Count the work that survives
A cheap generation is not a cheap approved asset. The gap contains briefing, reference gathering, selection, revisions, retouching, legal review, accessibility, exports and the conversation in which somebody finally accepts the result. Our budgeting notebook treats those as parts of production, not inconvenient costs outside the AI story.
Our position
Price a useful deliverable and name the assumptions behind it. How many distinct routes will be explored? How many approval rounds are included? Who owns the product-accuracy check? What constitutes a change of brief? A small set of agreed answers can protect both the client and the production team from a seemingly unlimited supply of almost-finished work.
Faster exploration is valuable when it resolves an uncertainty. It is less valuable when it shifts that uncertainty to a reviewer. Do not use the count of generated images as a proxy for progress, and do not treat time saved on one stage as a saving across the whole project.
A workable review
Track generation, selection, finishing and approval separately on the next job. Note which revisions were caused by technical failures and which came from unresolved strategy. Then compare approved outputs of similar scope. The result will be a modest local benchmark, not an industry-wide promise—but it will be something the team can use when it writes the next estimate.
Sources & limits
Source facts and editorial interpretation are distinguished in the text. A brand or agency account is not an independent effectiveness audit.
- Ad Field Notes — Production budgets research notebook — source publication date not stated.
Send a correction with the passage and supporting source.