By Daniel Rohregger, Microsoft MVP Published
A client told me recently that he’d spent four hours trying to build an agent for a task that simply couldn’t work from a technical standpoint. My honest answer: I could have told you that in four minutes.
That’s exactly why every good agent project starts with an honest feasibility check, not with building. And for that check, I use a format in client projects that keeps proving its worth: the agent hackathon.
The principle: feasibility before the full build
An agent hackathon isn’t a contest for the slickest demo. It’s a way to find out, quickly and honestly, whether an idea holds up technically at all, before weeks go into a live connection to SQL Server, Snowflake, or an ERP system that turns out not to be needed. The two days are a deliberately hard deadline: they keep an idea from getting lost in endless scope debates before anyone has tested whether it even works.
Why business and IT have to show up together
The most important ingredient is who’s on the team, not the technology. Microsoft’s own guidance for Copilot Studio projects lists core roles including a project lead, solution architecture, the actual agent development, someone for success metrics, and someone for security and compliance. That’s a lot more than “one business person plus one IT person.” But the basic rule is what counts: the business side brings process knowledge, test questions, and the call on whether the result actually works for the job. IT, or the platform team, brings the technical build and a realistic sense of what’s possible and how much effort it takes. Neither side can carry the format alone. And that’s not a special case. It’s how agents get built in organizations, long after the hackathon.
The lead-up matters more than the day itself
A good agent hackathon starts two to four weeks before the event, not on the morning of day one. Each team submits, in advance, a specific process that people really do by hand today, and answers four questions: What does the process look like now? What should it look like afterward? What exactly is painful about it today? And which data sources would an agent need? The submissions get prioritized, so that on hackathon day, people only work on ideas with a realistic shot, not on whatever someone comes up with at nine that morning.
Check feasibility first, then build
At the heart of the format is a principle that matches Microsoft’s own approach to assessing risk: instead of building the live connection to a core system right away, you first use Agent Builder with sample or exported data to check whether the basic task can be solved at all. Only if this quick check comes back positive (Microsoft calls it a time-boxed “spike” for the riskiest open question) do you move on to Copilot Studio, with real connectors, multi-step flows, and governance. If it comes back negative, that’s not a failed hackathon day. It’s an early, cheap pivot, long before anyone has promised a finished feature. If you just want a quick read on a single idea, the free pre-check at Agent-Architekt (in German) tells you whether your agent idea is feasible in Copilot Studio.
A two-day structure that works
Day 1: from idea to tested feasibility. Kickoff and a short platform intro, then each team pitches its pre-submitted idea in a few minutes. Next comes the actual feasibility workshop: Is the data source available? Is there system access? Is the rule clear enough for an agent, or does it need human judgment? Are there compliance issues that stop the project cold? Each team then picks the smallest test slice that still means something, not the whole process, and starts on a first prototype, tested against sample or exported data.
Day 2: from prototype to decision. Teams keep building and testing with real questions from the business, backed by dedicated contacts from the platform team. In the afternoon, each team gives a live demo, judged on the same three criteria: business value, technical feasibility, and realistic effort to get to production. Every idea ends with one of three decisions (pursue, park, or drop) and one named person who owns that decision from then on.
The pitfalls that come afterward
The hackathon day itself is rarely the problem. Three things decide whether a good demo turns into something that runs in production:
- Too much scope. Teams try to map the entire process instead of one meaningful slice. A facilitator who narrows things down early matters more than extra build time.
- Synthetic or missing data. Test data that has nothing to do with real production data tells you very little about whether a prototype will hold up later. Preparing a few anonymized datasets that fit the topic is worth more than any extra hour of build time.
- No owner afterward. The most reliable reason a good prototype never makes it to production: after the event, nobody is specifically responsible for it. That’s why the owner question belongs at the end of the agenda, not in an email that goes out at some point after the event.
What usually comes out of a format like this
Across several rounds, the same kinds of use cases keep coming up, no matter the industry: a document or contract review agent that checks incoming paperwork against a fixed set of criteria; a hotline or support agent that serves as the first point of contact, connected to an existing ticketing or knowledge system; a qualification agent that sorts and pre-assesses incoming requests or business cases before a human decides; an onboarding agent that walks people through an internal process step by step; and an information agent that answers questions from a data export first, before anyone even considers a live connection.
If you want to run a format like this for several teams or departments on a fixed date, that’s a job for custom projects.
The full talk this format comes from is on YouTube (in German): “Copilot Studio & Agents: So bereitest du dein Unternehmen vor” (m365 Show, Mirko Peters’ channel).
Sources: Microsoft Learn — Organize hackathons (Power Platform), Design training programs and events, Build your team, Define value before you build your agent, Prioritize risks and identify workarounds, Agent in a Day; AngelHack — The AI Internal Hackathon Playbook, How To Run An AI Agent Hackathon: A 2026 Playbook.
