Build a B2B ideal customer profile with AI and real evidence

Distinguish the account from the person

An ideal customer profile describes a suitable organization. A buyer persona describes a person within that organization. Mixing the two can produce a profile that sounds specific but offers little guidance: a busy marketing manager who wants growth might work in thousands of companies with very different needs.

Start with the account's situation. What recurring problem exists, what makes it urgent and what resources are available to solve it? Then identify the people who experience the problem, evaluate a solution and authorize spending. Your campaign may reach a user while the decision depends on someone else.

Read successful and unsuccessful accounts together

Prepare anonymized summaries of customers who achieved a useful outcome, deals that stalled and buyers who chose another approach. Include onboarding effort, sales-cycle context and support requirements where you have permission to use them. Revenue alone can hide an account that was expensive to serve or never reached the intended outcome.

Look for recurring conditions rather than attractive labels. An illustrative workflow product might fit agencies managing several simultaneous client approvals better than agencies of a particular headcount. That is a hypothesis to investigate, not a universal rule about the agency market. Include cases that do not fit the pattern.

• Problem: which task repeatedly causes delay or rework?
• Trigger: what changed that makes action necessary now?
• Ownership: who is accountable for fixing the problem?
• Readiness: what process or data must already exist?
• Disqualifier: what requirement cannot your product satisfy?

Write two competing ICP hypotheses

Create two plausible segments rather than one polished description. For each, state the buying situation, current workaround, expected benefit and obstacle to adoption. A segment defined by urgent client deadlines may behave differently from one defined by internal reporting complexity, even when the companies look similar in a database.

Add explicit reasons to reject each hypothesis. For example, a customer may require an integration you do not offer or implementation help you cannot provide. Disqualifiers protect both sides from a poor fit. They also make your marketing more credible by preventing every visitor from being treated as an ideal prospect.

Use Mirror to challenge the purchase logic

Upload the two segment briefs and the evidence behind them. Explore each situation separately in Mirror, keeping the product capabilities consistent. Ask how the user, manager, finance approver and existing supplier might influence the decision. Review whether each represented role is relevant to the actual account.

Pay particular attention to contradictions. A team can urgently need a solution and still lack authority to buy it. An executive can approve a budget while users resist the workflow change. Simulated objections are prompts for investigation, not proof that a segment will convert.

Turn the profile into an account-screening checklist

Translate the strongest hypothesis into questions a salesperson can answer from a discovery call or a permitted account review. Replace vague criteria such as innovative company with observable conditions such as a team currently coordinating approvals across multiple client projects.

Run a small research batch before expanding acquisition spending. Record the fit criteria, the source of each answer and the outcome of the conversation. Compare not only positive replies but also onboarding feasibility and the customer's ability to reach a useful result. Keep unknown answers visible instead of converting them into a favorable score.

Use the ICP to sharpen your next campaign

A working profile should change something concrete: which prospects receive outreach, which examples appear on a landing page or which discovery question is asked first. If it cannot change a decision, it is probably still a descriptive document rather than an operating tool.

Bring one candidate segment and a short evidence packet into Mirror to explore its buying barriers. Use the report to refine your checklist, then test it in real conversations. The outcome to seek is a clearer account-selection decision, not a larger list of AI-generated personas.

Common questions