AI-Agent Client Discovery Questionnaire: Five Questions That Find the Real Scope

A long questionnaire does not guarantee a useful brief. The goal is to ask a small number of compound questions that let you extract the agent’s audience, purpose, common intents, authoritative facts, prohibited topics, response behavior, and next step.

Ask for one rough answer at a time, summarize what you understood, and request clarification only where the answer changes the build. Do not make the client repeat information already supplied in a website, document, or earlier answer.

Ask: “Who should use this agent, what should it help them accomplish, and what five questions do those people ask most often?”

Extract the primary audience, secondary audiences, desired outcome, repeated intents, vocabulary, and a first set of launch tests. If the answer includes several unrelated jobs, choose the highest-value safe starting job rather than building a general assistant.

Ask: “In one rough message, tell me what you offer and the facts people need before taking the next step—such as services, products, prices, inclusions, process, timing, hours, locations, policies, proof, and current limitations.”

Turn the answer into explicit facts and FAQs. Record conditions, dates, locations, plans, exclusions, and source ownership. Mark anything uncertain for confirmation instead of smoothing it into confident marketing copy.

Ask: “What must the agent never answer, promise, reveal, or decide; what should it say when information is missing; and exactly where should it send someone who needs a human?”

Capture sensitive information, regulated advice, private records, guarantees, negotiations, emergency matters, unsupported comparisons, and any action requiring authorization. Make the handoff actionable with a person, team, form, email, phone number, or approved link.

Ask: “How should it sound, how detailed should answers be, what should it call the business and visitor, and what action should it normally recommend next?”

Convert vague answers such as “professional but friendly” into observable guidance: preferred terminology, response length, use of bullets, questions it may ask, phrases to avoid, and the single primary next step.

Ask: “Which websites, documents, FAQs, notes, and contact links may the agent use; who approves them; and should the final agent use a Qlynk page, a website widget, or both?”

Record source permission, authority, owner, effective date, review trigger, privacy level, and publishing destination. Confirm who will create the client-owned account and who can approve publishing and website installation.

• Agent briefName, type, audience, job, welcome message, suggested questions, and primary next step.
• Knowledge planApproved facts, FAQs, links, documents, gaps, conflicts, and source owners.
• Behavior rulesAllowed topics, blocked topics, uncertainty response, tone, response length, and escalation.
• Launch planAccount ownership, publishing surface, website installation, acceptance tests, approval, and maintenance.