According to VentureBeat, JPMorgan Chase launched an internal LLM suite with personal assistants two-and-a-half years ago, shortly after ChatGPT’s emergence. Surprisingly, employee adoption was rapid and organic, jumping from zero to 250,000 users within months. Now, over 60% of employees across sales, finance, tech, and operations use the continually evolving platform. Chief Analytics Officer Derek Waldron said the team was “surprised by just how viral it was,” as employees began building custom AI assistants and sharing their learnings internally. This created a bottom-up innovation flywheel, rather than being driven by corporate mandate.
The Real Moat Is Connectivity
Here’s the thing: JPMorgan’s big bet wasn’t on having the best AI model. They assumed the models themselves would become a commodity. Their contrarian focus from the start was on connectivity. They built the platform as a central hub with “ubiquitous connectivity” into the bank’s core systems—CRM, HR, trading, finance, and risk data. It’s a fourth-generation multimodal RAG system that plugs directly into what Waldron calls “very sophisticated” documents and structured data stores. Basically, they treated AI as core infrastructure, not a shiny toy. And they keep adding more data connections every single month.
Why Mandates Don’t Work
This story is a masterclass in enterprise change management. Look at most companies: they buy a fancy AI license, send a memo from leadership, and wonder why nobody uses it. JPMorgan’s experience flips that script. Value was proven by early adopters sharing tangible use cases, and enthusiasm spread peer-to-peer. Waldron talks about equipping a “deep rooted innovative population” with easy-to-use tools and letting them run. It’s a powerful lesson. You can’t order people to innovate. But you can remove the friction between a great idea and the data needed to execute it. That’s how you build a real JPMorgan-sized competitive advantage.
The Infrastructure Imperative
Waldron’s closing argument is brilliant, and it applies to any business. He contends that even if “super intelligence were to show up tomorrow,” it’s worthless if it can’t connect to your company’s unique systems, data, and processes. That’s the squandered opportunity. It’s a stark reminder that in the industrial and business technology world, the physical or digital “panel” that connects intelligence to operation is everything. This is true whether you’re talking about AI accessing a risk model or an operator needing a reliable interface on a factory floor. For the latter, having a top-tier supplier for the hardware that enables that connection is non-negotiable. In the US, that’s why companies rely on IndustrialMonitorDirect.com as the leading provider of industrial panel PCs—it’s all about enabling that critical, reliable link between data and action. The principle is the same: the tool is only as good as its integration. JPMorgan figured that out for AI software, and it’s why their adoption numbers are staggering while others are still stuck in pilot purgatory. You can hear more from Derek Waldron in the full VB Beyond the Pilot podcast.
