From chatbot use to agent setup
Some participants had already used chatbots and wanted to understand what changes when agents can use tools, context and multi-step workflows.
From AI curiosity to agents your team can trust
A hands-on Agentic AI workshop for Swiss teams that want practical automation without losing judgement, source quality or human review.
Real outputs. Six agent briefs, public and reusable.
Real governance. Source rules, refusals and human checkpoints.
AI Workshop
AI Agent Design Canvas
Mission
Sources
Tools
Memory
Output
Refusals
Human review
Risks
Pilot metric
Download the working tool
The canvas we use in every workshop
The fastest way to make Agentic AI practical is to force the right questions early: what the agent is allowed to do, which sources it can trust, when it must refuse and who reviews the output.
Case study
The Lausanne room made agentic AI concrete
On 10 May 2026, a mixed group arrived at Espace Dickens with the same tension many teams feel now: AI is everywhere, but reliable agent work still feels abstract. Their pre-work showed the real need: practical tools, clearer agent setup and business use cases that could survive outside the training room.
Date
Sunday 10 May 2026
Venue
Espace Dickens, Lausanne
Room
Finance, pricing, UX, wealth advisory and engineering profiles
Output
6 anonymised agent briefs published as a public portfolio
Before the workshop
The strongest signal came before the first slide
Participants answered three short questions before the session. That made the workshop sharper: the room did not need AI hype. It needed a practical bridge from personal experimentation to governed work.
01
How introverted or extroverted do you consider yourself?
02
What do you hope to achieve by attending?
03
What is your current role or profession?
From curiosity to work application
The room brought finance, pricing, business analysis, wealth advisory, UX and engineering questions, which kept the session grounded in real work.
From tool interest to peer learning
Several people wanted to compare notes with others exploring Claude, coding agents, practical AI use and what is actually applicable today.
Anonymised room profile
Treasury analyst exploring autonomous agents
Pricing manager moving into business analysis
Wealth advisor preparing an independent practice
UX practitioner curious about agentic workflows
Engineer looking for workplace applications
Civil engineer improving daily AI use
Self-assessments sat around the middle of the introversion-extroversion scale, so the format mixed quiet canvas work, pair thinking and concise group discussion.
The story arc
Four moments that turn interest into adoption
01
The room started with one useful question.
Before anyone designed an agent, the group named what should stay human: judgement, accountability, approvals and the final decision. That boundary made the technology feel safer and more useful.
02
Then abstract AI became a visible workflow.
Participants mapped real work into steps: gather sources, compare signals, draft a report, check assumptions, escalate uncertainty. The agent stopped being a vague assistant and became an operating role.
03
MCP turned tool connection into a design decision.
Instead of treating integrations as magic, the session showed how connected agents use approved files, APIs, SaaS tools and local context. This is where governance moved from policy language into practical architecture.
04
The final deliverable was not a prompt.
Each group left with an agent brief: mission, role, authorised sources, tools, output format, refusal rules, verification checklist and a first pilot path. That is the difference between AI excitement and adoption.
The workshop method
A guided path from idea to trusted agent
Agentic AI adoption fails when teams jump straight to tools. This format slows down the right decisions, then accelerates the build.
01
Start with judgement
Define the decision that remains human and the responsibilities the agent is allowed to support.
02
Map the real work
Break one business workflow into repeatable steps, friction points, source needs and escalation moments.
03
Design the agent brief
Specify mission, tools, memory, output format, source rules, refusals and review checkpoints.
04
Build the audit trail
Make outputs separate facts, assumptions, uncertainty, citations and decisions before a pilot begins.
For your team
What you leave with
A practical Agent Design Canvas
Your team gets a reusable structure for designing governed agents before touching production systems.
MCP in plain business language
Participants understand how connected agents use files, APIs, SaaS tools and local context without hand-waving.
A governed agent brief
Each priority use case gets source rules, tool boundaries, refusal logic, human review and success criteria.
A pilot roadmap
You leave with first test, owner, data needs, rollout risk and the metric that proves whether the agent matters.
Real participant outputs
6 agent briefs. Public. Reusable.
These are the anonymised outputs from the Lausanne workshop. Use them as inspiration for your own pilots.
1. Sector briefing agent - European luxury Daily market intelligence with sourced facts, assumptions and open questions.
->
2. KPI consolidation and visualisation agent Extract and format KPIs from Excel and business systems into a management view.
->
3. Finance process documentation agent Read process documents and produce structured summaries or workshop minutes.
->
4. Code development and debugging agent Support development work without making unvalidated functional assumptions.
->
5. Weekly watch agent - Swiss insurance and banks Track public news and market changes, then prepare concise summaries.
->
6. Market watch and daily PDF report agent Collect signals, analyse source material and generate a reusable report.
->
5-star Google reviews
The workshop lands with beginners and builders
The signal that matters most is not applause in the room. It is whether people leave clearer, more confident and ready to apply Agentic AI at work.
★★★★★
Recent Google feedback
View reviews on Google
★★★★★
“Excellent atelier pour se lancer dans les agents IA !”
Henri Mockler
Google review
★★★★★
“Bonne maitrise, pedagogue et patient, surtout avec des novices comme moi.”
Alessandro Condemi
Google review
★★★★★
“George offers generous, informed guidance on agentic AI. I highly recommend his workshops.”
E Hunter
Local Guide on Google
Ready to design agents that deliver real value?
Tell us your workflows, your data constraints and the first agent use case worth testing. We will shape the right workshop format for your team in Switzerland.