Product Updates
AI TAM Query Assistant for ICP Search
DiscoLike’s AI TAM Query Assistant turns a short ICP description into a proposed search configuration using patterns from prior customer searches.
· updated October 6, 2026
A founder may know exactly what a good customer looks like and still struggle to write a useful account search. The AI TAM Query Assistant helps translate that business knowledge into search text, phrase matches, and filters.
Give it a short description of your ideal customer. Review the proposed search before using it to build a larger target account list.
How prior searches inform the assistant
DiscoLike’s starting material was thousands of saved and exported searches from hundreds of GTM teams. These searches paired ICP descriptions with example domains, phrase matches, and industry choices.
We used Claude Code to analyze patterns in that work. The aim was to capture practical search knowledge: how operators describe a niche, which examples clarify it, and which constraints narrow the results.
A saved or exported search indicates that an operator found it useful. It does not prove that every account qualified or later became a customer.
What to include in your request
Name the business activity and the customer it serves. “Software companies” is a starting point; “software vendors selling appointment scheduling to independent clinics” gives the assistant a clearer distinction.
Add geography, size requirements, and representative domains where you have them. Explain any must-have criteria separately so you can check how the assistant applies them.
Review, sample, and refine
Read the proposed ICP text before running a large search. Look for assumptions you did not intend, then inspect a sample of companies.
If the results include the wrong business model, revise the description. If a filter removes known matches, examine whether that filter is essential or whether it relies on incomplete evidence.
The DiscoLike MCP: Refine Your ICP Query Plan article explains how an agent can inspect and revise the same query plan.
Try DiscoLike with one customer segment and a few examples you can independently verify.
Related posts:
