AIfor Enterprise · AI programmes
The AI pilot worked.
Nothing reached production.
For enterprise teams that own an AI programme, before the next commitment is signed.
The AI programme has a name, a budget and a steering group. It does not have an owner, and nobody has written down what success would look like.
AI readiness · Governance and risk · CTO, CIO, CRO, transformation lead
Or it never got that far
Six ways an AI programme stalls
A pilot that stalls is the version you can point at. More often the programme slows earlier, somewhere nobody owns, and the slowdown rarely reaches a status report, because every function is reporting green against its own scope.
Nobody is using it
It shipped. The training happened. The usage numbers have not moved, and nobody wants to be the one who says so.
Risk will not sign
Legal and risk have questions nobody can answer, and no one can say what would change their answer.
The data is not there
The use case assumes data that does not exist in the shape it needs. That became clear after the build started.
Everyone has their own
Four functions are each running an AI project. Nobody holds the total, the overlap, or the combined exposure.
The vendor decision
A platform is about to be chosen, and nothing defines what it has to prove before the first renewal.
Four answers at the board
The board asked what the AI strategy is. Engineering, product, risk and finance each gave a different answer.
Each of these has an answer, and the answer is knowable before the next commitment is signed. They are hard because each function sees only the one nearest to it.
Start from what is stuck
Find the line that sounds like you
What you are committing to
Typical shape, not a fixed package
AI Readiness Assessment
Typically three to four weeks where one use case is in scope. Longer where the programme spans several and the stakeholder count climbs.
The Briefing
Typically one session, built around your situation.
Who else gets asked
Procurement will compare them anyway
This spend is a line item and it goes through committee, so it gets compared whether or not the comparison is ever written down. Each of the four below is the right answer to some version of this problem.
A large consultancy.
They will bring a team, a method and a deck, and they can absorb a programme this size without noticing. They will also take a quarter to arrive at the position you need this month, and the people who present the work are rarely the people who stay for it.
The internal team.
They understand the organisation better than any outsider will, which is the argument for them and also the limit. The disagreement that stalled the programme runs between them, and somebody inside a disagreement cannot be the one who closes it.
The vendor already in the account.
Cheapest to start, because the relationship and the paperwork exist. Their read on what you need is bounded by what they sell, and neither side can fully discount that when the recommendation arrives.
Deferring to next quarter.
It costs nobody anything personally, which is the whole reason it wins. The budget holds. The steering group keeps meeting. The decision that was due this quarter comes back next quarter with more sunk cost attached to it.
The case for us is a narrow one. We sit outside the disagreement, and we are gone once it is settled. On a programme this size that is a small line against a decision that is not.
Why SHA/RP AI
What sits behind the work
Production, not demos.
Raphaël worked in product at Google, then led product at Horangi Cyber Security from MVP to its acquisition by Bitdefender. The work is judged on what ships.
We run on it ourselves.
SHA/RP is run by the same kind of AI team we set up for clients: chiefs of staff, automated workflows, and the groundwork done before the day starts.
A straight answer.
The readiness assessment can come back No-Go, and on a first call we will tell you if AI should not take a piece of work.
Yours to keep.
Documented, your team trained, and nothing that depends on us after handover.
On the record
“… He doesn't jump to solutions. He gets to the crux of the problem first, builds solid hypotheses, and does the foundational research to back them up. The rigor was real. …”
Executive checklist
Get the Executive AI Pilot-to-Production Checklist (PDF)
23 operational checks across five areas, to move AI past proof of concept.
Recognise your organisation in this?
That is enough to start. You do not need a business case, a budget line or an answer to what it has to prove.
Request a call. The first step is a conversation to see whether we are a good fit for each other.
No obligation either way. We will come back with a tailored answer, or tell you it is not us.
Or write to us directly at contact@sha-rp.com.