Workshop

Frame and prototype an AI agent for your SME

Three hours in a small group to spot your tasks worth automating, map your processes and leave with a local AI agent prototype, without exposing your data.

Organise a workshop

workshop · three hours, 4 to 10 peopleIndicative

0h00Choose the taskthree candidates per participant, from their own work
0h30Map the processinputs, outputs, sensitive data, decision point
1h05Decide on hostingcloud, local, or hybrid, depending on data sensitivity
1h35Build the prototypea local no-code agent, assembled live on a fictional case
2h10Break itnormal case, ambiguous case, refused case — can the agent say no?
2h40Cost the effort and gohardware, oversight, maintenance, 90-day plan

Indicative schedule. The workshop follows your availability, and the time spent on each step follows the group's questions.

The deliverables

What you leave with

Not a slide catalogue: six documents each participant fills in on their own process during the session and reuses afterwards on their own.

Selection matrixgrid

Task	Freq.	Risk

Frequency, time spent, feasibility, sensitive data, risk in case of error.

Mapping templatetemplate

Inputs
Outputs
Data handled
Human decision

Confidentiality checklistsheet

Nature of the data handled
Retention period
Hosting location
Notice to the people concerned

Agent test gridprotocol

Normal casethe agent handles it
Ambiguous casethe agent asks
Refused casethe agent stops
Missing documentthe agent flags it

Local agent exampledemo

01reads a submitted document
02extracts the information from it
03proposes a reply
04stops before any action

On a fictional case, on the company's own hardware. No real data during the session.

90-day rollout planoutline

D0Task chosen, owner assigned
D+30Test on a bounded scope
D+90Criteria to extend, or stop

Selection matrix: A ready-to-use tool for assessing each candidate task: frequency, time spent, feasibility, sensitive data, risk if it goes wrong. You apply it to your own tasks during the session.
Mapping template: The template for describing a process: inputs, outputs, data handled, approval rules, point of human decision. Filled in once, it serves as the basis for the prototype and then the rollout.
Confidentiality checklist: The GDPR and AI Act points to check before handing a task to an agent: type of data, retention period, hosting, information to the people concerned. Adapted to the size of an SME.
Agent test grid: The minimum to test before deploying: normal case, ambiguous case, refused case, missing document. Each case is described with a concrete example and the expected behaviour.
Local agent example: A demonstration agent that can be deployed on a workstation, which reads a document, extracts the information and proposes a reply. A concrete starting point each participant can adapt.
90-day rollout plan: A template for organising what comes next: chosen task, hardware needed, person in charge of supervision, test milestones, criterion for moving to a wider rollout.

The programme

Six steps, from use case to rollout

A short, concrete format, with an exercise applied to your own work at each step. By the end, you know which task to automate, with which data and under which conditions.

Spotting a useful use case: A task becomes a good candidate when it is frequent, repetitive, time-consuming and limited in scope. Each participant lists three candidates taken from their own work, not from generic examples.
Mapping processes and data: A simple matrix per participant: inputs, outputs, sensitive data, approval rules and the point of human decision. The exercise shows what the agent can do alone and what must stay under control.
Choosing cloud, local or hybrid: The criteria are how sensitive the data is, the volume, in-house skills and the operating budget. For customer or personal data, local hosting stays the dominant criterion: the agent runs on the company's own hardware.
Building a limited prototype: A live demonstration of a local no-code agent on a fictional case: reading a document, extracting the useful information, proposing a reply, human approval before any action. The principle is visible without touching real data.
Testing failure cases and enforcing human approval: A minimal test grid: normal case, ambiguous case, refused case. We check that the agent knows how to say it does not know. Any consequential action, such as sending a message or changing a file, goes through human approval.
Working out costs and preparing rollout: Costs are expressed in effort: hardware, licence, supervision time, maintenance. Each participant leaves with a simple 90-day plan: task to test, person in charge, decision criterion.

Who it is for

A workshop designed for teams with no technical background

Executives, operational managers, independent professionals and project teams: the format adapts to your starting point, not the other way round.

SME owner who wants to understand AI: You want to know, concretely, what AI can do in your business, with no jargon and no promises. The workshop gives you criteria for spotting a genuine use case and avoiding projects with no return.
Operations manager looking for time savings: You know the repetitive tasks that slow your team down. The workshop helps measure their real cost in time and choose the one an agent can take on first.
Team with no technical background: No developer is needed. The exercises and demonstrations use no-code tools, and each participant leaves with materials they can reuse on their own.
Independent professional: Accountant, solicitor, notary, consulting firm: the confidentiality of files is the first constraint. The workshop shows how a local agent handles documents without sending them to an outside cloud.
Association or local authority: Limited budgets and sensitive data, that of members or residents, call for sober solutions. The workshop format lets you assess a precise use case before any commitment.
Group preparing an automation project: You are considering several agents or a wider automation effort. The workshop helps frame the first cases, identify the data involved and prepare an informed decision.

The format

How the half-day runs

Four to ten participants, no technical background needed, exercises on your real work and a live demonstration. Here is how it runs.

A group of 4 to 10 people: A size that gives everyone time to do the exercises on their own work and ask their questions. A whole team can attend together.
In person or by video call: The workshop runs at your premises or remotely. By video call, the exercises stay individual and the materials are shared on screen; in person, exchanges between participants are more direct.
No technical background needed: No need to know how to code or to master a particular tool. The useful notions are explained as the session goes, starting from concrete situations.
Exercises on your own work: Each exercise is done on a real task chosen by the participant: filing documents, replying to recurring requests, entering data. What you produce during the workshop belongs to you.
A live demonstration of a local agent: A no-code agent is built in front of you on a fictional case: it reads a document, extracts the information, proposes a reply and stops before any action. Strengths and limits are visible under real conditions.
One deliverable per participant: Each participant leaves with their prioritised use cases, their filled-in map and their 90-day plan. The workshop produces materials you can use, not just notes.

Why this workshop

Why a workshop rather than classic training

Training passes on general knowledge; a workshop produces decisions and materials. Here is what changes in practice.

Immediate practice: Training passes on general knowledge; the workshop works directly on your tasks. You test the method on your own work rather than on examples from elsewhere.
A small group: Four to ten people, a format where everyone takes part. Each exercise is discussed during the session, instead of staying homework for later.
Personal use cases: You leave with a list of tasks ranked by priority, specific to your company. The choice criterion is not the technology, but the time the task costs you.
A prototype that leaves with you: The local agent example seen in the demonstration can be reused as a starting point back home. You get a concrete measure of what a rollout would mean.
A moderate entry cost: The half-day limits the commitment before deciding on a wider project. It is a way to test whether an AI agent makes sense without launching a full undertaking.
An informed decision for what comes next: By the end of the workshop, you know whether an agent makes sense, which task to start with and which data is involved. You can then choose an audit, training or development with full knowledge of the facts.

FAQ

Frequently asked questions

Answers to the most common questions before booking an AI agent workshop.

Do we need to know how to code?

No. The workshop is designed for people with no technical background. The demonstrations use no-code tools and each exercise is guided step by step.

What should we prepare before the workshop?

A list of repetitive tasks that cost you time: sorting documents, replying to recurring requests, entering data. That is the only prerequisite, nothing else.

Do we leave with a working prototype?

Each participant leaves with a limited demonstration prototype, which serves as a starting point. It shows the principle on a fictional case; adapting it to your real data is the next step.

Does the workshop expose our data?

No. The exercises and the demonstration use fictional data, and the architecture shown is local, on the company's own hardware. No real data is sent to an outside service during the session.

How many people can take part?

The workshop takes 4 to 10 people. Below that, the exchanges are less rich; above it, it becomes hard to run a personalised exercise for everyone.

What happens after the workshop?

You can continue with an audit of your processes or bespoke AI agent development. The workshop gives you what you need to choose what comes next with full knowledge of the facts.

Organise a workshop at your company

A group of 4 to 10 people, in person or by video call. Each participant leaves with their prioritised use cases, their map and a demo prototype. Contact us for availability and the full schedule.

Book a call

To go further

Other formats complement this offer, depending on where you're starting from.