USA 
United Kingdom
JPMorgan Chase says more than 60% of its employees now actively use an internal suite of AI tools, following rapid organic adoption that began shortly after launch.
The platform was introduced around two-and-a-half years ago by Derek Waldron and his technical team, not long after the emergence of ChatGPT. At the time, skepticism toward generative AI remained high, and expectations for uptake were uncertain.
Within months, employee usage grew from zero to roughly 250,000 workers. The tools are now used across sales, finance, technology, operations, and other departments, according to the company.
“We were surprised by just how viral it was,” said Waldron, JPMorgan’s Chief Analytics Officer, speaking on a podcast.
He said employees opted into the platform voluntarily and began experimenting beyond basic prompting. Workers built and customized personal AI assistants with specific personas, instructions, and roles, then shared their insights on internal platforms.
Bottom-up adoption fuels innovation
Rather than relying on mandates or formal rollouts, JPMorgan’s adoption was driven by early users sharing practical use cases with colleagues. That peer-to-peer momentum helped create what Waldron described as an internal innovation flywheel.
Employees “weren’t just designing prompts,” he said, but actively developing assistants tailored to their roles and workflows. The result was a pattern of bottom-up engagement that spread across teams and functions.
“It’s this deep rooted innovative population,” Waldron said. “If we can continue to equip them with really easy to use, powerful capabilities, they can turbocharge the next evolution of this journey.”
Many organizations are continuing to struggle with employee engagement around AI, particularly where tools are introduced as top-down initiatives.
AI treated as core infrastructure
JPMorgan’s approach also reflects early architectural decisions that treated AI as foundational infrastructure rather than a standalone novelty. The company assumed early on that AI models themselves would become commoditized.
Instead, Waldron said the real challenge lay in connectivity across systems, data, and tools. That focus led to early investment in multi-modal retrieval-augmented generation, now in its fourth generation and incorporating multiple data types.
The AI suite sits at the center of an enterprise platform with connectors designed to support analysis and preparation. Employees can access documents, knowledge bases, structured data stores, and systems spanning CRM, HR, trading, finance, and risk.
“We built the platform around this type of ubiquitous connectivity,” Waldron said, adding that new connections continue to be added regularly.
He argued that AI’s expanding capabilities deliver little value without meaningful access to enterprise data and workflows.
“Even if super intelligence were to show up tomorrow, there’s no value that can be optimally extracted if that superintelligence can’t connect into the systems, the data, the tools, the knowledge, the processes that exist within the enterprise,” he said.
