How engagements work
The commercial terms are part of the method.
How an engagement is scoped, priced, contracted, and exited is not administrative detail that follows the real decision — it is most of what determines whether an AI programme ends on evidence or on politics. So it is published here, in full, before you talk to us.
Nothing below is an opening position to be negotiated back from. It is how we prefer to work, and the reasoning is included so you can tell the difference between a principle and a preference.
From first email to first gate
Five steps, with the honest duration of each. Nothing here requires you to commit to a programme in order to find out whether there is one.
- 01Reply within two working days
You write to us
A person reads it and replies with a point of view on what you sent — not a calendar link and a brochure. If your problem is not one we are the right firm for, that reply says so.
- 0230 minutes · no charge · mutual NDA offered first
A scoping conversation
We establish whether there is a real problem worth diagnosing, who owns it, and what evidence already exists. If there is not, we say so on the call rather than proposing a workshop to find out.
- 03Within one week of the call
A fixed-fee diagnosis proposal
Scope, price, dates, named people, and the exact artifacts you will receive. One document. No commitment beyond it, and no dependency on what happens after it.
- 042–4 weeks · fixed fee
The diagnosis runs
Inside your operation, against your data, with your people. It ends in the Diagnose exit-gate artifacts — including the evaluation thresholds any pilot would have to clear, written before there is anything to evaluate.
- 05Yours either way
You decide at the gate
Proceed to a pilot with us, take the diagnosis to another firm, run it in-house, or stop. All four are ordinary outcomes. The artifacts are yours in every one of them.
How it is priced
- The diagnosis is fixed-scope and fixed-fee
- You know the price, the dates, and the deliverables before you commit. It is deliberately sold as a complete thing rather than a qualification exercise for a larger programme — an entry point that only pays off if you buy the next phase is not an entry point, it is a funnel.
- Delivery is priced one phase at a time
- Each phase is priced against the gate it has to clear. You buy Pilot, not a programme with a Pilot at the front of it. The commercial unit and the evidence unit are the same size on purpose: it is what makes stopping cheap enough to actually do.
- Fixed price where the work is bounded
- Our standing preference is a fixed price per phase. Where a workstream is genuinely unbounded — an unmapped data estate, an integration into a system nobody can describe — we say so and price that part openly, rather than burying the uncertainty in a rate card and calling it flexibility.
- No leverage pyramid
- Small teams of senior people who have shipped in enterprise conditions. There is no bench whose utilisation your programme exists to solve, and no gap between the people who pitch and the people who deliver.
- Change is a written change
- Scope moves when both sides agree it should, in writing, with the price and the gate criteria updated together. Scope does not move because a status meeting drifted.
What the paper says
A master agreement plus a statement of work per phase. The parts that differ from what you will be used to are these.
- Gate criteria are a schedule to the contract
- The thresholds a phase must clear are annexed to the statement of work and signed before the work starts. Missing them has a defined consequence written in the same document — which is the entire mechanism that makes a gate a gate rather than a slide.
- Termination for convenience at every gate
- You can end the engagement at any gate, without cause and without an exit fee. We would rather lose a phase than hold a client by making leaving expensive.
- You own what we make
- Deliverables, code, evaluation suites, and documentation are yours on payment. We retain our pre-existing methods and tooling and grant you a licence to use anything of ours that ends up embedded in your deliverables — stated plainly here because the reverse is common and rarely highlighted.
- Confidentiality runs both ways, and publicity is opt-in
- Mutual NDA before material information. Our default is that the engagement is confidential: no logo, no case study, no anonymised retelling, unless you affirmatively ask for it. You are not buying a marketing asset for us.
- Liability, insurance, and audit are negotiated, not deflected
- We expect these clauses to be discussed properly. Where a requirement is disproportionate to the engagement we will say so and explain why, rather than signing it and hoping it is never exercised.
- Subcontracting is named or it does not happen
- We do not put a party you have not been told about onto your engagement. If specialist help is the right answer, you are told who, why, and under what terms before they start.
What your procurement team can ask us for
All of it available on request, most of it before you have committed to anything. If something on your framework’s list is missing here, ask — the answer will be what we hold, not what we intend to hold.
- Company registration and tax details
- Evidence of insurance to the levels your policy requires
- A signed data processing agreement and named sub-processor list
- Security questionnaire responses in your format, with named owners
- Sample master agreement and statement of work, including the gate schedule
- Conflict-of-interest declaration covering adjacent work
- References, spoken rather than written, where the referee has agreed to speak — we ask, we never assume, and we will tell you plainly where we do not yet have one for work like yours
- Supplier policy statements your framework requires, with a plain answer on which we hold and which we do not
Assurances we do not hold are published rather than deflected — the full list, including the certifications we lack, is on the security page.
Or take the consolidated version, built for your procurement file
What we need from you
The reciprocal half. An engagement that does not get these produces a worse answer, and we would rather agree them now than discover them in week two.
- One accountable executive
- Someone who can decide, not only convene. Gates need a signature from a person with the authority to stop the programme as well as advance it.
- Access to the operation, not the slides about it
- Time with the people who do the work, in the systems where the work happens. A diagnosis conducted entirely in workshops describes the organisation’s self-image, not its operation.
- Data access agreed before day one
- Or an explicit decision to run without it — which is a legitimate choice, provided the diagnosis is honest about what that costs in confidence and says so in the report.
- Named counterparts for security and data
- Identified at contracting, not discovered at the point they become a blocker. Most stalled phases stall here.
- A willingness to hear no
- The most valuable output we produce is sometimes a recommendation not to build. That only has value in an organisation prepared to receive it.
How you leave
At any gate, without cause and without an exit fee. What you keep is the same whether you leave at the first gate or the last: every artifact produced to that point, the code and evaluation suites in your repositories, infrastructure in your accounts, and a handover pack written for your engineers. In the default deployment shape that is already true on the day you go — your accounts, your repositories, your keys, nothing held only by us. Where you contracted the written exception of an environment we operate, the handover pack names every credential and configuration that transfers and the date it does, which is the migration date that exception had to carry before the work started. Designing for all of this from day one costs us a retention mechanism we would rather not have.
Straight answers
- Can you start next week?
- Rarely. Diagnoses are staffed with the senior people who will actually run them, so we would rather give you a date we can keep than one that sounds better on this call.
- Will you work on our paper?
- Usually, yes. We read it properly and come back with a short list of specific clauses rather than a wholesale redline, and the gate schedule is annexed either way.
- Can we skip the diagnosis?
- Sometimes. If you already hold a credible baseline and written evaluation thresholds, we will start at Pilot and say so. If you do not, skipping Diagnose means the pilot has nothing to be judged against, and we will decline rather than build something unjudgeable.
- What if the pilot fails?
- You receive a written no-go with the evidence, the reasoning, and what would have to change for the answer to change. That document is a deliverable, not an apology — it is what stops the same idea being re-piloted in eighteen months by a different team with the same result.
- Do you resell anyone’s platform?
- No. We hold no reseller margin and no vendor quota, which is why a recommendation to buy rather than build costs us nothing to make.
The next step is a 30-minute conversation.
Not a pitch, and not a discovery workshop. We establish whether there is a problem worth diagnosing — and if there is not, you will hear that on the call.