Promptathon
A Promptathon is a time-boxed peer-learning event modelled on the hackathon, in which small teams write, test and refine prompts for generative AI tools to solve set challenges. The aim is less to find a winning prompt than to get people working hands-on with AI and learning from each other. The term was first used by the Department of Computer Science at the University of Hamburg in 2023, and the Corporate Learning Community has published an open guideline for running one.
Facilitation script
- 1
Welcome participants, walk through the schedule and state the ground rules, including which data must stay out of the tools.
20 min - 2
Give a short primer on generative AI and on writing prompts, and hand out the cheat sheet. Shorten this if the group is experienced.
20 min - 3
Challenge sponsors present their challenges and what a finished result should include.
15 min - 4
Form teams of three to five with mixed experience and let each team choose or receive a challenge.
10 min - 5
Teams work: write prompts, test, compare outputs and refine. Sponsors and helpers circulate. Give a time check at the halfway point.
120 min - 6
Teams present prompts and results in a few minutes each, followed by questions and feedback.
45 min - 7
Make sure every result is on the shared board, collect participant feedback and say where the prompt collection will live.
15 min
Tips
Write challenges from real work tasks and test each one yourself beforehand, so you know it can be solved in the time and with the tools provided.
Ask teams to keep their failed prompts as well as their best one, because the comparison is where most of the learning sits.
Sort out tool access and data rules before the day: an hour lost to logins or to uncertainty about confidential data costs more than any other planning gap.
Common pitfalls
Challenges that are too abstract or too large, so teams spend the working phase debating the task and produce nothing to show
Tool access that fails on the day, such as missing licences or blocked sites, which stalls whole teams during the most valuable part of the event
No stated rule on sensitive data, with the result that participants paste confidential material into external tools or avoid realistic tasks altogether
Teams made up only of experienced users next to teams of beginners, which removes the peer learning the format is built on
No documentation after the event, so the prompts are lost and the day has no effect on daily work
Variations
The guideline describes events from a few hours to one or two days. A short version for a single department can run in half a day with three or four challenges. Online: teams work in breakout rooms on a shared board and present by screen share. Hybrid needs a helper dedicated to the remote teams. Without a jury the event stays a learning format; with a jury and criteria it becomes a light competition, which suits some cultures and puts off beginners in others.
Where it fits
When to use it
An organisation has given staff access to generative AI tools and most people have barely tried them
You want AI use cases to come from the people who do the work instead of from a central team
Experienced and inexperienced users should learn from each other in mixed teams
A community or cross-department learning day needs a hands-on format that keeps people active for several hours
You want a documented set of tested prompts that others can reuse after the event
When not to use it
Participants have no approved access to AI tools or the data rules are unclear. Settle both first, for example with an AI Working Agreements session
The goal is working software or an integrated product. A hackathon or a Design Sprint fits that better
You have under two hours. Run a short demo and paired practice instead of a full event
The group is a single small team. A normal working session on one real task gives more than an event format
The challenges would require confidential customer or personal data. Rewrite them with dummy data or choose a different format
Related methods
Frequently asked questions
What is Promptathon?▾
A Promptathon is a time-boxed event in which small teams write, test and improve prompts for generative AI tools to solve prepared challenges, then show their results to each other. It borrows the structure of a hackathon and puts the focus on peer learning. Typical events last from a few hours to a full day.
How is a Promptathon different from a hackathon?▾
A hackathon usually aims at a working prototype and needs people who can build one. A Promptathon needs no programming: the work product is a prompt and its output, so anyone who can describe a task in words can take part. The learning matters more than the result.
Do participants need prior AI experience for a Promptathon?▾
No. The format is meant for mixed groups, and a short primer at the start covers the basics. Mixing experience levels within each team is part of the design, because beginners learn from watching others prompt and experienced users learn from explaining.
How many people can take part in a Promptathon?▾
The published guideline describes events of roughly 20 to 200 participants working in teams of three to five. Larger events need more challenges, a sponsor for each challenge and enough helpers, and usually run the presentations in parallel tracks.
What makes a good Promptathon challenge?▾
A good challenge comes from real work, can be attempted in the available time with the available tools, and has a result people can judge, such as a draft, an analysis or a plan. It should not depend on confidential data. Test each challenge before the event.
Plan your next workshop with AI
Workshop Weaver helps you combine methods like Promptathon into a complete, timed agenda in minutes.
Try it freeMethod descriptions on Workshop Weaver are original content written by our team, based on established facilitation practices. This method was inspired by work from Corporate Learning Community (Martina Pumpat, Jochen Robes, Marcel Kirchner), Promptathon guideline v1.1, 2024.