Skip to main content
The quality of your AI agent depends directly on the quality, completeness, and structure of the business information you provide. A well-configured agent is not a matter of settings alone — it is a reflection of how clearly your business knowledge has been prepared and organized. This guide helps you understand what information to prepare at each stage of your agent’s development, so it can handle real sales conversations effectively.

​
How to use this guide

AI-agent preparation is organized into four levels. Each level builds on the previous one — start with Level 1 and move forward as your agent matures.
You do not need to complete all four levels before launching. A Level 1 agent is already functional. Higher levels add capability, precision, and control over time.
LevelNameWhat it enables
1BasicAnswers common questions, handles basic conversations
2Sales-ReadyQualifies leads, follows your sales logic, handles objections
3OperationalOperates within your full business process end to end
4Optimization & ControlImproves over time, maintains consistency and quality

​
Minimum viable information set for launch

If a client cannot complete all four levels immediately, they can still launch with a minimum viable information set. This allows the agent to start working while the remaining levels are completed over time. The minimum set includes:
  • primary AI-agent goal;
  • short business description;
  • product or service list;
  • product availability source or inventory update rules, if relevant;
  • target customer types;
  • tone of voice rules;
  • FAQ;
  • pricing or pricing-response rules;
  • working hours;
  • communication channels;
  • sales, booking, or support logic;
  • handoff rules;
  • prohibited topics;
  • 5–10 examples of good conversations.
This minimum set is enough for an initial launch, but it should be improved after reviewing real conversations. If the business has a website, add the URL so the AI agent can send it to customers when they ask for more information, want to review services, pricing, portfolio, or contact details. The more complete the information, the more reliably the agent will perform.
If customers may ask whether a product is currently available, define where the AI agent should get this information from. For static or rarely changing availability, provide an updated product list or availability rules. For real-time availability, connect the agent to the business’s inventory, stock management, ERP, POS, or e-commerce system so it can use current data instead of outdated manual information.

​
Main rule: the agent cannot work well with information it does not have

This is the core principle behind all four levels. The AI agent cannot reliably:
  • follow rules that were never defined;
  • answer questions using information that was never provided;
  • represent a process that the business itself has not documented.
To make the AI agent work like the best team member, the business should provide:
  • what the agent should know;
  • what goal it should achieve;
  • how it should speak;
  • what questions it should ask;
  • what actions it can perform;
  • when it should hand off to a human;
  • what it must never say;
  • how a good conversation looks;
  • how success will be measured;
  • how the agent will be improved over time.
Why it mattersEvery gap in the agent’s knowledge is a gap in its behavior. The business controls what the agent knows. The more clearly that knowledge is defined, the more the agent can be trusted to represent the business.