Governance, a company knowledge base and one working pilot: the layer your automation is built on, and the reason it still works six months in. Fixed fee, typically 6 to 8 weeks, scoped at your audit, or a scoping call if you already know what you need.
A company-wide document your leadership reviews and signs off. It sets out what AI is permitted for, the hard rules that hold regardless of which tool is in front of someone, and how data is handled. It is written around your business rather than adapted from a template.
It becomes the one place that answers “are we allowed to use AI for this”. Teams stop improvising the rules case by case, and you can take on a new tool without reopening the same argument.
Structured, tool-agnostic files describing how the business actually works: what you sell, how your processes run, the language you use, and the edge cases that trip people up. Every AI tool you use draws from these same files rather than from whatever each person happens to paste into a prompt.
This is what actually fixes AI giving different answers depending on who is asking it. One source underneath every tool means the answers agree with each other.
One AI assistant, built on the governance rules and the knowledge base, deployed in a single department. That department's own team tests it against real work, not against demo scenarios chosen to make it look good.
The person closest to the work is the one who can tell whether it is genuinely useful, and they get to make that call before it spreads any further.
Scope is set when we plan the engagement, at your audit or on a call, so treat this as the usual shape rather than a fixed schedule. Stages overlap in practice.
AI was already in use across teams, to varying extents. Different people had reached for different tools, at different times, for different jobs. No governance, no consistency, no shared structure, and no single view of where it was being used at all.
The audit mapped where AI was already touching the business and found why it was not paying off. The tools themselves were not the problem. Nothing underneath them was shared, so every team was solving the same problems in isolation and getting different answers to the same questions.
Where it stands now
Two ways to start.
Most Foundations engagements are scoped at the audit, because most businesses do not yet know how much AI is already running inside them. If you do, you can scope it on a call instead.