Project, then ongoing

AI trust program

You ship AI and a buyer has started asking whether it is safe. You need an answer that survives their security review.

Good fit: AI is in the product and enterprise buyers are starting to ask hard questions about it.

What this covers

Model and data inventory. What you run, on whose data, and where that data goes.
Controls mapped to NIST AI RMF or ISO 42001, in the places a buyer will actually look.
Vendor and model risk, including the models you did not train.
The gate that decides how an AI feature reaches an enterprise customer.
The customer-facing story that comes out of it, in language your sales team can say out loud.

Worth saying plainly. The boundary is what makes the rest of it credible.

Red team your model: That is a specialist engagement. We will tell you when you need one and what to ask for.
Certify your AI: Nobody can. What exists is evidence that holds up under questioning, which is what we build.
Write your AI features: We govern how they ship. Your team builds them.

Week one: Inventory. What models you run, what data reaches them, and what leaves your boundary.
Week two: The gaps a buyer will find, ranked by which deal each one blocks.