Can an AI Agent Book You?
The next wave after recommendations: AI agents that enquire and book on a customer's behalf. What agent-friendly means in practice, and the honest timeline for caring.
So far this cluster has covered assistants that recommend — they name you, and the human does the rest. The next shift is already visible at the edges: assistants that act. "Find me a physio nearby and book something this week" is a request the newest AI agents attempt end-to-end: search, choose, open the website, fill the form, book the slot.
Today that flow succeeds rarely and haltingly. But the direction is unambiguous, and it introduces a genuinely new question about your website: not "does it persuade a human?" but "can a machine, acting for a human, complete a transaction on it?"
The agent test: an AI agent trying to hire you needs three things your site either provides or doesn't — a clear path (an obvious, working way to enquire or book), machine-navigable steps (real forms and buttons, not phone-only funnels or embedded puzzles), and a confirmable outcome (a definite "your request was received" a machine can recognise). Every gap ends the attempt — and the agent simply proceeds to the next name on its list.
What actually blocks an agent
The obstacles are mundane, and most of them quietly annoy humans too:
Phone-only conversion. If the only way in is a phone call, an agent (and the after-hours human) is done before starting. A working form or booking path is the minimum machine-accessible door.
Forms that fight. Multi-step widgets that break, CAPTCHAs guarding a simple enquiry, required fields nobody understands. Every one of these filters out agents today and impatient humans every day.
Ambiguous endings. The form submits into silence — no confirmation page, no acknowledgement. An agent can't tell its user "done"; it reports failure and moves on. (The humans, meanwhile, just wonder if you got it.)
Self-booking, or not. An agent that can complete "book Tuesday 2pm" against a real calendar is delivering exactly what its user asked. Practices with self-booking are, without any extra work, already agent-transactable; practices without it cap the agent at "I found their form."
Notice the pattern: everything that makes a site agent-friendly makes it human-friendly first. That's what makes this page actionable now rather than speculative — the to-do list is the same conversion hygiene Get Chosen already demands, and the agent wave just raises its stakes.
The honest timeline
Don't rebuild anything for agents this quarter — the traffic doesn't justify it yet, and the standards are still settling. Do fix everything on the agent test that's also a human-conversion fix, which is nearly all of it. And run one experiment worth five minutes: ask an agent-capable assistant to enquire with your practice, and watch where it gets stuck. Wherever the machine stalls at your front door tonight, some tired human stalled last week — the agent is just the first visitor who files a report.
Questions practices actually ask
Is anyone actually booking through agents today? In professional services, a trickle at the frontier — early enough that almost nobody is ready, late enough that "ready" is becoming a real differentiator. The moment to be early is while it's cheap.
Is it safe to let machines book my calendar? With the same guardrails good self-booking already has — confirmations, deposits where appropriate, qualification questions — an agent booking behaves like any online booking: it's the client's intent, machine-delivered.
Do I need special code or a standard for this? Emerging conventions exist and will matter eventually; today the practical answer is simpler: real forms, real confirmations, real booking. The standards, when they settle, will sit on top of exactly that.
How does this relate to being recommended? It's the same funnel, one step deeper: recommendation gets you named; transactability lets the naming become a booking without a human ever visiting. The practices that win both steps get customers who never saw a search result or a website.
Part of Get Found by AI — the deep dive on the second front door.