You tell AI: “Book a restaurant for four on Saturday evening, near the metro, under eight hundred per person.” It returns a booking confirmation. You haven’t even opened the restaurant’s homepage.
Convenient. For consumers, that is exactly the point of delegating. But for the restaurant working hard to shoot short videos and improve its search rankings, the entry point to the marketing battlefield has already shifted before the customer even walks in.
I tend to think the capability of AI agents that deserves the most scrutiny is deciding for us which differences we needn’t even look at.
This is more troubling than “everyone will just choose the cheapest option.” At least prices are clearly listed; omitted value never even gets a chance to make its case.
How Many Unspoken Conditions Hide Behind “Take Care of It for Me”?
Take the discussion around using AI to blanket the job market with applications as a mirror: applying faster does not mean either side understands the other better. Bring the same logic to reservations and price comparisons, and the problem follows.
“Find the best-value plumber or electrician.” Does value mean a low call-out fee, or someone willing to photograph the cause of a leak and explain which parts don’t need replacing? “Book a restaurant suitable for older people.” Does suitable mean having an elevator, or servers willing to wait for Grandpa to finish speaking?
All of these things can be turned into data, but someone first has to think to ask, someone has to take responsibility for recording them, and businesses’ own claims cannot be the only source.
If an agent has only prices, distances, and star ratings, comparing those is perfectly reasonable. The trouble arises when the interface presents “these are all I can compare” as though it means “this is what suits you best.”
Just like that, missing data becomes a business’s rejection.
Businesses May Have Acquired an Unpaid Purchasing Manager
Imagine two nearby auto repair shops. One quotes a lower price for maintenance; the other charges a little more, keeps inspection photos, and is willing to investigate a recurring strange noise.
If the agent receives only “find cheap maintenance,” there is nothing wrong with choosing the first. But when this kind of delegation happens repeatedly, the second shop may be forced to spend its energy breaking down quotes and adjusting line items to make the number look better at first glance.
There is a tension here: consumers really do need someone to screen out inflated quotes; professionals also need room to explain things they cannot confirm until they begin the work. If an agent treats every follow-up question as transaction friction, the few minutes it saves may be precisely where professional judgment takes place.
Whoever gets to define the comparison fields begins to shape how work is priced.
The Marketing Audience Now Includes the System That Eliminates Customers’ Options
For local businesses, this makes AI search and recommendations very concrete. “Attentive service” offers almost nothing to compare; clearly stating whether wheelchair users can be seated, what a quote includes, and how cancellations are charged gives agents usable material for judgment.
But clear information does not guarantee a recommendation. How recommendation slots are allocated, whether payment can influence rankings, and whose system an order passes through all affect exposure. A business might buy one fewer click only to pay an additional commission on a sale; it might also have no way of knowing why the customer chose it.
The digital marketing question I care about has therefore changed: beyond helping AI understand a service, we also need to know what it leaves out. Otherwise, even the most polished content may be reduced to a single price field.
On Saturday afternoon, the restaurant owner receives a reservation from an agent: “Four people, near the entrance.” He glances at the note and asks in reply: “Any older guests? It’s drafty by the entrance. They’d be more comfortable inside.”
A spinner turns on the screen. The agent is deciding whether this question is worth bothering the customer with.