Team project brief · Paris 2026
This brief shows the team registration form, in one place. Click any question with a + to see what a strong answer looks like, with an example from a past hackathon.
Section No.
Examples
Three recent public hackathon outcomes. The pattern that wins: a real problem, a narrow slice, a working demo on Friday.
Enterprise risk / AI governance
HackEurope 2024 · 30 hours · Grand winner, 1,000 plus builders
A privacy gate that intercepts sensitive inputs (PII, medical records, card numbers) before they reach ChatGPT, Gemini, or Claude.
Beta pilot with University of Bath's Acceler8 society. Target: regulated sectors (healthcare, finance).
Team: 1 compliance lead, 2 engineers, 1 PM
Healthcare / patient experience
Anthropic Hackathon · June 2025 · 3rd place, 13K entries
Physician notes turned into a plain-language care summary the patient can understand, with risk flags and follow-up reminders.
Piloting with 2 hospital networks in Belgium. FDA path in scope.
Team: 1 clinician, 2 engineers, 1 designer
Accessibility / content ops
Google Cloud x Formula E · London, July 2024 · Grand winner
A Vertex AI agent turns a 2-hour Formula E race broadcast into a 2-minute multi-language summary podcast for visually-impaired fans.
Live partnership with the RNIB (Royal National Institute of Blind People). Rolling out for Season 12 race weekends.
Team: 1 PM, 2 engineers, 1 audio lead
Section No.
The form
Sixteen questions. About 10 minutes once your team has agreed on the project. Five questions need a written answer; the rows marked with + open guidance and an example for those five. The rest are simple fields.
Strong projects start with a real, costly problem your CEO has already named. Not an idea your team might explore. Not a long-term roadmap item. A problem your team feels every week, that your CEO would fund this quarter.
Name the problem clearly. Who has it. When it happens in their day. What they do today to work around it. The more specific your team is about the person and the moment, the easier it is to build something they would actually use.
Your team's words are fine. Aim for 3 to 5 clear sentences as a starting point. Longer answers are welcome where they help reviewers understand the moment, the role, and the cost.
A strong answer
“Our company is a mid-market B2B SaaS platform with about 1,100 customers and a 9-person support team. Inbound tickets land in three places: email, the in-app help widget, and a Slack channel reserved for top-tier accounts. We field 400 to 600 tickets per week. Support managers triage every ticket by hand, deciding urgency, picking the right queue (billing, product bug, integration, security), and rewriting the same first-response message they have written hundreds of times. High-urgency tickets sit for 30 to 90 minutes before the right rep sees them, and the product team only catches recurring issues during quarterly reviews. Our CEO has asked support and product to cut response time on top-priority tickets in half, and to surface emerging product issues to engineering inside the same week the tickets are filed. In customer exit interviews this year, 2 of our 4 enterprise non-renewals named slow support response as a top-3 reason for leaving, which puts roughly 180,000 dollars of ARR at risk per missed renewal.”
Names the role, the moment, and the measurable cost.
A weak answer
“We want to use AI to make our team faster.”
No role, no moment, no number.
Judges will ask what your project saves or earns before they ask what you built. Decide the business value first. Your team will use it as a target during the 24 hours, and as a clear answer when you present on Friday.
Your best estimate is fine. Revenue earned. Cost saved. Hours returned. Risk reduced. One number for one dimension is enough. The 30, 60, and 90-day reviews after Paris will measure your project against this same number, so an honest estimate now is worth more than an ambitious one. Show the math if it helps reviewers see how you got there. Longer answers are welcome.
A strong answer
“Support managers spend roughly 18 hours per week each on triage and first-response drafting. Across our 3 senior support managers, that is 54 hours per week, or about 2,700 hours per year. At a fully loaded cost of 80 dollars per hour, the triage workload alone costs 216,000 dollars per year. We target a 60 percent reduction in manual triage time once the assistant is classifying urgency and drafting responses for human approval. That is roughly 32 hours per week back to the team, or 130,000 dollars per year in returned capacity. Separately, the weekly recurring-issue summary is expected to save the product team 6 to 8 hours per week of ticket review and shorten the cycle from problem detection to product fix from about 6 weeks to 2 weeks. We will measure against these numbers at the 30, 60, and 90-day reviews after Paris. The 32 hours per week we expect back will go to proactive outreach on at-risk accounts and a tighter weekly product-feedback loop, both of which our CEO has asked for and neither of which we have headcount for today.”
One number, one calculation, one comparable unit.
A weak answer
“Significant productivity gains across the company.”
Adjectives, no numbers.
Arrive with a rough plan, not a finished system. Two days is enough to prove a small piece works on real data. It is not enough to build a complete product. Pick a small, sharp piece and finish it. Name the components your team plans to use, the data you will start with, the person who reviews the AI's output, and the narrow first slice you intend to ship by Friday. Longer answers are welcome.
A strong answer
“Gemini reads each incoming ticket, the customer's account metadata, and the last 90 days of ticket history for that account. It outputs three things for the support rep to review before any customer sees them: a confidence-scored urgency label (P0 to P3), a suggested routing queue with the reason it picked that queue, and a draft first-response message keyed to the ticket type. The support rep approves, edits, or rejects each suggestion inside our existing Zendesk console, so nothing reaches the customer without a human in the loop. A second weekly pass clusters the past seven days of tickets, surfaces the top three recurring product issues, and writes a short briefing for the head of product. We will measure the classifier against support managers' own labels on a 500-ticket holdout set, with a target of 85 percent precision on P0 and P1 urgency labels. Below 70 percent model confidence on a ticket, the system shows the suggestion as low confidence and the rep starts from a blank draft. For the hackathon, we will start with approximately 6,000 historical tickets (3 months at our current volume), each run through a PII redaction step that strips customer names, email addresses, and account IDs before anything reaches Gemini, and ship the urgency classifier and response drafter for one ticket type: integration errors.”
Names the components, the person who reviews, and the narrow first slice.
A weak answer
“Use AI and machine learning to automate everything.”
A list of capabilities, no system.
Tell us which AI tools your team plans to use. Gemini and Google Cloud are the defaults in the room. If your team is fluent in another tool (Claude Code, Cursor, Lovable, Base44, GitHub Copilot, others), name it on the form. We will route the request to Google for approval before the event.
Your team's AI experience level helps us match the right mentors to your team. There is no wrong answer. "Exploring" teams get more direct support from engineers. "Advanced" teams get more detailed architecture feedback. A few sentences about what your team has shipped, used, or experimented with helps us match the right mentors and the right Friday judges.
Submission deadline is set 2 to 3 weeks before the event. Eurazeo will confirm the exact date when invitations go out.
Confirm your teamSection No.
What comes next
Three things happen after Eurazeo receives your team's form.
Mid-June
A 60-minute walkthrough of Google AI tools. A 30-minute walkthrough of the hackathon itself. Both are optional. We share recordings if your team cannot join live.
Late June
Each person on your team fills out a 5-minute form. Dietary needs, accessibility, and 5 questions on AI experience. The same questions appear at the 30-day review, so you can show your CEO how much your team learned.
July 9 and 10
Google engineers, mentors, and judges are in the room with your team. We match them to teams whose approach fits their expertise. Bring your laptops, your team, and the prototype you have started.
Ready to confirm your team?
Once your team submits, Eurazeo and Yelin will respond with logistics and the mid-June enablement schedule.