Your China-based production partner proposes an annual AI creative service. It will remember approved product facts, learn your review preferences and suggest what to make next. The proposal includes setup and ongoing fees. Before negotiating the total, ask what “exclusive to your brand” actually describes.
It might describe your product knowledge, a custom review configuration or a newly developed module. It might also be used loosely for access to software the agency operates for many clients. Those are different purchases. ONCE recommends identifying the asset first, agreeing its intended use second, and attaching a fee to the actual work and restrictions.
Why this question deserves a place in the brief
On November 2, 2026, Publicis announced that it had completed its LiveRamp acquisition. Its official announcement positions data and AI capabilities inside clients’ own environments, with client control of data, AI intellectual property and technology choices. These are company commitments, not independent proof that every implementation delivers them or that agency practice has universally changed.
Smaller brands need not replicate a holding company’s infrastructure to learn from that positioning. They can ask a practical question: as the service becomes more knowledgeable about this brand, what exactly accumulates for the brand, and what remains the provider’s reusable method?
Build two inventories before comparing prices
The ONCE approach starts with one inventory of project-specific material and another of methods the provider proposes to reuse. Mixed components need their own explanation. These are commercial scoping categories, not declarations of legal ownership.
| Component | What the proposal should clarify |
|---|---|
| Your footage, product documentation and review history | The intended project uses, with any proposed training or cross-project use stated separately rather than hidden under “improvement.” |
| Brand vocabulary, review rules, shot annotations and project findings | The files or configurations delivered, the agreed exclusive uses, and the conditions attached to findings. |
| Existing agency checks, blank templates and operating methods | What remains reusable and whether you are buying a service, an operational copy or another specified deliverable. |
| A general tool combined with brand-specific configuration | Which parts can be separated and how the combined component can be delivered and used if separation is impractical. |
| Third-party models, plugins and licensed material | The dependencies, who pays for them and what the provider can actually deliver under the applicable arrangements. |
Derived material deserves particular attention. A recommendation configuration can preserve knowledge from a project without retaining the original spreadsheet. Removing a brand name from a prompt or a creative ranking does not, by itself, establish that it is suitable for reuse.
Reed Smith’s September 30, 2026 analysis of AI in media planning and buying separately raises output ownership, use of brand data and training purposes. That distinction informs the questions here; the article is not being treated as a universal legal rule.
Use synthetic substitution to expose a mixed component
Consider a hypothetical overseas appliance brand working with a Shenzhen production team. The proposed service contains a blank script checker that flags missing model numbers. It also contains a configuration recommending compact-storage stories for that brand’s UK campaign. This is an illustration, not an ONCE engagement or test result.
In a separate demonstration environment, replace project material with a clearly fictional brand and synthetic product documents. Does the checker still identify a missing model number? Can the provider describe that general capability without using the first brand’s facts? Does the recommendation configuration still reproduce the first brand’s market assumptions?
This exercise helps make the components visible. It does not establish permission to reuse them, prove that sensitive information has been removed, or require deleting production data. A functioning demonstration and an agreed use arrangement answer different questions.
The resulting brief can be precise: deliver our approved product facts and brand-specific rules in an agreed form; identify the general checker; explain any inseparable configuration and its operating dependencies. That is more actionable than asking for exclusive ownership of “everything AI.”
Price the work and the restriction separately
Use this comparison structure, with zero or omitted items where appropriate:
Total fee = integration and preparation + dedicated development + ongoing operation and maintenance + third-party charges + separately agreed exclusivity restrictions.
Preparation covers the material and connections to be organized. Development names what will be built and how it will be accepted. Operations specifies update frequency, human review, support and duration. Third-party charges identify subscriptions or usage costs and the responsible payer. Work already included in one line should not reappear under another name.
An exclusivity fee needs an identifiable additional restriction: for example, withholding a specific custom module from a defined set of other projects for an agreed period. Define the object, scope and duration before pricing it. Avoid charging again for protection already promised in the service, or presenting limited exclusivity as exclusive control over the supplier’s entire category expertise.
A brand with stable product facts and occasional review may need a well-maintained project folder and human support rather than an annual system. Frequent changes and repeated operational use can justify a different service. Equally, receiving a configuration does not make future hosting, updates or maintenance free.
A scalable production business can retain methods without pooling client findings
For ONCE’s longer-term move from project production toward managed content services, the reusable layer could include blank checklists, quality checks, collaboration methods and tools whose reuse has been agreed. Brand-specific facts, rules and findings require their own arrangements.
That distinction also matters when assessing business value. A large collection of client files is not automatically a commercially usable data asset. A repeatable method with clear provenance, permitted uses, deliverables and maintenance costs is easier to evaluate. Any improvement in margin, retention or valuation would still need operating evidence; none is claimed here.
For an overseas team sourcing production in China, start with one mixed component rather than a demand to acquire the agency’s entire stack. Have the partner identify the brand-specific material and the reusable method, demonstrate the separation with synthetic inputs, and map the result to the quote. That conversation makes “exclusive” a defined purchase before it becomes an annual commitment.
Explore ONCE’s production planning articles when translating that scope into shoot, edit and review work.
Sources and scope
Primary signal: Publicis Groupe, acquisition completion announcement, November 2, 2026. Supporting professional analysis: Reed Smith, “AI in media planning & buying: Current trends & risk considerations,” September 30, 2026. Both checked November 4. The inventories, synthetic substitution exercise, pricing structure and appliance example are ONCE’s proposed methods. No client systems, private data, model tests or realized financial results were examined.
