Back to home

The methodology

Proving a licensed professional stayed in responsible charge of AI-assisted work.

Six gates, from the moment an AI tool enters a project to the moment a deliverable is sealed. Each one leaves a record, so what a carrier or a board eventually asks about can be shown rather than asserted.

Records, not verdicts. The methodology documents that a licensed professional exercised responsible charge. It does not score the tool, rate the vendor, or judge the engineering — that judgment is the professional’s, and stays theirs.

1 · Why gates

Accountability is a sequence, not a signature.

Responsible charge is not something exercised at the end. It is the direction given before the work starts, the judgment applied while it is produced, and the review that happens before anything enters a sealed deliverable. A signature at the end records the conclusion of that sequence — it does not substitute for it.

So the methodology is a sequence too. Each gate is a point where something is decided and recorded while it is still true, rather than reconstructed afterward from memory. A record assembled after a claim arrives is worth considerably less than one written as the work was done.

2 · The flow

Six gates, in order.

This is the shape of the work. What each gate asks for in detail — and what it takes to satisfy one — is part of the working platform.

Gate 1

Register & declare

Say which AI is in use before it touches the work.

Gate 2

Define task & competence

State what the AI is being asked to do, and who is competent to judge the result.

Gate 3

Qualify the tool

Establish the tool is fit for this task, on this project.

Three ways in

· Existing tool

· Built for this project

· Reusable tool, re-qualified

Gate 4

Direct & develop

The professional sets the approach and stays in charge while the work is produced.

Gate 5

Verify

Check the output against the professional’s own judgment before it goes anywhere.

Gate 6

Seal & preserve

Sign, and freeze the record of who was accountable and how.

A tool can enter the work in one of three ways, and gate 3 is where that is settled. The rest of the sequence is the same whichever way it arrived.

3 · What each gate records

Written while it is still true.

  1. 1

    Register & declare

    A declaration: which AI is in use on this deliverable — or that none is.

  2. 2

    Define task & competence

    A competence statement: the task, and the licensed professional’s basis for judging the result.

  3. 3

    Qualify the tool

    A qualification record: that the tool was found fit for this task, on this project — and by which of the three ways it entered.

  4. 4

    Direct & develop

    A direction record: the inputs, assumptions, and acceptance criteria the professional set, and their engagement as the work was produced.

  5. 5

    Verify

    A verification artifact: the independent check performed, and the disposition — adopted, corrected, rejected, or overridden.

  6. 6

    Seal & preserve

    A frozen seal record: who was accountable, the licence held at signing, and the artifacts behind the decision.

None of these is a score. Each is a fact about what happened and who was accountable for it, recorded at the time and kept.

4 · The record

What comes out the other end.

For each sealed deliverable that AI touched, the gates produce one responsibility-chain record: who was accountable, with the licence they held at the moment of signing; which AI touched which deliverable and the role it played; that the professional directed and reviewed the work before it was sealed, rather than after; and all of it time-stamped and preserved.

It is proof of accountability, not a judgment of the engineering. The professional’s judgment stays theirs — the record makes the fact that they exercised it provable.

5 · What this does not claim

Evidence, not coverage. Records, not judgment.

  • This does not place, price, or guarantee insurance. It produces evidence a carrier or broker may find useful; the coverage decision is entirely theirs.
  • This does not certify that an AI tool is safe, and it does not rate or rank tools. Whether a tool is fit for a given task is decided by the professional accountable for the result.
  • This does not review the engineering. A complete record says a licensed professional was accountable and exercised judgment — it says nothing about whether that judgment was correct.
  • These records are the firm’s own — declarations and artifacts made by the accountable professional, not independently verified or attested by us. They make responsible charge visible and reviewable; they are not a third-party audit of it.
  • Licensure decisions, seal placement, and responsible-charge determinations remain the sole responsibility of the licensed professional.

6 · Status

Published as research, in active development.

What is published here is the framework: the gates, their order, and what each is for. The detail of what satisfies a gate is part of the working platform and is being built in conversation with the profession — with licensed practitioners, and with the carriers and brokers who price this risk.

It is written by a practicing professional engineer who seals work under the same rules it documents, and it will change as the standards and the law it tracks change.

If this is your world, I’d value the conversation.

The research is free to read and published in full. The tooling that applies this methodology is in development.