Leeds Business School students build their own AI agents
Two workshops with 218 students lifted confidence in building agents by 86%

Leeds University Business School
Leeds

The challenge
What we set out to solve
Business students are about to graduate into workplaces that AI is changing fast, and employers are looking for people who can already use it. Leeds University Business School had run one series with AI Lab, where every team built one agent. This time they wanted more building, tied to real business problems, for a much bigger cohort: 218 students, most new to AI. The aim was for every student to leave with something they had built and could talk about to an employer.
What we delivered
How the lab was designed
- Two three-hour hands-on workshops on 21 October 2026
- A pre-session video so students arrived knowing the basics
- A printed Use Case Design Spec for every student
- A live agent build in Microsoft Copilot Agent Builder, followed by the whole room
- A team agent for a business brief, built and pitched to the room
- Three AI Lab volunteers helping on the tables
What happened
On the day, in the room
Over 100 students packed into one room, with another group of similar size earlier the same day. Only a small number had ever built an AI agent. Most had only used chat tools. The facilitators told them nobody in the room was an expert, not even them. Then they built. Most of the room had a working agent on their own laptop, following along live. A student stood up and explained, correctly, how several agents can work together under a lead one, and got a round of applause. Then came the fun part. Ten teams of around ten, most meeting for the first time, had half an hour to pick a problem from a bank's brief, design an agent for it and build it. Pitch order was decided by a spin of the wheel. One agent sorted customer complaints by risk and removed the duplicates. Another scored a competitor on how reliable and responsive it was. The winners built Campus Cash, a budgeting tool for first-year students: type in an £80 night out and the week's budget updates. The judges said they could actually see a bank building it. Not everything worked. Screens dropped, a demo stalled, and one team's agent failed in front of the room, so they pitched without it. The room still clapped every team up and the vibe was buzzing!
What people built
Real tools, made by real attendees
Gallery
From the day







Related stories
More proof from the labs

T Level students built working AI agents in two hours
AI Lab Leeds ran a hands-on session with twenty T Level students from Leeds City College's Printworks and Quarry Hill campuses. Students worked in Copilot, Gemini and Claude, learned the difference between an assistant, a skill and an agent, then built their own. They left with working agents they had designed and tested themselves, from a CV builder to a budget tech buyer.
Read the story
Already using AI, missing the guardrails
This LUBS group didn't need persuading. Between them they were already running six AI tools across research, job applications, revision and planning. What none of them had was a way to judge when to trust the output. The session put reliability, bias and accountability at the centre, then split the room into three teams, who each designed and built a working agent in three hours in Microsoft Copilot.
Read the story