AI agents: how a team of agents builds your training kit
· 3 min read
You hear a lot about AI agents these days. The idea is not magic; it is a smart way to organise work. This article explains it simply and shows how Conti uses it to build your training kit.
What is an AI agent?
In computer science, an intelligent agent is a system that takes in information, makes decisions and acts to reach a goal. When several agents work together, each on its own task, we call it a multi-agent system.
Picture the kit team at a professional training house: an instructional designer who sets the structure, a content writer, an activity designer, an assessment writer and a references checker. Each one is a specialist and picks up the work of the one before. That is exactly what a multi-agent system imitates.
Why not let one model write everything?
Because long, compound tasks get much better when they are split. Anthropic's article on building effective agents describes common patterns such as prompt chaining and orchestrator-workers, where a big task becomes smaller steps that are each easy to check.
The golden rule: one small, clear task per agent, and a review point between the important stages.
Conti's stages: a production line of agents
A kit in Conti goes through fixed stages, each with a clear role:
Stage | What happens |
|---|---|
Technical conception | The kit's structure: days, sessions, axes and topics |
Introduction | The programme, why it matters and who it is for |
Objectives | General and detailed, measurable objectives |
Content | The material for every axis |
Activities | Exercises and interactive activities tied to the content |
Tests | Pre- and post-tests and evaluation forms |
References | A documented reference list |
Export | Word, PDF and PowerPoint files |
The key point in this line is the stop after the technical conception (for kits and books): Conti shows the structure as an editable tree and does not start writing until you approve it. So the content is never built on a structure that does not fit you.
What this means for you as a trainer
Higher quality and consistency: each agent focuses on one task and receives the output of earlier stages.
Transparency: you watch the stages complete one by one, with the time and credits used.
Flexibility: if production stops for any reason, you resume from where it stopped instead of starting over, because each stage's state is saved.
Fair cost: credits are counted per completed stage.
Limits you should know
Agents are powerful tools, but they do not replace your expertise. You know your audience and its context, so your review of the structure and the final files stays an essential part of the process.
The bottom line
AI agents are not one "robot" writing everything. They are an organised team with clear stages and review points. That is what lets Conti hand you a complete kit, not just text. Start from the training kit page.
