How to Run a 90-Minute AI Design Sprint (with prompts)
A repeatable framework for product teams to think, design, and ship smarter with AI.
Most teams still run ideation sessions with a whiteboard, a problem statement, and a flurry of post-its. To be honest, I’ve always loved a good Design sprint, especially in person and I hope those don’t go away for anyone because they’re an awesome way to learn and connect together.
But with AI, the way we generate, evaluate, and shape ideas has fundamentally shifted. You can collapse days of thinking into a focused 90-minute sprint if you know how to structure it well.
This is the format designed to move fast without losing the depth. It blends design thinking, systems thinking, and agent-era AI capabilities into a repeatable flow you can run any time your team needs clarity.
Here’s the 90-minute AI Design Sprint, step by step with prompts you can copy, paste, and use today.
0–10 minutes: Set the stage
Before you touch a model or start generating ideas, you need clarity on the problem’s physics: time to deep on this.
Use this phase to:
Frame the opportunity
Define the user, the workflow, and the pain
Clarify constraints
Agree on what “a good outcome” looks like
Here’s the fastest way to do that with an AI copilot.
Prompt
Act as a senior product strategist. I want to run an AI design sprint.
Help me clarify the problem space by giving me:
1. A sharp problem statement (one sentence)
2. The 3–5 most important user types affected
3. The core job-to-be-done
4. The measurable pain or friction
5. The constraints we need to respect (technical, ethical, time, data)
6. What “a meaningful win” looks like
Here’s the context: [paste your context]
Outcome
A clean, shareable 1-page brief that becomes the anchor for the rest of the sprint.
10–30 minutes: Explore the solution space (Diverge)
Now that the problem is framed, it’s time to stretch the edges of what’s possible.
The goal in this phase is quantity + diversity, not refinement (my in-person human alone teams have come up with 500+ ideas in a day).
You want ideas across different horizons:
Incremental (what AI can automate today)
Adjacent (new workflows, rewired experiences)
Transformational (agentic systems, reimagined jobs, new value pools)
This is where AI is uniquely powerful and it can generate literally hundreds of possibilities, but only if you prompt it with range.
Prompt
Act as a hybrid PM + AI architect + UX Lead.
Generate a diverse set of solution directions for this problem.
Give me 12 concepts broken into these categories:
1. Quick wins (can be built in < 4 weeks)
2. Workflow upgrades (AI enhances, predicts, or personalizes steps)
3. Agentic concepts (AI acts autonomously with user oversight)
4. Wildcards (non-obvious, high-upside explorations)
For each concept include:
- A one-sentence summary
- The value to the user
- Why AI is meaningfully suited for this
- Major risks or dependencies
Outcome
A wide-ranging landscape of possible directions giving you raw material for convergence.
30–50 minutes: Converge and pick the strongest concepts
This phase is about reducing noise.
The team identifies what’s actually worth developing.
You can use this window to evaluate ideas through three lenses:
Desirability (Does anyone want this?)
Feasibility (Can we build it soon?)
Differentiation (Does this meaningfully stand apart?)
AI can help you triangulate quickly and reduce those blind spots.
Prompt
Evaluate these concepts using a desirability/feasibility/differentiation lens.
1. Score each concept from 1–5 on each dimension.
2. Identify the top 3 concepts worth exploring.
3. For each top concept, explain:
- Why it stands out
- What assumptions must be tested first
- What success would look like in 30 days
Outcome
A shortlist of 1–3 concepts with a clear rationale which ultimately is the sprint’s direction of travel.
50–70 minutes: Prototype with AI
This is where the sprint becomes really interesting.
The goal is to turn abstract ideas into something concrete enough that you can react to it, like a rough flow, an interaction, a storyboard, a system map, or a lightweight UI.
At this stage we’re not aiming for polish, instead this is all about articulating things clearly.
Use this phase to:
Visualize the user journey
Identify the key moments of intelligence
Sketch interaction patterns
Explore different modes (chat, agents, multimodal, automation, sensing, prediction)
Reveal hidden assumptions
Prompt
Act as a product designer + AI prototyper.
Using this chosen concept, generate:
1. A lightweight user journey (5–7 steps)
2. A proposed interaction model (chat, agentic, UI-led, hybrid…)
3. A rough UI or interaction sketch described in words
4. The data flow and where AI is adding value
5. The biggest unknowns we should test
Make it concise and visual where possible.
Outcome
A tangible proto-concept you can show to your team, share with a stakeholder, or iterate on immediately without waiting for engineering.
70–90 minutes: Stress-test and evaluate
The biggest mistake teams make is stopping at the prototype.
The real insights comes from interrogating it, it’s time to bash it like a piñata.
In this final phase, you pressure-test the prototype from multiple angles:
What breaks?
What’s unsafe?
What’s technically unrealistic?
Where could hallucinations or overconfidence cause harm?
What edge cases destroy the user experience?
What needs human oversight?
This is where AI helps you think like a safety engineer, a systems designer, a red-team analyst, a skeptical stakeholder…
Prompt
Act as a red-team reviewer + responsible AI specialist.
Stress-test this prototype by identifying:
1. Failure modes (UX, technical, safety, ethical)
2. Edge cases and scenarios it will struggle with
3. Risks that require guardrails or human oversight
4. Data quality or model limitations
5. The smallest testable version (an MVP we can validate in < 2 weeks)
6. A confidence score and what would increase it
Outcome
A grounded, reality-checked concept with clear next steps, known risks, and a path to validation.
Now to Wrap-up
A 90-minute AI Design Sprint won’t replace deep research or long-form strategy work.
But it will give you something most teams lack: momentum, clarity, and a concrete starting point.
Instead of spending days circling the problem, you walk away with:
A sharp problem definition
A diverse landscape of possible solutions
A clear set of top concepts
A lightweight prototype
Known risks and next steps
And most importantly I think, a direction the team can act on
Run this sprint once, and you’ll see how quickly it becomes the default way your team tackles ambiguity. AI use on teams isn’t really about accelerating the work imho, but the greatest value is in accelerating understanding.
If you use this sprint with your team (or adapt it), I’d love to hear how it goes.
And if you know a PM, designer, founder, or researcher who’d find it useful, feel free to share the post with them.