How to run an AI hackathon: a practical guide for organizers

AI has changed what a team can build in a weekend. With tools like Cursor or Claude at hand, a working prototype in 48 hours is no longer the exception, it's the baseline. And when every team can ship something that looks finished, a polished demo stops telling you much.

That changes the organizer's job. Building is no longer the hard part. What makes an AI hackathon succeed is everything around the code: the problems you choose, the access you set up, the rules you write and the way you judge.

An AI hackathon is not a regular hackathon with AI added on top. Here is how to run one, based on what we have learned from recent events on TAIKAI, from a two-day business lab with 79 teams to sold-out community hacker houses.

Start with a real problem, not "build something with AI"

The fastest way to get vague demos is a vague brief. "Build something with AI" invites chatbots and wrappers that look impressive on stage and go nowhere after it.

At the SpaceXSI Digital Transformation Lab, each of the 79 participating teams arrived with a problem from its own operations: approving invoices faster, ending manual weekly reports, aligning shoe lasts without handling them, turning metal sheet offcuts back into stock. Teams spent their time building, not debating what to build.

A good AI hackathon challenge has three things:

  • A real user who feels the pain today
  • Real data teams can work with, or a realistic sample of it
  • A clear test of what "working" means at the end

If you can't name all three, the challenge isn't ready yet.

Choose the right AI hackathon format for your goal

The format follows the goal. An event built to drive AI adoption in SMEs looks nothing like one built to grow a developer community. These are the four formats we see working best right now:

FormatBest forExample on TAIKAIPlan for
Business lab (in person, 2 days)AI adoption in companiesSpaceXSI Digital Transformation Lab: 218 participants from 79 companies in BragaInviting companies, not only developers, so attendees can act on results
Community hacker house (in person, 1 day)Ecosystem, talent, communityAI Hacker House Solstice (Porto) and Equinox (Lisbon): registrations sold out, €5,000 in prizes across both editionsSponsor tracks, a venue partner, a strong waitlist
Public innovation hackathon (hybrid)Regional and public sector innovationHack4Guimarães 2035: five challenges, linked to the city's bid to become a European Capital of InnovationLocal partners and challenges tied to the region's priorities
Internal hackathonAI adoption inside your own teamOur internal hackathons at LayerXTime protected from day-to-day work, and a path to production

If you're unsure, ask one question: what should exist the week after the event? Signed pilots point to a business lab. New members and partners point to a hacker house. Tools your team uses point to an internal hackathon.

Sort tools, access and data before day one

In a regular hackathon, the first hour goes to forming teams. In an AI hackathon, it often goes to fighting logins, API limits and missing credits. That hour is the most expensive one of the event.

Before day one, make sure you have:

  1. Accounts and credits for every tool you provide, tested by each participant in advance
  2. A setup deadline with a reminder, so problems surface days before, not minutes before
  3. Data ready to use, anonymised or synthetic where needed, with clear permission to use it
  4. A compliance check if real company data is involved. Our guide to AI hackathons and EU compliance covers what enterprise teams need to know

Set clear rules for using AI

Every participant will ask the same question: can I use AI for everything? If your rules don't answer it, teams will answer it themselves, differently, and someone will feel cheated at the end.

Write down, before registrations open:

  • Which tools are allowed, and whether teams can bring their own
  • What must be disclosed: ask teams to state which models and tools they used, as part of the submission
  • What counts as pre-existing work: templates and open-source libraries are usually fine, a product built last month usually isn't
  • What happens with data: what teams can paste into external tools, and what must stay in approved environments

The goal is not to limit AI. It's to make sure every team plays the same game, and that judges know what they're looking at.

Rethink how you judge projects

AI makes it easy to produce a convincing interface in an afternoon. So judging on how a project looks rewards the wrong thing. At the SpaceXSI Lab, the winning team, MAKE IT, earned first place with a live system that turned 253 datasheets into searchable data, not with a better slide.

A few changes that help:

  • Ask for a live demo on real data, not a recorded video or a clickable prototype
  • Score the problem as much as the solution: how real is the pain, and how close is the project to being used?
  • Keep judging close to the work: at the SpaceXSI Lab, judging happened the same day, across four rooms, so every team got time with the jury

The right judging model also depends on your format. A business lab needs a jury that understands the companies' problems. At our AI Hacker Houses, we use public voting on TAIKAI instead, which suits a community event where participants know each other's work.

Plan what happens after the demo

Most of the value of an AI hackathon shows up after it ends, if someone plans for it. Without a next step, even the best prototypes end up as screenshots in a recap post.

At the SpaceXSI Lab, the follow-up was part of the design: every participating company, not only the winners, got a free month of the Enterprise plan and a session with the Enterprise team on implementation and scale. Our own internal hackathons at LayerX work the same way. Several of the internal tools we use today started as hackathon projects.

Before the event, decide:

  • Who owns each project once the weekend is over
  • What support continues: credits, mentoring, a pilot budget
  • When you check back, with a date in the calendar

For teams, our guide on what to do after a hackathon covers how to turn a prototype into a product.

Your AI hackathon checklist

  • Every challenge has a real user, real data and a clear test of success
  • The format matches what should exist the week after the event
  • Tools, accounts and credits are tested before a hard setup deadline
  • Data is cleared for use, and compliance is checked where real company data is involved
  • AI usage rules are written down before registrations open
  • Judging rewards live demos on real data, not polished slides
  • Every project has an owner and a follow-up date

Run your next AI hackathon with TAIKAI

TAIKAI handles registrations, team formation, submissions, judging and public voting in one place, and our team can help you design the challenges and the follow-up. Whether you're planning a business lab, a community event or an internal sprint, talk to us about hosting your AI hackathon, or see how our Workshop & AI Program helps teams turn AI training into real projects.

Rita Simões
Rita Simões
ai hackathonfor organizersfor companies
Ver todos

Assine a newsletter

Mantenha-se atualizado com a economia Web3 e tudo relacionado com o ecossistema!