Agent-led growth: test how assistants describe your business

An assistant can help a buyer research your company before a sales call. Whether that happens in your market is something to investigate. A missing price does not prove that ChatGPT chose a competitor, and publishing one does not guarantee a recommendation.

Here, agent-led growth means making your business easier for a buyer’s assistant to research, compare and approach. Neal Behrend at Insight Partners describes this buyer-side use of agents. Treat the idea as a hypothesis to test against your customers’ behaviour.

For a service business, the first useful task is often an accurate offer page and a clear way to contact you. A new checkout or a callable tool is a separate decision.

Publish the facts needed to judge fit

Choose one service and answer the questions that determine whether an enquiry is worth both parties’ time.

Keep the answers consistent across the service page, pricing page and booking page. If a price depends on discovery, explain the variables and the next step instead of inventing a fixed amount. Public case studies should distinguish the client’s result from a forecast for another business.

Our free 30-minute growth consultation, two-week paid AI Opportunity Audit and scoped pilot build are distinct engagements. The pilot covers an agreed workflow with acceptance and stop criteria. Clear names help people and assistants compare the work without confusing a consultation with a paid project.

Run two different tests

Discovery: use a genuine buyer question without your brand name or URL. For example, a hypothetical buyer might ask, “Which firms could assess an invoice-processing workflow for a Canadian services company, and what information should I compare?” Keep the question relevant to the customers you serve.

Readback: give the assistant the actual offer URL and ask it to state the deliverables, price and currency, exclusions and next step, citing the supporting pages. Ask it to mark missing information as unknown. This checks whether it represents your offer accurately; it does not show whether a buyer would discover you.

Record the exact prompt, date, assistant and available search mode, language, location context, full answer and cited URLs. Use fresh conversations to reduce carry-over. Try a small set of realistic questions and repeat observations, keeping the conditions visible. Do not turn a handful of repeated prompts into a market-wide ranking or a buyer-adoption statistic.

Compare the answer with the cited page. If an assistant confuses our free consultation with a paid engagement, check the source URL and the wording on that page.

After fixing a specific problem, repeat the same test and retain both outputs. A corrected readback is useful evidence of that correction. It is not evidence of increased sales.

Make the first improvement small and verifiable

Keep a short issue record: the buyer question, incorrect or missing answer, supporting URL, proposed change, owner and review date. Prioritise errors that change fit, such as the wrong currency, a missing exclusion or a broken booking route. You do not need a separate page for every possible wording of the same question.

Google’s guidance says Google Search ignores llms.txt for visibility and rankings. That statement is about Google Search. An index file may serve another documented use, but it is not evidence that a buyer found or chose you.

If accurate information and a working form meet the buyer’s need, that can be the complete improvement. Only consider agent-callable jobs when a useful task requires them, with defined access and approval. The commerce guide addresses the separate question of purchasing.

Measure the enquiry, then the outcome

Track observed mentions, accurate descriptions, visits, qualified enquiries and closed work separately. Where available, retain referral information and the buyer’s own account of how they found you. Neither captures every interaction, so keep unknown sources unknown. An assistant visit or a crawler request is not a lead.

A sensible review asks whether the enquiries fit your service, whether important facts are represented correctly, and whether enough evidence exists to justify further work. A small sample or a long sales cycle may leave the commercial result unresolved.

Our generative engine optimization service covers this discovery work. Start with the free growth consult if you want to discuss a specific finding. Bring the question, answer and source URL so the conversation begins with an observable problem.