A non-tech firm can begin with AI without a vendor in the room. The order is: frame the question the business is actually trying to answer; scope a small honest pilot the firm can run without a long-term contract; decide what AI belongs in the business and what does not; and write the reasoning down so the board, the operator, and the eventual vendor can all read it. Done in this order, the first AI engagement is a decision the firm owns. Done in any other order, it is a decision the vendor owns and the firm pays for.
Why the vendor goes first by default
The path of least resistance for a non-tech principal who has been told they need AI is to call a vendor. The vendor has a deck, a pilot offer, a case study, and a sales engineer who can speak the principal's language. The deck is professional. The pilot is free or near-free. The case study is real. None of these things are dishonest. They are all, however, optimised for the same outcome: a multi-year contract before the principal has the language to evaluate it.
The trouble is not the vendor. The trouble is that no one in the room has been hired to ask whether the question the vendor is answering is the question the business is actually trying to ask. The principal cannot ask it — the principal does not yet have the vocabulary. The internal team, if there is one, is too close to the operational pressure that prompted the call. The board wants a paragraph in the next deck that says the firm is AI-enabled. Everyone is reasonable and everyone is biased toward signing.
A senior counsel, sitting on the firm's side of the table, is the cheap part of the engagement. The expensive part is realising eighteen months later that the wrong thing was bought.
Frame the question first
Before any vendor is briefed, the question itself has to be framed in language the operator can defend. We need AI for customer support is not a question. We need to reduce the time-to-first-response on tier-one tickets without degrading resolution quality, on a queue that doubles in December, with no addition to headcount — that is a question. The first sentence is a vendor brief. The second is a business brief.
Most firms can produce the second sentence inside a single working session, with someone outside the firm asking the questions. The work is not hard. It is unfamiliar. Once it is on paper, the conversation with vendors becomes a different conversation entirely: it is a conversation about whether a specific product solves a specific problem, instead of whether the firm should be doing AI at all.
Scope a small honest pilot
The right pilot is the smallest one that produces a defensible read. It is bounded in time (six to twelve weeks), bounded in scope (one queue, one workflow, one decision), and bounded in commitment (no multi-year contract, no exclusivity, no auto-renewal). The pilot is designed to answer one question: does this approach, for this firm, on this workflow, do better than the existing baseline by enough margin to justify the operational change?
The pilot is not designed to prove the vendor's product. The vendor's product may or may not be the right one — the pilot tells the firm, not the vendor. If the vendor refuses a six to twelve week pilot with a clear exit, the vendor is telling the firm something important about the contract they intend to write. Listen.
The cheapest AI mistake is the one made before any code is written. The most expensive is the one defended for two years afterwards because the contract has nineteen months to run.
A pilot that ends with the approach worked, here is the next phase is a good outcome. A pilot that ends with the approach did not work, here is what we learned about the workflow, and the firm is no worse off is also a good outcome. A pilot that ends with we have signed for three years and we will figure out whether it works during the contract is a failure of the engagement, regardless of what the vendor's slides say.
Write the readiness memo
The deliverable from a readiness engagement is a single written document the principal can keep, hand to the board, and use as the framework for every AI proposal that comes in for the next two years. It contains: the firm's actual question, in business language; the firm's current AI footprint and what is real about it; the criteria the firm will use to evaluate any AI proposal; the workflows where AI is plausibly useful, and the workflows where it is not; the vendors the firm has spoken to and how they were evaluated; and a small set of named pilots, scoped, with named exits.
The memo is short. The memo is dated. The memo is signed by the principal — not by the vendor, not by the consultant. The principal owns the framework. Future vendor pitches are run against the framework, not against the principal's intuition under sales pressure.
What this is not
It is not anti-vendor. The right vendor, on the right workflow, after the right framing, is exactly what the firm wants. The point is the order: framing first, pilot second, vendor selected third, contract signed fourth. Most firms run this in reverse and pay for the order error in the contract.
It is also not a sales pitch for in-house AI. Building in-house, for most non-tech firms, is the wrong answer. The right answer is usually a small number of well-chosen vendors, evaluated honestly, with the firm's own decision framework underneath. A senior counsel can write that framework. The vendor cannot write it; the vendor is, structurally, the wrong person to write it.
If the firm is being approached by AI vendors, asked by the board for an AI strategy, or pulled toward a pilot it cannot honestly evaluate, the right first step is not another vendor call. It is a written readiness memorandum the firm will own afterwards. Write to the studio. We sit on your side of the table and we take no commissions from any vendor.