The Project
As AI use spreads, it brings a new kind of software with it: small tools that people build with AI help. Custom dashboards, internal utilities, one-off apps that solve a real problem for one team. They are useful, and they are exactly the kind of thing that ends up running somewhere nobody is watching.
My strategy is to give those tools a supported home early, while there are still only a few of them. I am championing that strategy and architecting the platform behind it. The work is in progress, so this page covers the problem, the strategy, and my role rather than results.
The problem
AI lowers the cost of building a tool far more than it lowers the cost of running one. Someone can produce a working dashboard in an afternoon. Hosting it, securing it, keeping its data access appropriate, and knowing who owns it next quarter are still real work. Without a paved path, each new tool either stalls or finds an unofficial home.
The strategy: make the safe path the easy path
It is the same principle behind our AI governance. An AI-built tool should ship the way any other internal service does: through the platform the teams already run, with infrastructure as code, GitOps, and the same review and change process. It builds on the guardrails we already have for AI use: usage policy, data and PHI rules, and model access approvals. The path also has to be short enough that people actually use it, or they will route around it.
My role
I am the one championing and architecting it, the same way I architected our Kubernetes platform and then handed implementation to the team. It pairs with my Zero Trust Agents work: one gives AI agents a safe place to act, and this gives what engineers build with AI a safe place to run.
Wrap Up
Getting approved AI use to most of engineering was the first half. The second half is making sure what they build with it can run safely and stay owned. I will add outcomes here once the platform is in use.