Skip to main content
← All posts
By Abhay Muttreja, Founder, SolvTreeAI

Which version of the campaign did the client approve

AI makes creative versions cheap. Keep one authoritative asset record and check every published file against the version that was approved.

·4 min read
  • AI Marketing
  • Creative Approvals
  • Asset Management

The short answer

The approved version is the one recorded as approved, against a specific file, by a named person, at a specific time. Not the latest file in the folder, not the one in the last email, and not the one the designer remembers. When AI tools can produce ten variants in a minute, campaigns need one asset record per piece, and a final check that what goes out matches what was signed off.

Key points

  • AI lowers the cost of making versions, which raises the cost of losing track of them.
  • Comments are not approvals. An approval names a version and a person.
  • Keep one record per asset that holds the approved version and its history.
  • Before anything is published or sent to print, compare it with the approved file.
  • Agree who can approve which kinds of change: copy, claims, imagery, pricing.

How version confusion actually happens

Campaign confusion is rarely one big mistake. It is a series of small, reasonable steps:

  • A client comments on a WhatsApp thread, another on email, a third in a shared document.

  • A designer exports “final”, “final2” and “final_v3_approved” into the same folder.

  • An AI tool regenerates an image to fix one detail and subtly changes another.

  • A media buyer uploads the file that was attached to the most recent message.

Each person acts on the version they saw. Nobody did anything wrong in isolation, yet the published ad is not the one the client approved. In real estate, where creatives carry prices, payment plans and regulatory details, that gap is more than embarrassing.

The authoritative asset record

For each asset in a campaign (an ad, a brochure page, a hoarding, a video cut), keep one record with:

  1. Asset ID that never changes, such as “LAUNCH-FB-04”.

  2. Current version with a version number and file link.

  3. Version history, with who changed what and why.

  4. Approval entries: approver, version approved, date and time, and scope (“copy approved, image pending”).

  5. Approved claims: price, offer, dates, project or registration details, each with its source.

  6. Distribution log: where it was published or sent, which version, by whom.

This can live in a project tool, a structured sheet, or a purpose-built system. The tool matters less than the rule: if it is not in the record, it is not approved.

An illustrative example: a residential launch campaign

An illustrative example, not a real client or project: consider a hypothetical real estate developer in Pune launching a residential tower, with an agency producing social ads, a brochure and site hoardings. The agency uses AI tools to draft copy variants and generate lifestyle imagery, with human designers finishing each piece.

The asset “LAUNCH-FB-04”, a carousel ad, goes through these steps:

  • v1: copy states “2 and 3 BHK from ₹78 lakh, all inclusive” with an AI-generated balcony image.

  • Comment, by the developer’s sales head on email: “Price is ₹78 lakh plus charges, not all inclusive. Balcony image shows a sea view; this site has a hill view.”

  • v2: the agency fixes the copy and regenerates the image. The new image shows a hill view, but the regeneration also adds a swimming pool on the balcony level that the project does not have.

  • Comment, by the marketing manager on WhatsApp: “Looks good.”

  • v3: the designer catches the pool, edits it out, and uploads v3.

Which version is approved? The marketing manager’s “looks good” was about v2, and it was not recorded as an approval. The sales head never saw v3.

With an asset record, the agency logs: “v3 submitted, changes: price wording, image regenerated, pool removed.” The developer’s approver, who is named in the record as authorised for price and project claims, approves v3 with the scope “copy and image”. The record links the approved file.

Before the media buyer uploads, a final check compares the file to be published with the approved v3 file: same file, same price text, same image. A mismatch blocks the upload and returns it to the asset owner. The distribution log records that v3 went to the platform on a given date.

Three weeks later the developer changes the offer. The record shows exactly which assets carry the old price and where each was published, so they can all be replaced.

An approval checklist for AI-assisted campaigns

  1. Does every asset have one ID and one record?

  2. Is each approval tied to a specific version number?

  3. Is the approver authorised for that type of change (price, legal or project details, imagery, tone)?

  4. Are factual claims listed with their source and checked against the latest approved information?

  5. After any AI regeneration, has someone checked the whole asset, not only the part that was meant to change?

  6. Is there a final comparison between the file being published and the approved file?

  7. Does the distribution log show where each version went?

  8. When an offer changes, can you list every live asset that carries it?

Where legal or regulatory wording is involved, have the developer’s own adviser confirm what each asset must carry.

Where SolvTree fits

This maps to our AI-first marketing work: social, design and campaigns produced with AI systems and human review, with approvals recorded against versions. We would start with one campaign and its approval flow, agree who signs off what, and make the final check routine.

Next step: Book a free 30-minute operational audit and we will map one campaign approval flow and its final asset record with you. The report is yours to keep.

Frequently asked questions

What makes a campaign asset version approved?
An authorised person’s approval must be recorded against a specific version, file, time and scope. A recent file or an informal comment does not establish approval of later changes.
What should be checked after AI regenerates an asset?
Review the whole asset, including claims and imagery outside the intended edit. Before publishing, compare the outgoing file with the approved version and record where that version is distributed.