I've stood up two data organizations. The first was at Novavax, in the middle of the Covid vaccine race, where the pressure to move fast shaped almost every hiring and tooling decision I made. The second was at a major blockchain foundation, where I had to be lean, and self-reliant - automation was a must.
I’ve wrestled with "what does a modern data team look like?" for most of the last twenty years. Have you noticed how quickly the conversation in our circles has shifted? Two years ago, agentic AI was a curiosity at most data leadership tables. Today, every director I talk to has watched some part of their team's work get absorbed into an agent in the past six months, while their stakeholder needs as evolved toward self-service and natural language; and most of these directors have the same private question: what does my team actually need to look like a year from now?
I don't fully know the answer. I do believe, though, that most of us - myself certainly included, in both of those past build-outs - have been spending most of our energy optimizing the team we have, when the more important question might be what kind of team we should be becoming. It's harder to redesign the shape than to make the existing shape run a little better. And the people best-positioned to redesign it are usually too deep in the running to do the redesigning.
I'm starting Intelligence Layer because I want to think about that redesign in public, with the people who are quietly doing it. Most published writing on agentic AI in the enterprise comes from vendors, analysts, and consultants. The voices I most want to hear are practitioners in the seat. I'm going to write from that seat, drawing on having stood up two very different data orgs and on the conversations I'm having with people running the shape today.
What you'll find here is one practitioner's working notes on the operating models, team designs, governance structures, and leadership patterns the next generation of data orgs may need. Some of it will be confident. Some of it will be openly uncertain. I'd rather write something I might later revise than wait until I'm sure.
This is for data leaders who suspect the thing they're managing isn't quite the thing they should be building, and who'd rather think about it with someone than alone.
If that's you, subscribe.
— Kyle Langham