Stop Building Isolated AI Teams: How Enterprise AI Actually Ships

Key Takeaways

• Enterprise approvals often stall for six months in internal review councils; counter this delay by shipping small, approved iterations rather than monolithic overhauls.
• Isolated AI innovation squads alienate veteran engineers who hold essential domain knowledge and platform context.
• Nick Kuhn advises running weekly office hours, lunch-and-learn sessions, and open forums to turn AI specialists into internal service hubs instead of top-down decree issuers.
• Enterprise platform deployment requires pairing agent buildpacks and Model Context Protocol (MCP) gateways with existing security sandboxes rather than building shadow infrastructure.

The AI Innovation Team Trap

Nick Kuhn from VMware Tanzu sees the same mistake repeat across large companies: management forms an isolated unit tasked with rolling out artificial intelligence everywhere. This group starts issuing top-down mandates. Existing engineers, who understand the core systems, get cut out of the conversation. Friction spikes immediately.

“I see in organizations like a new technology will come up, and they'll be like a new team will be built, and that new team will decree, but there's so much other, like really skilled people that could help,”▶ 37:53 Kuhn explains. When a specialist squad operates in a bubble, it creates internal resentment while missing the operational context needed to run workloads reliably.

Instead of acting like an ivory tower, Kuhn argues that the new unit must treat internal teams as partners. “The responsibility of the new team is like engage every other team because they have a lot of people that can help you out and achieve your goal,”▶ 41:16 Kuhn says. Host weekly office hours. Set up lunch-and-learn sessions. Open up shared forums where senior staff can test tools, ask blunt questions, and bring their domain expertise to agent workflows.

Surviving the Six-Month Committee Cycle

Enterprise adoption moves at the speed of review boards. Trying to push a massive architecture overhaul through security and compliance in one package guarantees paralysis.

“Most large enterprises have a council or committee for everything,”▶ 35:46 Kuhn notes. “So it's like, well, we shall go to the committee, and we shall present our idea, and we will wait 6 months and see what has the result been.”▶ 35:50

The countermeasure is running small, controlled experiments. Once you secure an initial sign-off from a committee, iterate quickly within those approved guardrails instead of asking for a brand new mandate every quarter.

Kuhn emphasizes this incremental path: “Take baby steps from everything, right? Because, especially when you talk about the enterprise scope, they can be a lot slower to adopt some of these newer technologies. So, you kind of layer that on, and like you don't have to necessarily adopt all of these things at once.”▶ 35:07

Combine standard developer tooling like agent buildpacks, AGENTS.md specifications, and Model Context Protocol (MCP) gateways with existing platform engineering practices. When you respect existing governance while making internal enablement accessible, enterprise resistance fades.

What to Do With This

If you lead an AI initiative or sell to enterprise buyers, cancel your next top-down presentation. Schedule two recurring 45-minute open office hour slots on the calendar this week. Invite legacy engineering leads to bring their messiest integration problems, and address their security blockers live.