An overseas brand asks an AI assistant to compare three production proposals from China. Each promises ten videos, one shoot day and two revision rounds. The spreadsheet looks comparable. The work may not be.
One supplier could mean ten openings cut around an approved script; another could mean ten scripts and settings. A useful service catalog must preserve those distinctions when a human or a machine extracts an offer. Otherwise faster comparison simply makes a scope mismatch easier to miss.
A media-buying signal—not a film-production standard
On September 22, 2026, IAB Tech Lab announced AAMP 3.0 and OpenProposal, describing agent-readable media products and proposals, including public packages alongside custom proposals. Its announced comment period runs through October 22, 2026. These facts were checked on October 18.
The official OpenProposal repository README labels it an unratified draft without expected production stability; machine-readable schemas remain unpublished (checked October 18, 2026).
The scope is media buying, not video-production procurement. It does not establish that brands generally buy production through agents, nor that ONCE has adopted the protocol. ONCE’s inference is a practical one: as proposals become easier to retrieve and compare, the conditions that make an offer valid need to travel with it.
Count the work before comparing the count
Consider a hypothetical product-video brief. A group of deliverables is built from approved footage in one route and requires new location and talent work in another. Identical file counts would conceal rather than resolve that difference. This is not an ONCE package or a market-price example.
The same problem applies to a shoot day. An offer might assume a ready sample, a confirmed action and a single setting. An unverified prototype or a different performance requirement could invalidate that assumption. The first question is whether the offer fits this brief—not which number is lower.
The current VID Project page provides a useful, limited peer observation. It scopes work around goals, deliverables, timeline and production requirements; its list of activities says “may include,” and raw footage is delivered when included in scope. The undated page was checked on October 18, 2026. This is service positioning, not evidence of results or value for money. Supplier capability and purchased scope remain different things.
The ONCE Condition-Preserving Card
ONCE proposes a compact record for each comparable service item. It is not an OpenProposal field mapping or a claim of protocol compatibility.
| Record | What must remain attached |
|---|---|
| Outcome | The production task and verifiable deliverables, not a label such as “premium content.” |
| Preconditions | What must be ready: sample, approved script, usable footage, setting or other required inputs. |
| Unit | What a video, version, scene or shoot day means, and the scope on which fees depend. |
| Unresolved scope | Confirmed inclusions, explicit exclusions, items needing clarification and items that do not apply. |
| Version and source | Revision, review date, explanation owner and original scope reference. |
Do not collapse the four scope states into a blank cell. Unknown does not mean excluded. Excluded does not mean available as an add-on. Not applicable does not mean free. An adaptation-only service, for example, should not acquire an optional actor simply because the catalog needs to fill an empty field. A new live-action requirement may call for another service and a fresh scope assessment.
The README also makes line items self-contained. Applied cautiously to production, each extracted offer should keep its own conditions rather than rely on a separate general disclaimer.
Change one condition and replay the recommendation
Before building a quoting assistant, review the catalog manually. Put the same service against the original brief, then a brief with one condition changed. Should the recommendation remain valid, request clarification or leave the shortlist? This is a proposed offline check; no test or result is reported here.
Suppose approved footage is available for social cutdowns. Change the brief to require a newly filmed human demonstration. Adding a talent tag is not enough: capture, casting, execution and usage arrangements may need reassessment. If the catalog still returns the original scope, it has not preserved the dependency.
Now change “usable footage confirmed” to “footage availability unconfirmed.” That should create a question, not an assumption of free preparation or an unchanged delivery date. Ask the supplier to identify which precondition prevents a firm recommendation.
A catalog can support discovery, qualification and clearer enquiries. It cannot by itself judge a distinctive creative approach, confirm real availability or settle the final project agreement. Complex storytelling, unusual locations and unverified products may still need a custom proposal.
Start with one maintainable item
For a brand coordinating a China-based team remotely, ask shortlisted suppliers to complete these five records against the same brief. Separate what your marketing team provides from what the manufacturer must confirm and what the production partner actually supplies. That is more useful than requesting a longer list of capabilities.
No party needs to expose confidential costs, client information or proprietary production methods to make an initial scope intelligible. Share only the information authorized for the comparison tool and the procurement task.
For ONCE, a potential long-term asset is the relationship between recurring requests, prerequisites, reasons for scope changes and the final agreed work. A future catalog or enquiry workflow could use those records, if validated. This is not a claim of existing automated quotation, adoption or margin improvement.
Begin with one service item and a changed brief. A reliable buying aid should know when it cannot yet compare—not merely how to produce a ranking.
