Advertising AI agent · Creative production
Creative-Versioning & Asset-Production AI Agent
Take a signed master and produce the size, format, language and market versions a campaign needs — inside the brand template, the platform spec and the rights a person cleared.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Rights sign-off
- Named person
- After launch
- Agent Care
What this agent does
Produces the versions, not the permissions
In
- 01
Take the signed master, the brand template and the version brief — sizes, formats, markets, languages and offers.
- 02
Read the rights record for every element in it: music, stock, talent, likeness, voice and font, with its terms.
Reason
- 03
Lay out each version to the template's safe areas and the platform's current spec, not a spec cached last quarter.
- 04
Swap copy, offer, price and legal line from the approved strings for that market, not from a translation it invents.
- 05
Check each version against the licence: territory, medium, duration, exclusivity and the markets talent may not appear in.
Decide
- 06
Hold any version whose rights cover, disclosure or claim differs from the master a person signed.
- 07
Route every offer, price, claim and legal-line change to the person who approves it in that market.
Out
- 08
Package for trafficking with the file spec, the naming convention and the flight dates the licence allows.
- 09
Retain the master version, the rights terms read, the strings used and every approval, asset by asset.
Product statement
The agent versions inside the rules and rights a person set. It does not clear a right, sign a master, or change what a version claims.
Example workflow
One version, end to end
- 1Master and brief receivedA signed master, the brand template and the version matrix — sizes, formats, markets, languages and offers
- 2Rights readMusic, stock, talent, likeness and voice terms for every element — territory, medium, duration and exclusivity
- 3Versions composedEach size built to the template's safe areas and the platform's current spec, with the legal line kept in frame
- 4Localised and checkedCopy, offer, price and legal line swapped from the approved market strings, then checked against that market's rules
No human action required
Stages 1 to 4 run without a person in the loop — the rights read, the layout and the localisation are finished before anyone is asked to look at a version.
Reaches the approver ready to traffic.
Held with the element and the term named.
Business affairs and brand approver
Business affairs clears the right; the brand or market approver signs anything the version changes about the offer, the claim or the legal line.
Approve · Amend · Send back
Approved — handed back
- 6Packaged for traffickingFile spec, naming, flight dates and the platform's ad-review requirements; released only against the approvals on file
- 7Outcome evaluatedSpec rejections, disclosure retention, rights-window breaches and localisation corrections, by market and format
Amendments
Every change made in review is counted against the version that caused it.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Clearing music, stock, talent or likeness rights.
- Signing the master a version is cut from.
- Extending a flight beyond the licensed window.
- Approving a price, offer or discount in a market.
Automation boundary
Agent acts unaided
- ✓Compose versions to the template's safe areas and the platform's spec.
- ✓Swap approved market strings for copy, offer and legal line.
- ✓Check every element against the licence terms on file.
- ✓Hold anything that changes the offer, the claim or a right.
Write actions run only inside the approval boundaries agreed during implementation. Clearing a right and releasing a version are not among them.
- Changing a claim, a superlative or a comparison.
- Removing or shrinking a legal line or disclosure.
- Using a generated or altered likeness or voice.
- Releasing a version to a platform or partner.
Example output
One version, annotated
Everything the agent proposes is attached to the master and the licence it came from.
Versioning output · single version
Illustrative example- Master
- Signed hero film
- Version requested
- Six-second cutdown, third market
- Music licence
- Read today
- Rights status
- Outside the licence
- Confidence
- 91%
- Legal line
- Retained at every size
As received
The master as it was signed and the brief as it was raised, with the date the licence was read.
Evidence used
- Sync licence, dated
- Talent territory grant
- Platform spec, dated
Why it is held
The sync licence covers two markets and this cutdown is scheduled in a third.
What the score decides
It decides how much an approver re-reads. A rights gap holds the version at any score.
Action
ApproveAmendSend backValue
Where AI adds value
The same four claims, placed at the point in the workflow where each one applies.
Where the value lands
Value 01 – 04
Every version requested
From the brief and the signed master
Apply the template and the licence as written
Use the template's safe areas, the platform's current spec and the rights terms on file — territory, medium, duration and exclusivity, read at production time.
Cut the resize-and-cutdown hours
Resizes, cutdowns, format swaps and market strings are produced before a designer opens the file, so review is a check rather than a rebuild.
Surface the rights gap before trafficking
Versions that fall outside a licence, drop a disclosure or change what the ad claims are held now, instead of being pulled out of a market later.
Build an evidence trail
Retain the master version, the rights terms read, the strings used, the spec checked and every approval — on both paths.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
DAM & creative tooling
- AEM Assets
- Bynder
- Frontify
- Creative Cloud
- Render APIs
Rights & talent
- Rights management
- Talent usage
- Music and stock licences
- Consents
Localisation
- Translation memory
- Market strings
- Legal-line library
- Native review
Agent
Creative versioning & asset production
- Reads rights
- Builds versions
- Holds changes
Trafficking & specs
- Campaign Manager 360
- Platform specs
- Ad-review policy
- Flight dates
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six layers between the model and the ad server
Each control wraps the one inside it. A version clears every layer before it is packaged, and clearing the right itself sits outside all six.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Decide whether the version may exist
- L3
- Decides what a person must approve
- L4 – L5
- Keep the disclosure and the trail intact
- L6
- Pulls automation back when signals degrade
How Nestack evaluates it
Evaluate every version — not only the master it came from.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the versions a person approves
- E1Final-output evaluationDid the version meet the platform's spec and the brand template?
- E2Step-level evaluationDid it read the current licence terms and the approved market strings?
- E3Tool evaluationDid it pull the current master and write to the right asset and campaign?
- E4Disclosure retentionDid the legal line and any required disclosure survive the crop?
- E5Slice evaluationHow does performance change across markets, formats and platforms?
- E6Business outcomeRejections, re-versions, and versions pulled after they ran.
Floor — the versions that can lawfully run where they run
Failure modes
Where each failure originates in the agent
Seven failure modes plotted against the five stages of the agent lifecycle.
Agent lifecycle
Direction of processing →
01 · Master & rights
2 modes
- CV-01
Version outside the licence
A market, medium or window the rights do not cover.
- CV-02
Talent in an excluded market
A performer runs where their contract bars it.
Stage gathers
The signed master, the template and the licences
02 · Composition
2 modes
- CV-03
Disclosure lost in the crop
The legal line falls outside the resized frame.
- CV-04
Template rules broken
Safe area, clear space or colour fails at a size.
Stage lays out
Layout, cutdown, safe areas and the platform spec check
03 · Localisation
1 mode
- CV-05
Offer localised wrong
Price, discount or terms do not match the market.
Stage swaps
Copy, offer, price and legal line for that market
04 · Output
1 mode
- CV-06
Product shown wrong
Generated art changes how the product looks.
Stage returns
The version the platform receives and the market sees
05 · Change / Version
1 mode
- CV-07
Stale master propagated
A superseded master versioned into a whole flight.
Stage tracks
Master, template, licence, spec and model changes
- Sev-1 · the version breaches a right or a rule
- Sev-2 · the version is wrong where it runs
- Sev-3 · the version is rejected or re-made
Affected slices
A clean master does not make a clean version set
Aggregate spec conformance can look acceptable while a few market and format cohorts carry most of the rejections, most of the rights holds and nearly all of the lost disclosures. Nestack reports performance by slice, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Right-to-left language markets | 5.3% | 3.2× | Review | |
| Vertical short-form cutdowns | 3.8% | 2.3× | Review | |
| Talent and music versions | 3.1% | 1.9× | Watch | |
| Repeat evergreen resizes | 1.3% | 0.8× | Normal |
Bar: failure-rate lift vs. evergreen-resize baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
The licence expires whether the campaign changes or not
Terms end, talent cycles renew and platform specs move under a live flight. A cycle that re-reads the version but never the date closes the wrong thing.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
A rejection, a rights hold or a lost disclosure moves in one market or format.
- 02
Diagnose
Traced to the licence record, the template, the string set or the master itself.
- 03
Improve
The term, template or string set is refreshed, re-approved by business affairs and version-linked.
- 04
Verify
Re-run across every size and market that master has already produced.
- 05
Learn
The breach becomes a regression case and a pre-flight check on the template.
Learn → Detect
The return edge. Every cycle also re-dates the licences and platform specs in play — a version that was clear last month may not be now.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, rights and template mapping, versioning and localisation, evaluation, trafficking, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Workflow discovery and automation-boundary definition.Week 1
- 02DAM, master and brand-template assessment.Week 1, Week 2
- 03Rights, talent and licence record mapping.Week 1, Week 2
- 04Version matrix — sizes, formats and markets.Week 2
- 05Template safe-area and brand rules.Week 2, Week 3
- 06Localisation strings and legal-line library.Week 3
- 07Platform spec and ad-policy sourcing.Week 3, Week 4
- 08Rights-window and exclusivity checks.Week 4
- 09Business-affairs and brand approval workflow.Week 4, Week 5
- 10Evaluation suite and regression cases.Week 4, Week 5
- 11Trafficking packaging and campaign integration.Week 5, Week 6
- 12Observability, deployment and Agent Care handover.Week 6
12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel
Final scope and sequence confirmed in discovery
Engagement tiers
What each tier includes
Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.
| Capability✓ in scope · — not at this tier | PilotOne master, one platform | ProductionProduction DAM integration | AdvancedMulti-market / multi-brand |
|---|---|---|---|
| Introduced at Pilot | |||
| Size and format versioning | ✓ | ✓ | ✓ |
| Brand-template and safe-area checks | ✓ | ✓ | ✓ |
| Platform spec conformance | ✓ | ✓ | ✓ |
| Baseline evaluation | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Rights and licence checks | — | ✓ | ✓ |
| Localisation, legal lines and category rules | — | ✓ | ✓ |
| Business-affairs approval workflow | — | ✓ | ✓ |
| Trafficking packaging and observability | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Rights-window monitoring across flights | — | — | ✓ |
| Native-reviewer localisation QA | — | — | ✓ |
| Multi-brand and enterprise controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on markets and formats in scope, DAM and creative-tooling integrations, rights sources, localisation depth, approval controls and deployment requirements.
- Separate from build
- Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.
What we need from you
What you bring, and what we build with it
Each input maps to a piece of build scope and a week in the delivery timeline.
You bring
We build with it
01A signed master and the brand template behind it
→DAM, master and brand-template assessment
Week 102Your rights records — music, stock, talent and likeness
→Rights, talent and licence record mapping
Week 103The version matrix you actually traffic
→Version matrix — sizes, formats and markets
Week 204Approved market strings, legal lines and disclosures
→Localisation strings and the legal-line library
Week 305Versions that were rejected, pulled or re-made
→Evaluation suite, regression cases and failure-mode testing
Week 406Named business-affairs and brand approvers
→Business-affairs and brand approval workflow
Week 407Access to the DAM, creative tooling and trafficking systems
→Trafficking packaging and campaign integration
Weeks 5–6
Nothing else is required
Deployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
Phases are drawn over the weeks they actually occupy. Week 5 carries both the native-reviewer pass and the first trafficked versions.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Masters, markets and formats in scope; who clears a right
W2Rights and template mapping, and the version matrix
W3Composition, localisation strings and disclosure rules
W4Approval workflow, evaluation suite and market slices
W5Trafficking packaging, native review and corrections
W6Versions trafficked under approval, then Agent Care handover
- Reading the band
- Rights and template mapping finish in week 2, before a single version is composed in week 3. The bars show that dependency, not a smooth ramp.
- At the end of W6
- Versions have gone out under approval and a named person cleared every right they rest on, then Agent Care takes over monitoring.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Advertising AI agent
Build a versioning agent around your masters and your rights.
Show us one master, the version matrix it has to fill and where your rights records live. We'll take that master across its markets and formats, and show you what the agent holds rather than what it produces.
- Typical delivery
- 4–6 weeks
- Deployment
- Your DAM and creative stack
- Rights sign-off
- A named person, always
- After launch
- Agent Care available
Nestack Agents · Creative versioning & asset production
AGT-AM-06 · Agent Care available after launch