Course
catalogue

Five modules, M1 to M5, from executive awareness to the in-situ agentization audit.

Every workshop is hands-on: each module ends with a working deliverable participants take home. No module is a slideshow.

Modules — M1 M2 M3 M4 M5

M1 — AI agent fundamentals

Duration : ½ day (3 h 30)

Objectives: understand what an AI agent actually does (plan, call tools, iterate until validated) and where it fails; distinguish chatbot, assisted tool and autonomous agent; choose the right level of automation for a given use case.
Audience: managers, project leads, decision makers; no technical prerequisites (demos are live, not video).
Prerequisites: a laptop for the hands-on part (30 min of guided practice).
Deliverables / measurable outcomes:
- every participant leaves with 3 candidate automations identified in their own workflows (use-case sheet provided);
- short comprehension test (10-question quiz) — target ≥ 8/10;
- an argued decision: “let’s talk it over” vs. “let’s move to M2/M3”.

Duration : 1 day (6 h 30)

Objectives: set up an agent on your real tasks (writing, document summarization, data processing, reporting e-mails); learn to write good instructions and verify outputs; know when the agent should stop and ask for a human opinion.
Audience: operational teams (administrative, technical, commercial); comfortable with office computing.
Prerequisites: having taken M1 or already an active user of AI tools; a concrete use case brought by the participant.
Deliverables / measurable outcomes:
- 1 automated and tested workflow per pair, replicable at their place (script + written instructions);
- a “trust” grid: for which type of output the agent may act alone, for which not — validated the same day from field feedback;
- a simple before/after measurement on the chosen use case (estimated processing time vs. time measured in session).

Duration : 2 days (12 h)

Objectives: understand and reproduce an agent-assisted development pipeline: an issue base (beads) as the source of truth, an iterative loop that works by itself (ralph), and specialized sub-agents that check each other (worker, tester, reviewer, independent verifier).
Audience: developers, DevOps, data engineers, tech leads.
Prerequisites: comfort with the terminal, git, and a scripting language (Python); having taken M2 (strongly recommended).
Content anchored in the real thing: teardown of an unsupervised night run (automatic task selection, ~5 h of execution, end-of-night report) and of a multi-sub-agent harness with cross-model verification.
Deliverables / measurable outcomes:
- the participant has run their own autonomous loop on a small repository (10+ iterations, zero intervention);
- their issue base is initialized and synced (2+ issues closed by the end of the workshop);
- an “agent rules” document (what it may / may not do alone — e.g. never commit/push without human validation).

Duration : 1 to 2 days (depending on hardware)

Objectives: run an LLM on a 12–24 GB card (llama.cpp and derivatives), understand quantization, context and throughput (tok/s), and turn it into a local service with an OpenAI-compatible API; generate images with Stable Diffusion while keeping control (LoRA, post-processing) — and, if time allows, an identity pipeline that keeps a face from a photo in the generated image.
Audience: ML/AI engineers, data scientists, infra teams who have hardware (or know how to size it) and confidentiality constraints.
Prerequisites: basic Linux, Docker, an NVIDIA GPU (or access to a remote machine; the workshop can run on provided hardware).
Deliverables / measurable outcomes:
- a measured local LLM server (tok/s throughput and retained context written in the session report — we learn to measure, not to trust other people’s benchmarks);
- a locally generated image + a reproducible script;
- a written sizing: which GPU, which quantization, which model for their case (document taken home per pair).

Duration : 2 to 3 days on-site

Objectives: map the company’s workflows, score what can be automated now with agent tools (and what cannot), quantify the gain per workflow, and produce a prioritized roadmap leadership can execute without me.
Audience: executives + 3 to 6 key people; the audit is done inside the company, on its working data (anonymized if needed).
Prerequisites: access to 2 to 3 people per service involved for 1-hour interviews.
Deliverables / measurable outcomes:
- a workflow map (10–25 workflows covered, maturity score 0–4 for each);
- a prioritized roadmap: top 5 automations with estimated gain (hours/week) and risk, sequenced over 6 months;
- a stack recommendation (local vs. hybrid) with cost/risk justification — written to be read by the executive committee.