Acceptance Criteria

Family. The survey belongs to the family of factual-findings engagements (ISRS 4400 revised). It reports what the elements establish at a date. It expresses no opinion, issues no certificate, and produces no score or risk level.

Standing. No professional body licenses or guarantees the author. The method is published, the findings are traceable to the elements they rest on, and both are open to attack.

Unit of work. One workflow: the one the director designates. Not a department, not a tool inventory, not everything the organisation runs.

The AI reaches the European market. Where the organisation is established does not matter, and neither does the nationality of its clients. What matters is the market the system reaches.

A single workflow is designated. One output, one recipient, named by the director. If the boundary cannot be drawn in one sentence, there is no engagement yet.

Dated elements exist and can be produced. Screenshots, invoices, tickets, exports, message threads, contracts. The survey measures the distance between what is declared and what the elements establish. Where nothing was recorded, nothing can be read, and we say so rather than infer.

The director is the interlocutor. For the Baseline, the director alone. For the Full Map, the director and three to four of the people who run the workflow. An intermediary may introduce the engagement; an intermediary may not stand in for the director.

A conclusion is required in advance. A case where the answer is known before the elements are read is not a survey.

A certificate, an opinion or a score is wanted. We produce none of the three. A request to have the survey serve as evidence of conformity toward a third party is refused, because the document does not carry that and would be misread as carrying it.

No element can be produced. A workflow that has left no trace at all is not one we can survey. We will say what would need to exist; we will not reconstruct it from conversation.

The scope will not close. A case that grows during scoping, or that resists designation of a single workflow, is refused rather than surveyed partially and reported as if whole.

Independence would be compromised. Two of the three possible issues of a Baseline open a paid engagement whose price is public. Where our interest in the outcome cannot be disclosed alongside the finding, we do not take the case.

The time available does not allow the work. A deadline that forces us to shorten the reading of elements is refused. We would rather decline than deliver a survey whose limitations are the schedule's.

Public errata. An error found in a delivered document is corrected in a public errata log, under its date, with what was said and what is now said. A delivered document is never silently amended.

Refund, capped. Where an error changes what the document establishes, the engagement is refunded, capped at the price paid for it. That cap is the whole of it: we never owe more than we were paid, and we say so here rather than leave it to be discovered in a dispute.

Opening date. The log opens with the first delivered survey. Until one is delivered there is nothing to log, and we would rather say that than publish an empty page as evidence of rigour.

Each quarter. How many engagements were accepted, how many were refused and on which ground, and how the delivered surveys concluded. The denominator is published with the figures.

Opening date. The series opens with the first delivered survey. Publishing a rate over zero engagements would be a rate about nothing.