The inversion
Traditional production is front-loaded: crew, location, talent, equipment. Producing a second version is expensive, so the process optimises for getting it right once, and enormous planning effort exists to avoid reshoots.
AI production inverts this. Generating an alternative is close to free, so the constraint moves to deciding which alternative is good. The scarce resource becomes attention, and a pipeline that does not budget for judgement produces a large volume of mediocre output very quickly.
Where the cost actually sits
Direction, before anything is generated: what the piece is, who it is for, what it must communicate in its first seconds. Skipping this produces output that is technically fine and strategically pointless.
Selection, after generation: reviewing candidates and choosing. This scales with volume and does not automate, because it is the taste judgement being paid for.
Revision, which is genuinely cheaper than traditional reshoots but not free — each round still costs review attention.
Compliance and provenance, which is a real line item now rather than an afterthought: disclosure, consent records, licensing and retention.
Why per-asset pricing misleads
A cost-per-asset figure is close to meaningless without knowing how many were generated to produce it and how much direction preceded them. A studio quoting a low per-asset price with no selection process is selling generation, which the client could buy directly from a tool for less.
The value in an agency arrangement is the direction and the selection. Pricing that hides those makes it impossible to compare offers meaningfully.
What we will publish, and what we will not
Quantivo's creative benchmark tracks production time, cost per delivered asset, approval rate and revision rounds on our own products. When there is enough data to report with a stated method, we will publish it with its limitations attached.
We are not publishing figures before then. An unsourced cost-per-asset number is trivially easy to produce and worth nothing, and this is a space with a lot of them.