How to Run an AI Agents Workshop for Students (90-Minute Format)

Written by

Vasco PedroFounder & CEO

Vasco Pedro is the founder and CEO of Spinnable, the platform behind autonomous AI workers. He writes about AI workers, team automation, and the future of work.

Vasco Pedro
5 min read
Published: November 12, 2026 (6d ago)

Most students have used a chatbot. Very few have built an AI agent that does a real job, on a schedule, with a teammate's inbox or a shared folder. That gap is exactly what a good classroom workshop closes in 90 minutes.

This guide is for lecturers, programme coordinators and student club leads who want to run a hands-on AI agents workshop for students. It covers the format, a minute-by-minute agenda, the exercises that work, and how to set things up so every student leaves with a working AI worker rather than a slide deck.

Why a hands-on AI agents workshop beats a lecture

A lecture about AI agents gives students vocabulary. A workshop gives them a habit. When students build an agent themselves, three things happen that a talk can't produce:

  • They see the difference between chatting and delegating. A chatbot answers a question. An AI worker takes a task, uses tools (email, documents, the web) and comes back with a result, often on a schedule.
  • They learn to write instructions, not prompts. Describing a job clearly enough for an agent to do it is the same skill as briefing a colleague. It transfers to every course and every internship.
  • They meet the limits early. Agents get things wrong. A workshop is a safe place to discover where human review is needed, before it matters.

In our own sessions with university classes, about three in four students who signed up during the session built a working AI worker before it ended. Building in the first half hour is what makes that happen.

The 90-minute workshop format

Ninety minutes fits a standard class slot and is long enough for every student to finish one build. Here is the agenda we use.

TimeBlockWhat students do
0–10 minFramingSee one live example: an AI worker that researches a company and drafts a one-page brief.
10–20 minSign up togetherEveryone creates an account and joins the class workspace at the same time.
20–50 minBuild 1: a research workerCreate a worker, give it a role and instructions, run it on a real topic from the course.
50–70 minBuild 2: a team jobIn groups, set up a worker that serves the whole project team (meeting notes, task tracking, weekly summary).
70–85 minReview and limitsGroups share one result, point out one mistake the agent made and how they caught it.
85–90 minNext stepsEach student schedules one recurring task for the coming week.

The single most important block is the ten minutes of signing up together. If students sign up "later", most never do. Doing it live, with the presenter walking through each screen, is what turns a demo into a hands-on session.

Planning a session? Create a free Spinnable workspace and try the agenda yourself first.

Three exercises that work in a classroom

1. The company research brief

Ask students to build a worker whose job is to research a company relevant to the course (a case study, a potential client, a competitor) and return a one-page brief with sources. It is concrete, it shows web research and writing in one task, and the output is easy to check against reality.

2. The project team assistant

Each project group sets up one worker for the team: it summarises meeting notes, keeps a list of who owes what, and sends a short weekly recap. This is where students see an agent as a teammate rather than a tool, and it carries on long after the workshop ends.

3. The critical review

Have students give their worker a task where accuracy matters (for example, summarising a reading with page references) and then check it line by line. Finding the agent's mistakes is the most valuable ten minutes of the session. It teaches the review habit that keeps AI use honest.

How to set it up so the room isn't waiting

Workshops fail on logistics, not content. A short checklist:

  1. Create one class workspace in advance. In Spinnable, a workspace can let anyone with the same email domain join automatically, so students with a university address land in the right place without individual invites.
  2. Prepare one shared example worker. Students copy it and change the instructions, instead of starting from a blank page.
  3. Pick course-relevant topics beforehand. "Research this company" works far better than "try anything".
  4. Set spending limits. A shared budget with per-member credit limits keeps a room of 40 students from surprising anyone. The Usage page shows exactly who used what afterwards.
  5. Use a short, tracked sign-up link on the first slide. It saves typing and lets you see how many students actually joined.

Academic integrity: set the rules at the start

Students will ask, so answer it in the framing block. A simple rule set works well:

  • Agents can research, summarise, organise and draft. Students own the final work and every claim in it.
  • Any submitted work that used an agent says so, and how.
  • Sources the agent finds must be opened and checked before they are cited.
  • Follow your institution's AI policy where it is stricter.

Framing agents as research assistants with a duty to check, rather than as essay writers, keeps the session aligned with most university policies.

What it costs a class

Per-seat AI tools get expensive fast for a class of 40. Spinnable charges for the work the AI workers do, not for the number of people, so the whole class, and every project group, can join the same workspace at no per-seat cost. For a single workshop, most of the usage is the two short builds, and the per-member limits mean you decide the ceiling in advance.

Common mistakes to avoid

  • Too much talking before building. Keep the framing to ten minutes. Students engage once their own worker produces something.
  • Generic exercises. Tie every build to the course or the students' current project.
  • Skipping the review block. Without it, students leave thinking agents are always right, or never trustworthy. Both are wrong.
  • No follow-up task. A scheduled task for next week is what turns one session into a habit.

After the workshop

Send students the guide to AI agents for student teams so project groups can keep going, and point anyone curious about rolling this out more widely to how teams roll out AI workers and how to keep AI agent costs under control.

FAQ

Do students need coding skills?

No. Workers are set up by describing the job in plain language. Students with coding skills can go further, but nobody is left out.

How many students can one presenter handle?

Forty works with one presenter and one helper walking the room. The sign-up block is where a helper matters most.

Can it run online?

Yes, but keep cameras on during the builds and use breakout rooms for the team exercise.

What do students keep afterwards?

Their account, their workers and any scheduled tasks. The project team assistant usually keeps running for the rest of the semester.

Run your first AI agents workshop

You can try the whole flow yourself in about twenty minutes before you take it into a classroom: create a workspace, build the research worker, and invite a colleague to join it.

Start free on Spinnable and build your first AI worker →

Already have an account? Invite your class or your project team to your workspace. There's no per-seat cost, so everyone can join.

Share this post

View as Markdown

About the editorial team

Vasco Pedro

Vasco Pedro

Founder & CEO

Vasco Pedro is the founder and CEO of Spinnable, the platform behind autonomous AI workers. He writes about AI workers, team automation, and the future of work.

Get Spinnable in your Google results

Add us as a preferred source to see more of our articles in Google Search, Top Stories, and AI Overviews.

Read more

Your next team member is one click away

Choose a role or describe one. Ready in under a minute.

Start your free trial

Ready in 60 seconds

Pick a template or describe your ideal hire.

No technical setup

No coding, no complex integrations.

Free for 15 days

Full access. Cancel anytime.