Personalisation in an AI agent happens at two levels: how the agent speaks (set in the system prompt) and what it retrieves (determined by your knowledge base structure). Get both right and your agent feels less like a FAQ bot and more like a knowledgeable mentor who knows your learners.
Personalisation Starts in the System Prompt
The system prompt is the set of instructions you give your agent before it talks to anyone. It is where you tell the agent who it is speaking to, what tone to use, and how much assumed knowledge is appropriate. A beginner-focused educator running a “Getting Started with AI” programme should instruct the agent to avoid jargon, explain every term the first time it uses it, and use simple analogies. An advanced programme can set a different expectation entirely.
Think of the system prompt as the briefing you give a new teaching assistant before their first class. “These students are 45+ professionals who are new to AI but very experienced in their own fields. Treat their prior expertise with respect and connect new concepts back to what they already know.” That briefing changes every answer the assistant gives.
Structure Your Knowledge Base for Different Learner Stages
Personalisation also comes from how your content is organised. If your BetterDocs knowledge base has articles tagged by difficulty — beginner, intermediate, advanced — your agent can retrieve the most relevant article for the question being asked at that stage. A student in Week 1 asking “What is a prompt?” should get a different depth of answer than a student in Week 8 asking the same question.
You can also write separate articles for common “stuck points” at different levels. “Why is my prompt not working?” means something very different to a first-week learner versus someone building their first automation. Two articles, two audiences, better answers for both.
Where your platform allows it — FluentCommunity, for example — you can pass member data into the agent context so it knows which course the student is enrolled in, how far they have progressed, and what their stated goals are. That context lets the agent tailor its response without the student having to re-explain their situation every time.
What This Means for Educators
You do not need to build a different agent for every type of learner. One well-briefed agent with a well-organised knowledge base handles a wide range of learner needs. The investment is in writing good articles that speak to different experience levels — work that serves your human students just as much as it serves the agent.
The Simple Rule
Brief your agent on who it is talking to, structure your content by learner stage, and pass in as much context as your platform allows. That combination gets you eighty percent of the personalisation benefit with none of the complexity of building separate agents for separate audiences.