Inside the 3-Day AI Design Sprint

A service-design project starts open-ended. Two workshops in, the team has a prototype direction. Then the next steps drift. Schedule slips, scope expands, deliverables stay vague, weeks turn into months. The product owner wanted certainty in six weeks. Six months later they're still negotiating what "done" means.

The 3-Day AI Design Sprint replaces that pattern with a defined process. Three live days Monday through Wednesday, plus a Day 0 prep session the week before. The polished asset bundle lands in your inbox Friday end-of-day, or the following Monday at the latest: Figma library, prototype URL, research synthesis, signed roadmap, signed decision log.

The shape below is what the engagement looks like in practice. The structure has run before. The AI delta compresses the production layer, not the method.

The shape

Day 0 is where the project becomes possible. A 90-minute working session: align on the problem, draft the Outcome Contract, screen user-test participants, check tool access. By the end of Day 0 we both sign a contract that says exactly what gets built. How many components, how many screens, what tier of prototype, how many user sessions, how the roadmap is shaped. No scope drift, no surprises.

Monday is frame, decide, build. Morning: a sailboat retro on the current product surfaces what's slowing teams down. Three-direction ideation and a vote settle the question worth testing. The user journey gets detailed. Afternoon: the prototype build starts live. You watch how Lovable, v0, Figma Make, and the prompt patterns work in practice. AI is in the room, the tools build with the team, the facilitator drives decisions.

Tuesday is iteration plus user testing. Morning: top three changes on the prototype, live iteration to v2. Afternoon: five user-test sessions back to back. Stakeholders observe via shared link. Between sessions we iterate based on what we just learned.

Wednesday is synthesis and decisions. The research becomes a 90-day phased roadmap. The decision-maker joins for the prototype review and roadmap sign-off. The Decision Log gets signed in the room.

Friday or following Monday is delivery. The polished asset bundle ships: final Figma library, finalised prototype URL, research synthesis PDF, signed roadmap PDF, signed decision log. Day 3 was the live handoff of decisions. The polished assets follow within the same working week.

The half-back guarantee

50% on signed SOW, 50% on delivery. If the assets agreed in your Day 0 Outcome Contract aren't delivered, or are materially different from what was outlined, and you're not happy with what we did deliver, the final invoice isn't sent and half of the initial 50% is refunded. You pay 25% of the total, not 100%. Both sides have skin in the game.

Worked example on a €15,000 engagement: you pay €7,500 upfront. If delivery fails the two-condition test, the second €7,500 invoice doesn't ship and €3,750 of the initial payment is refunded. Net: you paid €3,750, I kept €3,750, the other €7,500 goes uncollected.

Two conditions, both required. Objective: assets not delivered, or materially different from the signed outline. Measurable per the Outcome Contract acceptance criteria. Subjective: you're not happy with what was delivered as a substitute. Pure subjective dissatisfaction with conforming work doesn't trigger the void. A small spec deviation you're happy with doesn't either.

The structure of Day 0 makes delivery failure avoidable rather than theoretical. Locked scope with count-based acceptance criteria. Signed Recruitment SLA. Verified tool access. Confirmed decision-maker attendance. That's why the half-back model works for both sides.

The guarantee is conditional on you meeting the Client Obligations in the contract. If user-test participants don't show up despite the Recruitment SLA, the research synthesis comes off the deliverable list. The rest continues. The void clause covers our work, not yours.

The four prototype tiers

You choose the tier on Day 0 based on what you need to test, not on budget. The price is the same.

Facade: realistic screens with no live data. Useful for brand validation and information-architecture testing.

Simulated-data: interaction logic with seeded data. One happy path plus one or two error states. Useful for workflow validation.

Spike: one narrow integration proof. Often a real LLM call returning real data, or a single backend integration. Useful for technical-risk reduction before committing engineering time.

Production-stub: deployable as a private demo URL, explicitly non-production-grade. Useful for stakeholder demos beyond the workshop, or for follow-on engineering scoping.

Tier changes mid-sprint require written consent. Silent tier downgrade, sold as Production-stub and delivered as Facade, is exactly the failure mode the Outcome Contract and the guarantee are designed to catch.

Heritage

The sprint structure is the one I led at Fraktio on the Admicom dashboard validation (Samuli Rantanen: "Expectations were high, and they were met. The Design Sprint went really well."), the FCG / Kuntarekry public-sector platform sprint with 11 cross-functional participants, already a 3-day Design Sprint with Jari Lepistö on cross-functional alignment, and an audit roll-forward tool sprint for a Big Four professional-services firm, sanitised reference, details on a fit call.

Those are pre-AI proof of the method. The AI-augmented version compresses what used to take 6 to 8 weeks at strategic-firm tier into three live days plus Day 0 plus delivery. New AI-augmented case studies come from the first engagements running on this format.

Pricing

Launch pricing is €15,000 for a solo sprint, moving toward €20,000 once 2 to 3 AI-augmented case studies have shipped. The Calmworks trio configuration (Vitali, Tarek for co-facilitation, Antonija for prototype build and research synthesis) is €35,000 with the same deliverables and a deeper bench. Available from Cycle 3 onward.

Add-ons scale with the engagement. Research depth tiers up: Expanded 8 sessions, Deep 12 sessions. Figma library handoff at €1,000 includes a 2-hour live training session, 30 days of async questions, and a 1-hour check-in at week four.

VAT excluded. Payment: 50% on signed SOW, 50% on delivery, net 14 each. The half-back void clause is invokable within the net-14 window after the final invoice.

What you walk away with

You don't get a deck of recommendations. You get a tested prototype direction, a Figma library you can build on, evidence from real users in your domain, a signed 90-day roadmap, and a decision log that captures every choice the room made, including the ones you discarded and why.

The intended handoff is your engineering team picks up the roadmap and prototype on Monday. The sprint is structured to end cleanly.

Twenty minutes on a call sorts whether this is the right fit and what tier of prototype makes sense for what you're testing. If it isn't the right fit, you'll hear so directly, with a pointer to what is.