“I can prompt AI, but I cannot design an agent.”
A prompt asks for an answer. An agent needs a goal, information, permitted actions, limits and a way to know when the task is done.
BCS AI Lab #1 · Hands-on AI agent workshop · Barcelona or online · 9 October, 19:00
Learn How to Build an AI Agent by Building Two
Move beyond asking AI for answers. In one evening, learn how to turn a goal into two working AI agents, with guidance, a clear method and a result you can adapt to your own work.
9 October 2026Starts at 19:00 CEST
Barcelona or onlineJoin on campus or remotely, live
One evening2.5 hours of guided building
Claude Code or CodexUse the one you already have. In English, on your laptop
What you build tonight
Two agents. One deal. You approve.
The buyer agent wants a lower price. The seller agent protects its floor. They trade structured offers until they reach a deal, hit the round limit or one of them refuses to break a rule. Nothing is final until a human approves it.
Negotiation runRound limit: 5
Buyer agent · Round 1Offer: €1,600 for 10 laptops, delivery in 14 days.
Seller agent · Round 1Counter: €2,100. Delivery in 14 days is possible.
Buyer agent · Round 2Offer: €1,800 if you add a 2-year warranty.
Seller agent · Round 2Counter: €1,950 with warranty. Below €1,900 is not allowed.
Buyer agent · Round 3Accept €1,950. Within budget, warranty included.
Result: Deal reached · Waiting for human approval€1,950 · 10 laptops · 14 days · 2-year warranty
Illustrative example of the exercise. Your agents, rules and numbers will differ.
The quick answer
What is this AI agent workshop?
BCS AI Lab is a hands-on AI agent workshop in Barcelona and online on 9 October 2026 at 19:00. It is for people who want practical AI agents training, not another presentation. In one evening, you build two agents and learn a method you can reuse for your own workflows.
- You finish one contained multi-agent case instead of collecting another unfinished demo.
- You learn a reusable method: role, goal, context, constraints, output, handoff, stop and approval.
- You can change the scenario after the workshop without rebuilding the logic from zero.
- The exercise shows both what AI agents can do and where deterministic rules still matter.
Why this case exists
You have seen the demos. You still have not built the system.
Tutorials make AI agents look simple until you need two roles to disagree, remember their limits and stop at the right moment. This workshop gives you a small enough problem to finish and a deep enough problem to understand.
“I copied a workflow, but I do not understand it.”
You will make the decisions yourself: what each agent knows, what it may reveal, when it can concede and when it must stop.
“The tools change before I finish the tutorial.”
The transferable skill is not memorising one interface. It is turning a real objective into roles, rules, handoffs, tests and human control.
Workshop agenda
From a first idea to two working AI agents
See the system before you build it
Run the finished negotiation once and identify the buyer goal, seller goal, handoff, stop and approval.
OutputA system map you can explain in plain language.
Build the buyer agent
Set the buyer’s budget, must-have terms, flexible preferences and walk-away rule.
OutputA buyer that can propose, counter and stop.
Build the seller agent
Add the seller’s price floor, permitted concessions, payment conditions and rejection rules.
OutputA second agent with genuinely conflicting incentives.
Connect the negotiation loop
Pass structured offers between the agents and add a maximum number of rounds.
OutputA complete exchange between buyer and seller agents with a visible end state.
Try to break it
Change one constraint, remove information and test whether either agent accepts an invalid deal.
OutputA small test sheet covering deal, no deal and rule broken.
Review, adapt and share
Discuss where the pattern can be reused and which real decisions must remain human-controlled.
OutputA reusable template and a clear next build.
Your finished result
You leave with something you can run again
The workshop ends when the system produces a clear result and you can explain why it made that decision. You keep the materials so you can change the scenario later.
Buyer and seller agents
Two reusable role templates with separate goals, information and limits.
Negotiation loop
A structured exchange with offers, counteroffers, a round limit and end states.
Constraint design sheet
A simple way to separate a preference, concession, hard rule and walk-away point.
Failure-case checklist
Scenarios for valid deals, impossible deals, missing data and broken constraints.
You also get the session recording, so you can repeat any step at your own pace.
BCS AI Lab series
One evening. One working agent. Every two weeks.
BCS AI Lab is a series of hands-on evenings. Each Lab takes one real use case and turns it into a working AI agent you can run again. This is Lab #1.
- A new use case in every Lab, from one clear problem to a working result
- The same reusable method: role, goal, rules, handoff, tests and human approval
- Every Lab is added to the BCS AI Agent Library: recording, materials and a short write-up
Reserve a place
Build two AI agents with us
Places are limited because every participant needs build support. Join BCS AI Lab on campus in Barcelona or online live and finish your first controlled multi-agent case in one evening.
9 October 2026 · 19:00 CEST
On campus: Barcelona Code School, Carrer de Paris 157, 08036 Barcelona
Online: join the same workshop live from anywhere
Workshop language: English
AI agent workshop FAQ
Questions before you join
Is this an artificial intelligence workshop for beginners?
The workshop is for confident computer users who already use an AI assistant such as ChatGPT or Claude but have not yet built a multi-agent system. You do not need to be an AI engineer, but you should be ready to work actively on a laptop.
Will I learn how to build an AI agent?
Yes. You will define a role, goal, context, constraints, output and stopping rule for a buyer agent, then repeat the method for a seller agent and connect the two roles in a controlled negotiation loop.
What makes this an agentic AI workshop?
The agents do more than generate a single answer. Each agent works toward a goal, evaluates an incoming offer against rules, chooses a next action and passes a structured response to the other agent. A human still approves the final outcome.
Do the AI agents make a real purchase or sale?
No. The negotiation happens in a safe training scenario. The agents do not move money, accept legal terms or send commitments. The exercise is designed to teach goals, constraints, handoffs, stopping rules and human approval.
What should I bring to the AI workshop?
Bring a laptop and charger with Claude Code or Codex installed, whichever you already have. Before the event, Barcelona Code School will send a short setup checklist so you can arrive ready to build.
Which AI tools do we use in the workshop?
You build with Claude Code or Codex. Use the one you already have: Claude Code comes with a paid Claude plan and Codex comes with a paid ChatGPT plan. The method you learn works the same in both.
Will there be a recording?
Yes. Every participant gets the session recording, so you can repeat any step after the workshop. The Lab is also added to the BCS AI Agent Library.
Can I join the AI agent workshop online?
Yes. You can attend at the Barcelona Code School campus or join the same live session online on 9 October 2026 at 19:00 CEST.
What will I take home?
You will take home the buyer and seller agent templates, the negotiation workflow, a constraint worksheet and a small test sheet for checking deal, no-deal and rule-breaking outcomes.
Is this AI workshop the same as the AI Agent & Automation Bootcamp?
No. BCS AI Lab is one evening and one bounded multi-agent case. The AI Agent & Automation Bootcamp is a four-week, full-time programme for building connected business automations, learning n8n and creating a portfolio-ready Virtual AI Department.
How much does the AI agent workshop cost?
The ticket costs €30. It covers the live workshop, the guided build, the session recording and the reusable learning materials described on the page. If you join the AI Agent & Automation Bootcamp within 7 days after the Lab, the €30 is credited to your tuition.
Your next step
Want to become an AI builder?
One evening shows you how two agents work together. The AI Agent & Automation Bootcamp takes you further: build connected business automations, learn n8n, add data, approvals and reliability, and finish with a portfolio-ready multi-agent system.
For AI Lab participants
Join the Bootcamp within 7 days after the Lab and your €30 ticket is credited to your tuition.