Moukangwe & Co., home

AI use needs a release standard

An approved AI tool still leaves important production decisions unresolved. A release standard connects permitted inputs, verification, supplier responsibilities and final approval, so that faster drafting remains supported by evidence and a clear allocation of professional judgement.

Moukangwe & Co. · Published Nov 16, 2026

Permission to use an AI tool settles only one part of a professional firm’s production process. The firm must also decide which information may enter the tool, which claims require verification and who can authorise the finished material. These decisions determine whether a faster drafting process produces work the firm is prepared to publish.

We advocate a release standard organised around the material being produced. It should connect permitted AI use to the firm’s existing responsibilities for accuracy, confidentiality and attribution. A responsible reviewer needs clear decision rights, access to the supporting evidence and sufficient time to exercise judgement.

Govern the use case as well as the tool

NIST’s AI Risk Management Framework treats governance as a continuing responsibility across the AI lifecycle and assigns importance to documented policies, roles and accountability (NIST, 2023). For professional material, the practical unit of control is the use case: the information supplied, the purpose of the draft and the consequences of releasing it.

The same approved tool might support an internal outline and a public service description. Those activities involve different assertions and audiences. A firm can permit assistance with structure while requiring direct verification of every claim about its credentials or delivery capability. Approval of the tool should make those distinctions easier to apply.

Input rules must be equally specific. A drafter needs to know whether a client document, an unpublished project account or a personal biography may be supplied in a particular working environment. Restrictions should identify who can decide an uncertain case. General instructions to exercise care transfer the unresolved decision to the person least equipped to make it.

Make review an exercise of judgement

Consider a hypothetical advisory firm drafting a new service page with AI assistance. The proposed wording describes experience in a specialist field and implies that a team is available to deliver the service. A polished draft can make those propositions appear settled before anyone has checked them.

The reviewer should separate the claims and examine their basis. Which assignments support the experience statement? Who performed the work? Does the present team have the capability being offered? Are the description and attribution consistent with the evidence? A reviewer who checks grammar alone has completed a narrower task than the firm’s release decision requires.

This exposes an important allocation problem. The communications team can control presentation while a practice lead holds the evidence of expertise. A release standard must bring those responsibilities together. Naming a human reviewer adds little if that person lacks access to the source material or authority to hold publication.

Preserve evidence that helps explain the result

Documentation should allow the firm to understand how a material assertion entered the published work and how it was checked. The useful record connects the sources considered, the consequential changes made during review and the version authorised for release. Record the tool and relevant operating context where they help explain the production decision.

The amount retained should be proportionate to the work. Keeping every interaction without a defined purpose can make the relevant evidence harder to retrieve and create further information-management obligations. Conversely, retaining only the finished page can conceal the point at which an unsupported claim was introduced. The firm needs a reasoned retention choice and a record that people can actually use.

Carry the standard through the supply chain

An agency’s involvement changes the production arrangement; the firm still needs an accountable owner for its published assertions. The brief should state the permitted AI use, the information restrictions and the evidence to accompany delivery. Acceptance should include the material claim checks required by the firm.

There is a commercial trade-off here. If a supplier is paid for rapid output while verification is treated as an unpriced extra, the contract and the control work against each other. Scope, fees and delivery time should provide for the review the firm expects. Otherwise the final reviewer inherits an avoidable conflict between the publication deadline and the standard of evidence.

Review what the standard achieves

Oversight should examine a small sample of released work, including supplier-produced material and items that required exceptions. Look for unsupported assertions, unclear attribution, missing release decisions and repeated corrections after publication. These findings show where the production process needs attention.

Where the tool, its operating conditions or the type of material changes, reconsider the relevant controls. A standard that was workable for internal drafting may need revision when the same workflow begins producing public professional claims. The review should lead to an assigned action, a revised instruction or a change in delegation.

The objective is a production process in which the firm can explain its decisions and stand behind the material it releases. Speed has commercial value when verification keeps pace with it. The release standard makes that responsibility explicit at the point where a draft becomes the firm’s public statement.

NIST (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0). NIST AI 100-1. Gaithersburg, MD: National Institute of Standards and Technology. See section 5.1, Govern.

Discuss this position

If this position bears on a decision your firm is facing, screening is the place to start.