Dishoom hired an agency to be found by AI. That is the line item existing.

Campaign reported on 28 August that Dishoom has appointed Brainlabs to improve how the restaurant group surfaces in AI discovery.

Read that as a budget line rather than a news item. Somebody at Dishoom worked out that a growing share of "where should we eat" now gets answered by a model rather than a search results page, decided that was worth paying an agency to influence, and signed something.

The capability has been discussed in this industry for eighteen months. The first UK client appointment most people will have noticed went to a restaurant group.

Why hospitality moved before B2B

The query shape suits it. "Where should I eat near King's Cross" is exactly the kind of question an assistant answers with a short list rather than ten blue links, and the list is genuinely short. Three names is a complete answer. Being the fourth is being nowhere.

Compare that to most B2B categories, where the buyer asks four or five increasingly specific questions over weeks and reads a shortlist of suppliers anyway. The cost of being absent from an AI answer is real, but it is diffuse and slow, and diffuse slow costs do not get budgets approved.

Restaurants moved first because the loss from being absent is immediate, countable and shows up in covers.

Categories where the loss is legible move first. That is a useful thing to know if you are trying to work out which of your clients will buy this and which will nod politely for another year.

The buyer now exists, which changes the sale

Until this month, most agency conversations about AI search were educational. You explained what an AI Overview was, cited the statistic about zero-click searches, and watched the client decide it was interesting but not this year's problem. Campaign's own reporting on 18 August put it at 80% of Google searches with AI-generated summaries ending without a click, which is the kind of number that produces agreement without producing a purchase order.

An appointment changes the conversation from education to comparison. The client's next question is not "why would we do this" but "who else does it, and are they better at it than you".

That is a much harder question, and most agencies are not ready for it. Explaining a problem well is not the same as being credible at fixing it, and the market has just moved from the first to the second.

What this exposes

Agencies have been publishing about generative search since it became a topic. Very few can name a client, a scope, or a thing they measured. That gap was invisible while nobody was buying. It stops being invisible the moment there is a comparable appointment in the trade press.

Three positions are now available, and only one of them is comfortable.

You can be the agency with the named client and the measured outcome. You can be honest that this is adjacent to what you do and you partner for it. Or you can carry on publishing about it while a specialist with one reference client takes the brief.

The third position is the most common and the weakest, because it looks identical to the first from the inside and nothing like it from the client's side of the table.

The narrower point for independents

None of this requires a large practice. It requires being able to describe, in specific terms, what you would do in the first ninety days and what you would refuse to promise.

Clients buying a new capability are not looking for the most confident supplier. They are looking for the one whose account of the work sounds like it came from having done it, which is a quality that survives being repeated to a finance director and does not survive being lifted from a whitepaper.

The Dishoom appointment will not be the last one this quarter. The agencies who get the next few will mostly be the ones who were specific in public before there was a brief to answer.

If you want to work out what your own account of this sounds like from the buyer's side, that is a conversation worth having.