Search Precision Consulting
Type to search · ↑↓ to navigate · Enter to open · Esc to close
AI Project Recovery
AI project rescue and recovery
stop, fix or scale, decided in weeks
When an AI initiative stalls, the first job is to establish its true position: what it returns against its business case, how it behaves on real data, and what it would cost to run. Our Stop, Fix or Scale Review does that over a planned two to four weeks and ends in a decision you make with us, with a costed plan any firm could carry out. We will lead the fix if you want us to.
The first conversation is confidential, without obligation and without charge.
This review is not independent. We say at the start that we would like to lead any fix, and your board should not rely on it as independent assurance. If you need a finding from a reviewer with nothing to gain, choose an Independent AI Review. Independent AI Review
Why do AI initiatives stall after the pilot?
Often because the pilot proved the technology and nobody re-tested the rest: the value case, the data at production volume, the cost to run, the controls and who owns the result.
Many Australian organisations are at that point now. The Reserve Bank's liaison with firms found enterprise use of AI "largely for pilot or experimental purposes", and the Digital Transformation Agency has told agencies to "stop counting pilots and start scaling what works".
The signs we look for:
- the benefit in the business case has not been measured on live work;
- the system behaves differently on production data than it did in the pilot;
- nobody can say what it will cost each month at full volume;
- approval points and an audit trail were meant to be added later;
- the sponsor has moved on, or ownership sits between two teams;
- the vendor and the organisation read the contract differently.
What does the review examine?
Six things, each against evidence: the value the initiative returns against its original business case; the data it depends on; how the model behaves on real data; what it would cost to run in production; the controls around it; and the position of the team and the vendor.
Where the system was built by another supplier, we examine it only with your written authority for every system in scope, including any hosted by a third party.
Who is the Stop, Fix or Scale Review for?
It is for the executive who owns a stalled AI initiative and wants a decision and a path: a sponsor, chief information officer, chief financial officer or chief operating officer, or an operating partner in a private equity portfolio company.
It fits when you have:
- a pilot that worked and has not reached production;
- an AI agent deployment that has overrun its budget or timeline;
- an assistant rollout that has lost the support of the people meant to use it;
- many pilots and nothing in production (see pilot triage, below).
It does not fit when:
- your board, a regulator or an audit committee needs an independent finding. Choose an Independent AI Review;
- you want evidence for a dispute with a vendor. The review excludes fault findings for litigation, and contractual questions go to your counsel;
- we have independently reviewed the system, or one it replaces, in the past 24 months. Our rules bar us from recovering it.
What do you receive?
The true position of the initiative, a decision taken with you, and a costed plan that stands on its own. The plan can go to any firm, including your own team.
- The true position: value against the business case, data, model behaviour on real data, cost to run, controls, and team and vendor position.
- The decision: stop, fix or scale, taken with the sponsor, with the reasons written down.
- The plan: if the answer is fix or scale, a costed recovery plan with its objectives in order, its milestones and the measures that show it is working. If the answer is stop, what to keep, what to close and how to close it cleanly.
- Our interest, declared: the scope states that we would like to lead the recovery, and the findings are written to stand without us.
How does the review run?
Over two to four weeks, led by a recovery lead from our Programme Recovery practice with an engineer who examines the system itself. The statement of work sets out the days each person spends.
Planned elapsed time Two to four weeks
| When | What happens |
|---|---|
| Week 1 | The business case, the contract and the delivery record; interviews with the sponsor, the team and the vendor; access to the system under your written authority. |
| Weeks 2 to 3 | The engineer tests the system on real data and costs its production run; the recovery lead tests the value case, the controls and the delivery position. |
| Final week | The position agreed with the sponsor; the stop, fix or scale decision; the plan. |
Two to four weeks is the planned elapsed time. We replace it with measured times as reviews complete.
Which door?
Choose an Independent AI Review if your board, a regulator or an audit committee needs a finding from a reviewer with nothing to gain from it. We cannot win the fix: after the review we do not build, fix or run the system, or a system that replaces it, for 24 months. If a fix is needed, we hand you a transition pack (the evidence, the recovery objectives in order, and the testing harness where we built one) so any firm can start in days.
Independent AI ReviewChoose the Stop, Fix or Scale Review if you want a decision and a path and do not need independence. We say at the start that we would like to lead the fix, and we never describe this review as independent.
What if we have many pilots and nothing in production?
Then the first decision is which few to fund. An AI Pilot Portfolio Triage ranks every pilot by value and feasibility, stops the weakest, and funds the few that can reach production, each with an owner and a gate.
It is planned at three to six weeks and starts the register of AI in use. As with the Stop, Fix or Scale Review, we declare that we would like to lead any recovery that follows.
Who leads the fix?
You decide. If you choose us, a recovery lead and Precision engineers re-baseline the initiative and take it to production or to a clean close, over a planned three to six months, and hand it back in a form your people can operate.
Embedded AI Recovery includes:
- a re-baselined plan, value case and cost to run;
- the engineering to finish, fix or retire the system;
- capability transfer to your team;
- a hand-over you can operate: infrastructure as code in your own cloud account, runbooks, the evaluation suite and a licence that lets you, your contractors and any later reviewer use and modify what we built.
Fees are scoped to your situation and quoted after the review.
If you recover another supplier's build, you confirm in writing your authority for us to access and modify that supplier's code and configuration, and your counsel handles any contractual question with the supplier.
After the recovery, we hand the recovered system over to your team or your existing operator.
Once we have recovered a system, we are its builder. Any later independent review of it goes to another firm. Read our independence and conflict rules
Can you recover a system another supplier built?
Yes, with your written authority for access to that supplier's code and configuration.
We keep the work on getting the system to production or to a clean close, and your counsel handles any question of fault or contract with the supplier.
What record stands behind the review?
The review is led from our Programme Recovery practice. These are the firm's own figures, as its home page states them.
- 90+ Client Programmes Recovered
- 30+ Years of Experience
Common Questions
The questions we are asked most
Why do AI projects fail to reach production?
A common reason is that the pilot proved the technology and nobody re-tested the value case, the data at production volume, the cost to run, the controls or who owns the result. A Stop, Fix or Scale Review tests all of them against evidence and ends in a decision.
How long does a Stop, Fix or Scale Review take?
The review is planned at two to four weeks from the start of work, depending on the size of the initiative and how quickly we can reach the system and its data.
How much does it cost?
The review is scoped to your situation, and its fee is agreed before work starts. If you choose us to lead the fix, that work is scoped and quoted separately after the review.
Is the review independent?
No, and we say so in the scope. We would like to lead any fix, so the review is written to stand without us, and your board should not rely on it as independent assurance. If you need independence, choose an Independent AI Review.
What if the answer is to stop?
Then the plan says what to keep, what to close and how to close it cleanly. Stopping a weak initiative early releases budget and people for the ones that can work.
Do we have to use you for the fix?
No. The plan is written so any firm, or your own team, can carry it out.
Can you take over a system a vendor built?
Yes, with your written authority to access and modify the vendor's code and configuration. The scope excludes findings of fault for litigation, and contractual questions go to your counsel.
How we work, in writing
A decision in weeks.
Bring us the initiative and its business case.
The first conversation is confidential, without obligation and without charge.