Should our agency offer AI discovery as a service?

Only if you can name who owns it internally, say what you would measure in the first ninety days, and tell a client plainly which parts you cannot influence. If you cannot do all three, you are selling a report rather than a service, and the client will work that out by month two.

The question is live because the appointments have started. Campaign reported on 28 August that Dishoom has hired Brainlabs specifically to improve its visibility in AI discovery. Once a category has a named client and a named agency, every other agency in the market gets asked the question by someone.

Most will answer yes on instinct. A good proportion should answer no.

What clients are actually buying

Not visibility. Visibility is the outcome they want and the thing you cannot promise, because the mechanism sits inside somebody else's model and changes without notice.

What they are buying is a considered account of how they are currently represented, a set of changes that make them easier for a model to describe accurately, and somebody who will keep watching as the answers move. The honest version of the offer is closer to reputation management than to search marketing, which is uncomfortable for agencies who priced it as the latter.

The clients who buy it twice are the ones who understood that distinction before they signed.

The four questions that decide whether you have an offer

Answer these before you put it on the website.

1. Who owns it on a Tuesday? Not who leads the practice in the deck. Who does the work when a client emails asking why a competitor is being recommended and they are not. If the answer is a rota of whoever is free, the offer will not survive its first difficult month.
2. What do you measure, and from when? You need a baseline captured before you touch anything, and a small number of things you will report on consistently. Agencies that skip the baseline spend the entire engagement arguing about whether anything improved.
3. What do you refuse to promise? Rankings in a system you do not control, and timelines tied to model release cycles. Saying this out loud in the pitch costs you a small number of briefs and saves you the ones that would have ended badly.
4. What happens when the model changes? It will, and the answers will move. If your commercial model assumes a fixed piece of work with a completion date, the first significant change will feel like failure to the client and unpaid work to you.

None of those are hard questions. They are just the ones that get skipped when a capability is launched off the back of market noise rather than a decision.

When the answer is no

You have no reference client and no appetite to be the first. Being genuinely first is expensive and often unprofitable. Doing it deliberately, at a reduced fee, in exchange for a case study you are allowed to name, is a strategy. Doing it accidentally on a full-fee retainer is how a capability gets abandoned after two clients.

Your existing clients are not in categories where the loss is legible. The businesses buying this now are ones where absence from an answer costs something countable and soon. If your client list is mostly long-consideration B2B, the appetite is a year or two out and you will spend the interim educating people who then buy from somebody else.

You would be reselling a tool with a layer of commentary. There is a version of this offer that is a dashboard subscription plus a monthly call. Clients recognise it faster than agencies expect, and it prices down towards the tool.

Nobody senior wants to own it. Capabilities that exist because the market seems to demand them, rather than because somebody in the building finds them interesting, decay quietly. This is the most reliable predictor of the four.

When the answer is yes

The strongest position is usually not a standalone practice. It is an extension of something you already do credibly, sold to clients who already trust you on the adjacent thing.

If you run communications for a client, how they are described by a model is a reputation question and you are already the person they ask about reputation. If you run their content, the material a model reads is largely material you wrote. Both of those are defensible. A newly invented practice with no relationship to your existing work is not, and it competes on price with specialists.

The other advantage of building it inside an existing relationship is that the first engagement is not a pitch. It is an extension of a conversation, which means the awkward questions about what you can and cannot promise get asked by somebody who already believes you.

The thing that actually sells it

Specificity in public, before there is a brief.

An agency that has written clearly about how this works, what it can and cannot achieve, and what it costs to do properly, is in a different conversation to one whose website added a service line last quarter. That is the same mechanism described in the visibility premium, applied to a capability rather than a business, and it is why the agencies who get the next few appointments will mostly be the ones who were specific about this months ago.

If you are working out whether this belongs in your offer, we can talk it through.