AI agent training · business and technical teams

AI agent training. Agents in your tools, and a human who signs off.

Kailli's AI agent training teaches your teams to design, deploy and oversee AI agents in the company's tools: choosing the tasks, plugging in connectors and MCP, deciding what a human approves, tracking costs. Two tracks, business or technical. 3 days (21 hours), in-house or in a multi-company session.

Pre-register See the program

No-commitment pre-registration: we check your needs and prerequisites, then send a written proposal before any final enrollment.

01Practical information

Practical information

The AI agent training — duration: 3 days (21 hours), in-house or multi-company, from €1,400 excl. VAT.

Duration
3 days (21 hours).
Format
in-house (your teams, your tools) or multi-company session.
Price
from €1,400 excl. VAT per participant.
Prerequisites
no code for the business track; Git and the terminal for the technical track. Access to the tools used, paid for by the client.
Qualiopi
DAZZ SOLUTIONS is a training provider Qualiopi-certified for training activities (certificate RNQ-23/10/1888-1, valid until January 18, 2027). This certification covers the provider's processes; it does not certify this course or the participant's skills.
Funding
full or partial funding can be explored with your OPCO (French skills-funding bodies), depending on your company, the available budget and the funder's prior approval.
Accessibility
a need related to a disability? Tell us from the first conversation: we'll look at possible adjustments with you.
Delivery and location
On-site anywhere in France or remotely in a virtual classroom; in-house or multi-company on request.
Capacity
3 participants minimum. The maximum number of participants is stated in the proposal before enrollment.
Lead time
Enrollment must be finalized at least 14 calendar days before the start, subject to available places and funding approval. Dates are agreed with the participant or the company; pre-registration does not reserve a place.
Follow-up and assessment
Before the training: a placement questionnaire and an interview to check expectations and prerequisites. During: observed hands-on exercises. At the end: a final practical exercise and a skills review. A separate satisfaction survey collects your feedback.
Learning resources
Summary PDF, practical guides and reusable templates, shareable exercises and answer keys, with no confidential data. A certificate of completion states the hours actually attended; the skills review is provided separately, with no diploma or vendor certification.
Disability contact
Morgan Godard: contact@kail.li. Adjustments are discussed during the interview, without entering any health data in the form.

Pre-register for this training

In brief

Who it's for
business and technical teams that already use AI and want to hand it tasks.
What you work on
your workflows, your tools and your data, not a demo.
Format
3 days (21 hours), in-house or multi-company, from €1,400 excl. VAT.
Quality
Qualiopi-certified for training activities (DAZZ SOLUTIONS); Qualiopi is France's national quality certification for training providers.

02Definition

AI agent: what does it mean for a business?

An AI agent is an AI model that receives a goal, chooses the steps itself and acts in your tools (files, email, CRM) until it gets the result. An assistant answers a request, an automation follows a script set in advance: the agent decides what comes next.

CriterionAssistant or chatbotAutomationAI agent
What it doesanswers a requestruns a fixed scriptcarries a task through to the result
Who decides the stepsyou, message by messagewhoever wrote the scriptthe agent, within set limits
In your toolsrarelyyes, depending on the scriptyes, through connectors and MCP
Main riska wrong answera script that breaksan unintended action

Agentic AI refers to this family of systems. An agent should only be deployed with three written answers: what it reads, what it does, who approves.

03Objectives and topics

The AI agent training program.

Five learning objectives, worked on a real workflow from your team. The balance depends on the track and the duration.

  1. Objective 1 · Choosing the tasks to hand to an agent

    • Spot tasks that are repetitive, multi-step and easy to check.
    • Rule out those where a mistake costs too much.
    • Write the task down: input, expected result, check criterion.

    Hands-on practice

    describe one of your team's tasks.

  2. Objective 2 · Starting from the tools already in place

    • Claude and Claude Cowork, ChatGPT's agents, Copilot Studio, n8n: what each one can do.
    • Choose based on where your data lives.
    • Subagents: breaking down a long task.

    Hands-on practice

    build a first agent on the team's tool.

  3. Objective 3 · Connecting the agent: connectors and MCP

    • The Model Context Protocol (MCP), an open standard for linking an agent to your tools.
    • Connectors: email, documents, CRM, tickets.
    • Read-only first; write access is decided action by action.

    Hands-on practice

    connect a tool, read-only.

  4. Objective 4 · Keeping a human in the loop

    • Place the approval steps: before sending, paying, deleting.
    • Keep a log of what the agent read and did.
    • Know how to stop an agent and take back control.

    Hands-on practice

    place the workflow's approval steps.

  5. Objective 5 · Security, data and costs

    • Permissions, secrets, personal data: what the agent never sees.
    • Track an agent's cost: calls, tokens, review time.
    • Team rules, a first basis for an AI usage policy.

    Hands-on practice

    write the agent's rules sheet.

04Two tracks

Two tracks: business or technical.

The objectives are the same; the tools and the level change depending on who is being trained.

  • Business track, no code

    Management, sales, HR, customer service. Agents in office tools, such as Claude Cowork for business: documents, email, CRM, with an approval before anything is sent.

  • Technical track

    Developers and data teams. Agents in the repository and the CI, MCP servers, subagents, permissions. To go further: Claude Code and MCP for technical teams.

A mixed team? We combine both tracks on the same workflow.

05In-house

Agents we use ourselves.

We train with the agents that run in-house at our company. These are not client cases.

  • Sales team: the agents prepare research and proposals; the salesperson checks and decides.
  • Leadership team: they gather documents and management data; leadership makes the calls.
  • Technical team: they work on code and tickets; the developers check and approve.

See the diagrams of these three systems

06The trainers

Your trainers work with agents.

Meet the trainers: two practitioners who oversee AI agents on real projects.

  • Rémi Carbonne

    Founder and CTO

    He codes and sells with AI every day, on projects that go into production.

  • Morgan Godard

    PhD in data science and trainer

    He knows why a model works, and he can explain it to a room with no technical background.

Frequently asked questions about the AI agent training.

What is an AI agent?

An AI agent is an AI model that receives a goal, chooses the steps and acts in your tools until it gets the result. It differs from a chatbot, which answers, and from an automation, which follows a fixed script.

Do you need to know how to code to take AI agent training?

Not for the business track: agents are configured in tools such as Claude Cowork or Copilot Studio. The technical track requires Git and the terminal.

Which tools do you use?

The ones your team already has: Claude and Cowork, ChatGPT's agents, Copilot Studio, n8n, or Claude Code for developers. Kailli sells no tools and takes no commission.

What is MCP (Model Context Protocol)?

The Model Context Protocol is an open standard launched by Anthropic in November 2024 that links an AI agent to your tools: documents, tickets, databases. Each MCP server opens a specific access, read or write.

Can agents act on their own in our tools?

Technically yes, if they have the access. In the training, the agent starts read-only, and every action that sends, pays or deletes goes through human approval.

Is the AI agent training eligible for CPF?

No. This training is not eligible for CPF (France's personal training account). Full or partial funding can be explored with your OPCO, depending on your company, the available budget and the funder's prior approval.

How much does the training cost, and how long does it last?

From €1,400 excl. VAT per participant, for 3 days (21 hours). Price confirmed by quote before enrollment.

Training pre-registration

AI agent training

Your profile

You fund the training yourself. After your request, we will contact you by email to schedule a phone call: be sure to check your spam folder. We will review your expectations, prerequisites and any adjustments needed, then you will receive a proposal specifying the program, duration, dates and price. No payment or commitment is required here. For a private individual, the applicable terms and contract must be approved before any final registration.

Your contact details are used to process this request and prepare the call. Retention: three years after the last exchange if no contract follows. Do not enter any health data. Information and rights: privacy policy

Before making any commitment, please read the documents: General terms and conditions of sale · Internal rules