Stickies, RIP? Has AI Killed the Design Sprint?

Short answer: the five-day, “everyone in a room with Post-its” ritual is pretty much a memory relegated to the nostalgic days of in-person workshops, stickies, sharpies and dot voting. AI hasn’t killed problem framing, prototyping, or testing with users but it has gutted the time and people required to do them. So what’s next? How do we reboot the design and discovery sprint, if at all?

What’s actually changed

• AI now does the grunt work of workshops and early analysis. FigJam and Miro can cluster and summarise sticky-notes in seconds; they’ll even turn piles of notes into drafts and summaries. That used to be hours of facilitator time and a designer or two doing affinity maps.

• Research synthesis has sped up materially. UserTesting’s AI Insight Summary and similar tools auto-pull themes and “moments” from sessions so teams jump to patterns faster (and with links back to the source). Nielsen Norman Group’s guidance: treat this as a first pass, then apply human judgement. https://www.usertesting.com/blog/ai-insight-summary
• Peer-reviewed studies show big productivity gains on ideation and analysis.
  • In a large field study at a Fortune 500, AI guidance made novice support agents 14–35% more productive; quality and retention improved. Translation for product: junior staff can get to “usable first drafts” faster.
  • In a Boston Consulting Group experiment, consultants with GPT-4 completed 12% more tasks, 25% faster, and produced >40% higher-quality outputs on creative/product tasks. This is ideation and framing being the heart of day 1–2 of a sprint. https://www.hbs.edu/ris/Publication Files/24-013_d9b45b68-9e74-42d6-a1c6-c72fb70c7282.pdf
  • HCI research finds LLMs accelerate brainwriting, idea selection and early thematic analysis. This is useful for compressing “diverge/converge” without ten people in a room.
• Macro data says a lot of work inside sprints is automatable. McKinsey estimates current gen-AI plus existing tech could automate 60–70% of activities people do today. Affinity mapping, desk research, and “first draft” prototyping are squarely in scope.
• Real world headcount signals are here even if outside design. Klarna’s AI assistant did the work of ~700 FTE support agents with comparable satisfaction scores. It showed how fast “patterned knowledge work” can shrink staffing. However – not all is well with this rapid efficiency shift – Klarna back-pedalled somewhat by re-hiring humans when customers complained! https://www.forbes.com/sites/quickerbettertech/2025/05/18/business-tech-news-klarna-reverses-on-ai-says-customers-like-talking-to-people/

So…do you need fewer people to ideate and research?

Short answer yes – fewer bodies, smaller rooms, shorter cycles. AI knocks out large chunks of time in: desk/competitive research, prompt-led ideation, clustering and early coding of qual data, survey/item drafting, and “blank canvas” UI first drafts. One strong PM/Designer plus a Research Lead can now cover what used to take a cross-functional workshop of 6–10.

But no, you can’t automate the judgement: framing trade-offs, choosing risky bets, recruiting the right participants, interpreting contradictions, and landing a decision in the organisation still need experienced humans. Nielsen Norman Group, among others, are clear: AI analysis is a head-start, not a substitute.

You’ll hire fewer facilitators and note-takers; you still need a sharp product lead, a researcher who designs good studies, and access to real users.

The design sprint isn’t dead but the five-day ceremony is

The GV sprint delivered value because it forced focus and a decision. As much as we used to love the process, that entire ‘theatre’ is now gone. Instead keep the focus and outcome positioning.

An AI-first, three-day sprint that actually fits 2025 could look like this: (Or even shorter!)

• Before the room (async, 24–48h):
  • PM/Designer drafts the problem statement, constraints, and success metric; AI produces a landscape scan, initial JTBD, and a set of rival patterns to review.
  • Research Lead uses AI to draft a test plan and recruit screener; tools prepare baseline themes from prior research.
• Day 1 (2–4 people): Align on the problem, generate options with AI (brainwriting), and let the tool cluster and rank by stated criteria. Decide one route.
• Day 2: Build a prototype using AI-assisted UI from your own libraries; keep fidelity low to avoid cargo-cult copying. (Remember Figma’s “Make Designs” misstep – AI can clone patterns too eagerly.) https://www.figma.com/blog/inside-figma-a-retrospective-on-make-designs/
• Day 3: Test with real users. AI drafts the discussion guide and produces the first-pass synthesis; the researcher writes the insight and decision.

This keeps the sprint’s intent (fast decisions, user-tested artefact) while stripping out the headcount and ceremony.

Quality risks and why “Stickies RIP” is not the whole story

• Homogenisation & IP. AI-generated UIs can drift into look-alikes. Figma paused its auto-design feature after outputs resembled Apple’s Weather app; it later relaunched with guardrails. Use your own systems and keep AI at first-draft only.
• Synthetic users are not users. NNG warns that AI-generated “users” and fully automated analysis miss context, emotion and tacit knowledge. Use them to augment, never to replace, primary research with humans.

Hiring implications for a modern product org

• Smaller, sharper cores. A PM/Design “duo” plus a Research Lead can cover most sprint work with AI, pulling in Engineering/Medical/Compliance for short, decisions-focused windows.
• Upskill over up-staff. The skills that matter: framing and metrics, prompt-led ideation, study design, critical reading of AI analysis, and strong writing for decisions. (McKinsey’s latest surveys show gen-AI use is now routine across leadership; treat AI literacy as table stakes.)
• Research remains a leverage point. One experienced researcher can now supervise AI-accelerated synthesis across multiple squads—if the org protects time with real users.

A pragmatic stance for CPOs

1. Retire the ceremony, keep the sprint’s spine. Time-box to 2–3 days with a small team; run AI for divergence and early synthesis; reserve humans for judgement and user contact.
2. Measure throughput, not theatre. Track time-to-decision and “ideas tested per month” before/after AI. Expect fewer meetings and faster learning cycles.
3. Guard against lowest-common-denominator design. Enforce use of your own design system; keep AI outputs at wireframe level; insist on live user sessions.
4. Hire for judgement; train for AI. Shrink the workshop-heavy headcount; grow the number of people who can design a good study, read messy evidence, and make a call.