1. Start with the repeated question
Write down the questions you answer again and again. A narrow, familiar job—explaining services, guiding guests, answering product FAQs, or clarifying a process—is a stronger starting point than “answer anything.”
2. Choose the agent type and audience
Select the closest role: Personal AI, Business, Property, Operations, Product, Support, or Custom Guide. Then identify who will ask the questions and what they should be able to accomplish.
3. Define scope before adding content
- PurposeDescribe the job in one or two sentences.
- Allowed topicsList the subjects that belong in the conversation.
- Blocked topicsName anything the agent should not discuss.
- Human handoffExplain when and how the visitor should contact a person.
4. Add the answers
Add facts, FAQs, links, profile context, and supported documents that directly support the job. Review each source first. Do not upload passwords, access codes, private customer information, or material you are not allowed to use.
5. Configure identity and behavior
Choose a clear agent name, welcome message, tone, response length, colors, and suggested questions. Make it obvious that visitors are speaking with an AI representative, not the person or team itself.
6. Test the edges, not just the happy path
Ask expected questions, vague questions, out-of-scope questions, blocked questions, and questions whose answer is missing. Confirm that the agent stays grounded, admits uncertainty, and uses the intended handoff.
7. Publish and maintain it
Share the Qlynk link where people already look for answers. Review conversations, correct outdated knowledge, and use knowledge gaps to decide what to add next. A trustworthy agent is maintained like any other source of public information.