Personal agents and what changes when your customer's agent already knows them.

Where the marketing budget stops working

Some customers' shopping is already decided before they ask anyone anything, because their personal AI agent already knows their taste, their usual spend, and what they've bought before. A small ceramics studio spends real, sensible effort making sure it shows up well across the surfaces a shopper might browse, a strong listing here, a good mention there, decent visibility on a comparison site, and every pound of that effort assumes the shopper is the one doing the looking. For this particular customer, none of it was ever going to be seen.

The studio's usual levers, a better product photo, a seasonal promotion, a boosted listing, are all designed to win attention from someone choosing where to look. A personal agent was never choosing where to look in the way those levers assume; it goes straight to checking specific shops' specific facts against what it already knows. Money spent sharpening any of those levers, for this customer specifically, buys nothing at all. The levers aren't broken. They're simply aimed at a decision this customer's agent never makes.

How the agent decides is a separate question, answered elsewhere

Exactly how a personal agent judges whether the studio's facts are a genuine match, and how that differs from a search engine ranking a page, is its own question, one this site answers in full elsewhere. What's specific to this piece is narrower: given that no amount of surface investment reaches this customer, where the studio's spend should go instead. The honest answer is a different kind of investment entirely, keeping the studio's own facts, exact glaze colours in stock, exact dimensions, exact lead time, accurate and current, which costs remarkably little next to a single seasonal ad push, and unlike that ad push, is the one thing that genuinely reaches this customer at all.

Two studios can be equally well regarded, equally well reviewed, equally visible across every surface a shopper might browse, and still get treated completely differently by this kind of agent, purely on whether their specific facts happen to answer the specific request in front of it. Reputation, in the sense a browsing shopper would recognise it, barely enters into it.

Redirecting the budget, not abandoning it

None of the usual surface-visibility spend stops earning its keep for shoppers who are still browsing; most customers still are. What changes is that a second, separate line item is now worth adding alongside it: not another surface to be seen on, but the ongoing cost of keeping specific facts current enough for an agent to match against a request that arrives already decided. Redirecting the whole budget there would be the wrong move. Ignoring that a second line item exists at all is the more common one.

Measuring whether the redirection paid off

The usual marketing numbers, click-through, time on listing, ad engagement, will show nothing for this customer, because none of those events happen on the way to this kind of sale. What moves instead is a quieter figure: repeat orders with no marketing spend attributable to them at all. That's not a hole in the reporting. It's the actual shape of a sale this kind of customer produces.

Keeping those specific facts checkable, so a personal agent finds a genuine match regardless of how it got there, is where Selfe's own work sits in this picture.

Should we stop investing in surface visibility, then?

No, most customers are still browsing and that spend still earns its keep for them. This is an additional line item, not a replacement.

Do we need different facts for this customer versus a regular shopper?

No, the same specific, current facts serve both. A personal agent just has no patience for anything that isn't already exact.