onboarded users, driven in part by employee demand.
3. As of the 2025 Investor Day: more than 200,000 employees globally have access to LLM Suite.
4. In the Commercial and Investment Bank letter: more than 65,000 CIB colleagues actively use LLM Suite.
5. In the same letter: over 90% of JPMorganChase engineers use AI code assistants.

What are the measured results inside the Commercial and Investment Bank?

The most specific operational numbers come from the Commercial and Investment Bank letter. The letter states that over 90% of engineers now use AI code assistants and that more than 65,000 CIB colleagues actively use LLM Suite. Engineers are applying the tools to transaction screening, pricing and risk tasks, which are high-throughput, judgment-heavy processes where a small efficiency gain compounds across thousands of daily decisions.

Heitsenrether quantified the employee-level benefit in an interview with American Banker: “On average, we're seeing people gaining an hour or two hours of productivity a week, which is really quite significant. And we're really still in the early stages.” The Investor Day transcript echoes that pattern, noting that certain key subsets of users report gaining several hours a week.

At 200,000 users, a one-hour-per-week gain represents 200,000 hours of employee time per week, before accounting for the higher reported gains among power users. JPMorganChase has not published a dollar figure for that time, but the scale of the deployment makes the productivity figure material to operating expense and capacity planning.

The North Star for LLM Suite is to position it as an AI hub for employees.Derek Waldron, Chief Analytics Officer, JPMorganChase

Reported LLM Suite metrics and sources

Metric	Reported value	Source
Onboarded users within eight months of summer 2024 release	200,000	JPMorganChase technology blog
Engineers using AI code assistants	Over 90%	JPMorganChase CIB letter
CIB colleagues actively using LLM Suite	More than 65,000	JPMorganChase CIB letter
Employees with global access to LLM Suite	More than 200,000	JPMorganChase 2025 Investor Day transcript
Average weekly productivity gain per employee	One to two hours	American Banker interview with Teresa Heitsenrether
Weekly productivity gain reported by key subsets	Several hours	JPMorganChase 2025 Investor Day transcript

How does LLM Suite handle model choice, security and data?

LLM Suite is designed as an abstraction layer, which means employees interact with a single internal interface while the platform routes requests to appropriate models. Providers including OpenAI and Anthropic are connected through this layer, and the bank can change or add models without asking users to learn new tools.

The security posture is central to the business case. By keeping model access inside a secure environment, JPMorganChase can apply bank-grade controls to prompts, outputs and data flows, which matters in a regulated industry where client information cannot be sent to public endpoints. The abstraction layer also gives the firm a single point for auditing usage, measuring adoption and enforcing policy.

This architecture differs from a shadow-IT pattern in which employees subscribe to consumer AI tools independently. JPMorganChase’s approach channels demand into a governed platform, which is why the bank could report both rapid adoption and consistent controls. For regulated enterprises, employee demand for AI is not something to suppress; it is something to route through infrastructure the organization can observe.

What does 90% engineer adoption mean for software delivery?

A 90% adoption rate for AI code assistants among engineers is high by most enterprise benchmarks, and it matters because software is an input to nearly every revenue process at JPMorganChase. The bank explicitly connects the tools to transaction screening, pricing and risk tasks, meaning the code being produced is embedded in the firm’s core workflows, not peripheral to them.

Engineer adoption at this level also changes the economics of software delivery. If the average engineer saves even one hour per week, and if code assistants reduce time spent on boilerplate, testing and review, the cumulative effect across the bank’s engineering population is equivalent to adding capacity without adding headcount. JPMorganChase has not disclosed a headcount-equivalent figure, but the stated adoption and productivity data point in that direction.

The CIB letter frames the AI coding tools as a productivity engine rather than a novelty. It describes engineers using AI assistants for transaction screening, pricing and risk tasks, where errors are expensive and where speed translates directly into client service. That pairing of high adoption with high-stakes use cases separates this deployment from a generic developer-tool rollout.

How does JPMorganChase’s deployment compare with the rest of the industry?

American Banker described JPMorganChase as the first big bank to roll out generative AI to nearly all employees through a single portal. Many financial institutions have experimented with point solutions in specific business lines, while JPMorganChase chose a firmwide platform with a common interface and a centralized abstraction layer.

The scale exceeds what is typical in enterprise AI rollouts. A 200,000-user deployment within eight months is difficult to benchmark against public disclosures, but it places LLM Suite in the same category as large internal technology platforms at major technology companies rather than in the category of pilot programs. The CIB letter’s 65,000 active users is itself larger than the total employee base of many mid-sized banks.

For competitors, the comparison is not about model quality, because most banks can buy access to the same frontier models. It is about organizational speed, governance and the ability to turn a voluntary tool into a workforce norm. JPMorganChase’s advantage, according to the bank’s own account, came from building the platform around employee demand rather than forcing adoption.

What are the limits of the reported productivity numbers?

The productivity figures are self-reported. Heitsenrether told American Banker that employees report gaining one to two hours per week, and the Investor Day transcript says certain key subsets of users report gaining several hours. These are survey-style self-assessments, not audited time studies, and the bank has not published the methodology or the sample size.

Adoption also does not equal value. A user can onboard to LLM Suite and still rely on it for low-value tasks, or the time saved may be redirected rather than captured as capacity. The bank has not disclosed a dollar-denominated return on investment, a revenue uplift attributable to LLM Suite, or a direct link between AI code assistants and software delivery outcomes such as release velocity or defect rates.

None of those gaps invalidate the reported milestones, but they should shape how peers read the story. The verifiable claims are the ones the bank documents: 200,000 onboarded users, more than 200,000 employees with access, over 90% engineer adoption of code assistants, more than 65,000 active CIB users, and an average self-reported gain of one to two hours per week.

What should peer executives take away from the JPMorganChase rollout?

One takeaway is that a governed AI platform can scale quickly if it is useful. JPMorganChase did not mandate LLM Suite; it built a secure abstraction layer, connected strong models and let employee demand pull adoption forward. The result was 200,000 onboarded users in eight months, a pace more commonly associated with consumer apps than with regulated financial institutions.

Another takeaway is that adoption metrics and productivity metrics should be reported separately. JPMorganChase discloses how many people have access, how many have onboarded, how many in the CIB actively use the platform, and how much time users say they save. Peer institutions should expect the same granularity from their own AI programs before making claims about return on investment.

A third takeaway is that AI strategy is workforce strategy. The bank’s named leaders describe LLM Suite as an AI hub for employees, not as a cost-cutting tool. The productivity gain of one to two hours per week per employee is modest enough to be credible and large enough to be material at the scale of a global bank. The metric that matters is not model quality in isolation but the combination of adoption, task fit and measured time savings.

We think that AI has the potential to really deliver amazing scale and efficiency as well as client benefit. It's a real priority for us, and we've been very focused on this for a long time.Teresa Heitsenrether, Chief Data and Analytics Officer, JPMorganChase

What comes next for JPMorganChase’s generative AI strategy?

Heitsenrether told American Banker that the firm has been focused on AI for a long time and sees potential for “amazing scale and efficiency as well as client benefit.” The Investor Day transcript frames current productivity gains as an early-stage result, which implies the bank expects measured benefits to grow as models improve, workflows mature and more employees become fluent in the tool.

The abstraction layer gives JPMorganChase optionality as the model market evolves. Because LLM Suite connects to multiple providers, the bank is not locked into a single frontier model and can route tasks to whichever model performs best on cost, latency or quality. That flexibility is likely to become more important as inference costs and the number of capable models both change.

For peer banks, the open question is whether they can replicate the organizational conditions that made this rollout work: a clear internal platform, executive sponsorship from the chief data and analytics office, and a culture where voluntary adoption outpaces mandates. JPMorganChase’s numbers are specific to its own workforce, but the pattern is transferable.

What is not yet public is the financial return on LLM Suite in dollar terms. The bank has disclosed adoption and time-savings data but has not assigned a dollar value to the productivity gain or tied it to specific revenue uplift. Until such figures are published, the strongest verifiable claim is the one the bank itself makes: a large, voluntary, governed deployment that employees use enough to report measurable weekly time savings.