What to ask an AI marketing supplier

A supplier offering to produce twice as much marketing for the same fee has made a proposal about production. The buyer still has to work out whether twice as much is useful, who will review it and what will happen to the work that already fills the week.

That distinction deserves attention as June closes. Gartner’s CMO spending analysis, published on 25 June, puts the average marketing budget at 7.8% of company revenue and describes continuing pressure on resources. An average is no prescription for an individual business. It does explain why a promise of cheaper or faster work will get a hearing during a budget review. Gartner’s 2026 CMO spending analysis

The buyer’s job is to turn that promise into something observable. The supplier’s preferred label, including “AI-native”, tells us too little about how a piece of work will be completed or what the client will receive. A useful discussion starts with a real assignment and follows it all the way through.

Put an ordinary brief in the room

A polished demonstration usually begins with the materials its presenter wants to use. That is reasonable for showing a feature. It leaves a procurement team with an incomplete view of how the service will cope with its own business.

Suppose a retailer needs a seasonal campaign adapted for several markets. Some prices are provisional. One product has different availability by country. The approved claim has a qualification that must remain visible, and a local team has a sound reason to reject the global headline. This is a hypothetical brief, but the tensions are concrete enough to test.

The supplier can show how it identifies missing information, develops an idea and handles those constraints. The buyer gets to see who intervenes when an output looks plausible but uses the wrong offer. An impressive first draft is still valuable. The route from that draft to an approved, usable asset is where much of the service becomes visible.

I would pay for a small, properly scoped trial here. Asking a prospective partner to build half a campaign for free rewards the firms with the largest pitch budgets and creates its own distorted comparison. A paid trial can use an agreed brief, clear acceptance criteria and enough awkward detail to make the result informative.

The incumbent process belongs in the comparison too. Without an account of its elapsed time, internal effort and revision burden, the buyer may simply be choosing the more persuasive demonstration.

Follow the work after the demonstration

The range of tasks on offer is expanding. On 10 June, Optimizely announced an AEO platform combining its observed AI-agent traffic data, Conductor’s search intelligence and agents intended to help teams act on discovery findings. That connects several stages of work in one proposition. The announcement describes capabilities; it does not establish the return a particular buyer will obtain. Optimizely’s platform announcement

For any supplier making a comparable promise, the revealing moment is the move from a recommendation to an action. A report may identify a content gap. Somebody must decide whether filling it serves a customer, whether the proposed claim is supportable and which existing page will need maintaining afterwards.

The trial should follow that decision. A report that a client team cannot use without another substantial round of analysis has a different value from an approved change ready for implementation. Both may be worthwhile services, provided the proposal says which one is being bought.

There will also be dependencies the supplier cannot remove. A product team may need to confirm a specification; an internal owner may have to approve an offer. Good partners make those dependencies visible early and explain how they affect delivery. A promise to take everything off the client’s hands deserves particularly close inspection when the supplier has yet to see the client’s data.

Count the effort on both sides

Imagine two trial proposals for the same assignment. One costs £6,000 and requires twenty hours of client review. The other costs £8,000 and requires four. Those figures are illustrative, not market rates. The cheaper proposal saves £2,000 in fees while consuming sixteen more hours inside the business. Whether that is a good exchange depends on whose time is needed and what it displaces.

If the reviewer is the only person who can approve the next product launch, those hours have a consequence beyond the hourly cost used in a spreadsheet. If the additional review develops a junior team’s judgement, some of the effort may be valuable. The buyer needs to see it rather than discover it after signing.

Software charges belong in the same conversation. A fixed service fee can sit beside variable usage costs, extra licences or work required to prepare the source material. The trial should expose the likely operating bill and the circumstances in which it changes. It should also show what the buyer can continue using if the relationship ends: approved assets, source material, records of decisions and access to its own work.

None of this requires a supplier to reveal every detail of its intellectual property. It requires enough visibility for the customer to understand the offer and take responsibility for buying it.

Agree what would earn the next assignment

A small trial cannot prove every promised commercial outcome. A campaign developed in two weeks may need considerably longer to show its effect, and a supplier should be able to say so without losing the pitch. Faster completion and fewer corrections can be observed sooner. Customer response needs its own evidence.

The contract can reflect that difference. The first assignment might establish whether the team can deliver acceptable work within the agreed constraints. The next can test that work in a defined market or channel, using a comparison that can reveal disappointment as well as success. The scope expands when the evidence gives the buyer a reason to expand it.

June’s budget conversation makes this an immediate buying decision. Before committing the next part of the year’s spend, a marketing leader can replace an abstract debate about the future agency with a much more useful discussion of an actual job. At the end of the trial, the people who will brief, approve and use the work should have seen the whole process, including the correction that the demonstration would normally leave out.