JPMorgan is pushing its engineers to use AI more and tracking them with internal dashboards visible to thousands of their peers, ramping up pressure on workers to prove they’re keeping up.
Loading audio narration…
The bank recently told developers in its roughly 65,000-person Global Technology division to show “meaningful improvement” in both the quality and volume of their code by using AI. Behind the scenes, it’s also built internal tools to measure that progress, including dashboards that rank engineers based on their usage of tools like GitHub Copilot and Anthropic’s Claude, and a formula to categorize users.
Business Insider reviewed screenshots of these systems and spoke to seven current and former developers. They described a growing push to produce more, faster, as managers become increasingly aware of how often employees use AI. While some said the tools are helpful, most were also worried about being flagged as underperformers if their usage doesn’t rise.
A JPMorgan spokesperson said the bank does not incorporate the data in “performance management,” and instead tracks it to measure the effectiveness of its AI investments.
“Gathering data and feedback enables us to provide training and support in the right ways to the right people,” the spokesperson said, adding that the bank is “acutely aware that the power of these tools are only realized when people, process, and technology come together.”
JPMorgan isn’t alone in intensifying its expectations around employees’ adoption of AI. Companies across industries are under pressure to prove their AI investments are paying off, including Meta and Google, which have set goals for AI-assisted code and allow managers to mandate the use of agents, respectively.
Tracking AI use at the individual level is not something that Sameer Gupta, head of AI in financial services in the Americas at EY, has previously heard of at banks.
While he couldn’t comment on specific banks’ activities, he’s seen firms track AI usage across teams, which he said can help a company determine which AI tools work, and how to address roadblocks.
“In general, people don’t like to be tracked. Set aside AI for a second,” Gupta said. “If somebody is tracking how many total hours in a day you are on Teams and videoconferencing, it makes you uncomfortable.”
Inside the effort
In public, JPMorgan’s leaders, including CEO Jamie Dimon, have touted the firm’s embrace of artificial intelligence, backed by its nearly $20 billion tech budget. During the bank’s most recent earnings call this month, he predicted that AI would soon be table stakes.
Most of the engineers Business Insider spoke to said that they wanted more clarity on which metrics matter.
Business Insider has seen two dashboards tracking engineers’ use of GitHub Copilot, a key hub combining access to multiple AI tools. One dashboard displayed data about employees’ AI adoption and usage rates, listing nearly 70,000 people as “provisioned users” of Copilot; about 24,000 were listed as recently “active” users as of late March, per a screenshot. Tens of thousands of employees’ records were available, listing the most recent day they’d been active using AI tools, their office location, and their reporting lines. The other dashboard characterizes them as “non,” “light,” or “heavy” users.
According to a screenshot of an internal Microsoft Teams channel viewed by Business Insider, a developer in the Global Technology group sent a message to dozens of colleagues warning them to check whether they had been added to what the sender called the “naughty list” of AI users.
The warning included a link to what appeared to be a large database of individual employees’ AI usage rates, with another link to where those who believed they were wrongly listed as under-users could seek their removal from the list. The developer also encouraged others to download Copilot if they hadn’t already and open it once a month to avoid getting flagged by the system.
One engineer said he checks one of the dashboards frequently, worried that the system might not be capturing the full extent of the activities he performs with AI.
Another who recently left the firm said a direct manager told them in a meeting several months before that they were “showing up as a nonuser of artificial intelligence” and should “get to using it.”
The multi-part formula
One intranet document outlines the bank’s daily scoring system that tracks how engineers use GitHub Copilot tools. Those interactions are weighted, with higher scores for proactive use, like starting a prompt, and lower scores for passive actions, like accepting code an LLM has written. The activity is aggregated into a quarterly score that rates engineers as above or below the median use rate in their line of business, “allowing us to compare performance” from one quarter to the next.
“The general vibe feels like we have to be going 110% every day, all week,” one midlevel developer said. “It’s so interesting. I thought this AI tool was supposed to make our lives easier, but it really seems to have stepped up how much we’re expected to do.”
Other engineers can relate — developers across white-collar industries have told Business Insider that AI has significantly heightened their workload, while others have reckoned with a loss of personal identity as LLMs have taken over tasks they spent years learning to perform.
As some engineers have slowly gained access to Claude Code, Anthropic’s AI coding agent, another dashboard has emerged. Screenshots show it tracked more than 600 people as of early April, including adoption rates and spending per user and in total.
Other documents showed that developers were working on several additional tools that could augment the bank’s tracking capabilities if adopted.
The specter of past tracking
JPMorgan’s history of tracking employee engagement, from office attendance to time spent on Zoom, is intensifying the anxiety for some, particularly in offices affected by pre-planned layoffs this year, three of the developers said.
The new goals for AI usage have some people “a little freaked out,” one programmer with almost half a decade of experience at the bank said. “In some areas, I think it’s probably scarier just because they realize that they have way too many people for the level of work that they need to get done.”
His manager’s words of advice? Get familiar with the tools and use them often, “because no matter where you end up going, whether it’s within the company or whether you get laid off in the coming years, this is what the job’s going to look like no matter what.”
Have a tip? Contact these reporters via email at [email protected] or SMS/Signal at 561-247-5758 or [email protected] or Signal at alicetecotzky.05. Use a personal email address, a nonwork WiFi network, and a nonwork device; here’s our guide to sharing information securely.
