top of page

Review and Approval Workflow for AI Campaign Assets: Rounds, Versions, Sign-Off

Oct 23
7 min read

Updated: 14 hours ago

A workable review and approval workflow for AI campaign assets rests on five rules: one named approver per side, a fixed number of feedback rounds, one consolidated comment list per round, a version label on every file, and a written sign-off that locks each asset once approved. Generating another version costs little, so without these limits feedback multiplies and nobody can say which file is final. In practice the studio delivers numbered batches. The agency gathers every comment from brand, legal and regional teams into a single list and sends it back once. The studio answers each point, delivers a new version, and the approver confirms it or asks for specific changes. Many projects plan two rounds on the base images and one on the variants, though scope changes that. After sign-off an asset changes only through a new, documented version. You do not need special software for this. A shared delivery folder, a naming convention and one comment list are enough.


The rest of this article explains each rule, so that agency and brand teams can write them into the brief before production starts.


Why approval needs its own rules in AI production


On a classic shoot, the day itself forces decisions: the light changes, the model leaves, the set of images is fixed when the studio closes.


Generated and hybrid production removes that pressure. A new version of an image takes far less effort than a reshoot, so reviewers feel free to ask for one more variation. Each request is reasonable alone. Together they produce rounds that never close and comments that contradict each other. The review process has to supply the discipline that the shoot schedule used to supply.


This article covers the stretch from first delivery to sign-off. How the project starts is described in What to Put in a Brief for an AI Production Studio, and how studio and agency divide the work is covered in How AI Production Studios Work With Ad Agencies.


Name who comments and who decides


Separate two roles that often get mixed up.


  • Reviewers give feedback. They can be brand managers, art directors, legal, regional marketing teams or the model's representation when likeness is involved.

  • The approver decides. One person per side, usually one at the agency and one at the brand, confirms that an asset is final.


Write both lists into the brief. Reviewers send their comments to one coordinator, not to the studio. The approver resolves conflicts between reviewers. If two reviewers disagree about the colour of a coat, the studio should not have to pick a side, and it should never receive both opinions as two separate instructions.


Plan the rounds before production starts


Agree the number of rounds in advance and tie each one to a purpose.


1. Base assets. The first round covers the hero images and the reference set that everything else will be built from. Corrections here are cheap. A correction made later has to be repeated in every variant that inherited the mistake.


2. Variants. The second round covers formats, market versions and seasonal changes. Reviewers check consistency against the approved base, not new creative direction.


3. Final check. A short last pass on the delivery files: crops, file formats, labels where disclosure applies.


The timing of these rounds, and how long each one realistically takes, is covered in How Long Does an AI Campaign Take?. A small campaign may merge two rounds; a multi-market rollout may need one per market.


A rule that helps in every setup: new creative ideas are not feedback on a variant. If an idea arrives in round two, the approver decides whether it becomes a change request with its own scope, or waits for the next campaign.


Bundle feedback instead of sending it piece by piece


Feedback that trickles in as separate emails creates versions nobody planned for. Ask for one comment list per round, built like this:


  • each comment refers to a specific file and version

  • each comment says what to change, not only what feels wrong

  • each comment has an owner on the review side

  • comments that conflict are resolved by the approver before the list goes to the studio


The studio replies to the same list and marks each point as done, declined with a reason, or needing a decision. That reply is part of the record. It shows later why an image looks the way it does.


Label every version


Every file needs a name that tells a stranger what it is. A simple scheme is enough: campaign, asset number, format, market and version, for example in the order they appear in your variant matrix. Two rules keep the scheme honest.


  • A version number changes whenever the image changes, even by a small retouch.

  • Nobody overwrites a delivered file. Corrections get a new version, and the earlier one stays available.


Without this, a team can find out at launch that two people approved two different images under the same file name. With it, you can answer which version a client saw and which one went live.


Keep the source of each generated asset together with its version: the reference images and product data it was built from, and the instructions the studio worked to. When a later question arises about product accuracy, that trail answers it.


What sign-off means and what it locks


Sign-off is a written statement from the approver, naming the asset, the version and the date. An email reply that lists file names is enough. A thumbs-up on a preview is not.


State in the brief what sign-off locks:


  • The image itself. After sign-off it does not change. A new idea means a new version with a new sign-off.

  • The base for variants. Variants are built from signed-off base assets only. If a base asset changes after variants exist, the project owner decides whether the variants are rebuilt, and what that costs in time.

  • The rights position. Sign-off records that the talent consent and usage terms in the brief cover this asset.


Also decide who may request a change after sign-off, and how. A short change request with a reason, a scope and a named decision-maker is usually enough. It keeps late requests possible without turning each one into an informal reopening of the whole round.


Keep one record per asset


A record per asset gets built as the project runs, not at the end. For each delivered asset it should show:


  • asset name and final version

  • which round or rounds it went through

  • the approver and the date of sign-off

  • the source material it was built from

  • whether talent consent covers this use

  • where a disclosure label is required, whether it has been applied


Fill it in as each round closes; it is hard to reconstruct afterwards. Legal teams and brand owners ask for exactly this information when a market raises a question. The disclosure point depends on the markets involved, and the rules differ between them, so this article does not set out what the labels must say. Review of the images themselves, covering skin, hands, fabric and product detail, is described in Quality Checks in Hybrid AI Production.


How The Bridges AI works


The Bridges AI builds hybrid campaigns from one controlled shoot with real talent and real product. Because generated assets trace back to that source, each version can point to the material it came from, which makes the record per asset easier to keep.


Frequently Asked Questions


How many review rounds does an AI campaign need?

Plan at least two: one on the base images and one on the variants. A final check on the delivery files is often a third, shorter pass. Scope, the number of markets and how many parties must approve change the count, so agree it in the brief instead of leaving it open.


Who should give the final approval on AI-generated campaign assets?

One named person on each side, usually one at the agency and one at the brand. Other reviewers give feedback, but only the approver confirms that an asset is final. Several people approving the same asset in parallel is the most common reason rounds do not close.


How do you manage versions of AI-generated images?

Give every file a name that carries campaign, asset, format, market and version. Change the version number whenever the image changes, and never overwrite a delivered file. This lets you show later which version was reviewed, approved and published.


Do you need special software to run reviews with an AI studio?

No. A shared delivery space, a naming convention and a single comment list per round are enough. The process decides whether reviews stay under control: named approvers, fixed rounds and written sign-off.


What should happen when a client wants changes after sign-off?

Treat it as a change request with a reason, a scope and a named decision-maker. If the change affects a base asset, the request must also say whether the variants built from it are rebuilt. This keeps late changes possible while the earlier approved version stays on record.


What belongs in the record for each approved asset?

The asset name and final version, the rounds it went through, who approved it and when, the source material it was built from, whether talent consent covers the use, and whether a disclosure label has been applied where one is required.


Planning a campaign with AI imagery or video? Request an AI campaign.


About The Bridges AI

The Bridges AI is a hybrid AI creative agency: a creative studio and production partner for fashion, beauty, luxury and hospitality brands and for the agencies that work with them. It combines booked real talent, real products and human creative direction with AI production, and offers licensed digital twins of real models, only of models under contract. The Bridges AI works in strategic partnership with Most Wanted Models®, the German model agency founded in 1997, which brings nearly 30 years of experience in bookings, buyouts and usage rights.


Written by the editorial team of The Bridges AI. Last updated October 2026.


Comments


bottom of page