The Tempo Problem
In a co-located sprint, momentum is automatic. You’re in a room together. There’s nowhere to hide. The clock is on the wall. Friday is coming whether you’re ready or not.
Async kills all of this. Without a forcing function, sprints stretch into weeks, energy dissipates, people context-switch, and nothing gets tested. The single biggest risk of an async sprint is not that the exercises won’t work — it’s that the sprint never finishes.
The entire async design revolves around solving the tempo problem.
Phase Structure
| Phase | Original Day | Time Window | Core Activity |
|---|---|---|---|
| Setup | Before sprint | 1–3 days | Challenge, team, platform config |
| Phase 1: Understand | Monday | 48 hours | Goal, map, experts, target |
| Phase 2: Diverge | Tuesday | 48 hours | Inspiration, sketching |
| Phase 3: Decide | Wednesday | 48 hours | Critique, vote, storyboard |
| Phase 4: Prototype | Thursday | 72 hours | Build the facade |
| Phase 5: Test | Friday | 48 hours | Customer interviews, debrief |
Total: 10–14 calendar days. Hard maximum. If the sprint isn’t done by day 14, it’s failed — declare it and either restart or abandon.
Why These Windows
- 48 hours gives everyone at least one full work cycle regardless of time zone. A person in Tokyo and a person in San Francisco both get two business days to contribute.
- 72 hours for prototyping because building takes longer when you can’t hand things to someone across the table.
- The windows INCLUDE weekends if they fall within the sprint. People don’t have to work weekends, but the clock doesn’t pause for them.
Deadlines Are Sacred
Every phase ends with a hard deadline. The AI enforces them.
What happens at a deadline:
- The AI closes submissions for that phase
- Late submissions are excluded (harsh but necessary — without this, sprints never finish)
- The AI processes whatever it has and opens the next phase
- If critical inputs are missing (e.g., Decider hasn’t voted), the AI escalates directly to the Decider before proceeding
The harshness is the point. In a co-located sprint, 5:00 PM on Wednesday comes whether you’ve finished your storyboard or not. The async equivalent is a deadline that closes a phase regardless of completeness. Teams learn fast to submit on time.
Deadline Warnings
The AI sends warnings at:
- Phase opens: “Phase 2 is now open. Deadline: [date/time UTC]”
- 24 hours remaining: “24 hours left in Phase 2. Here’s what’s still needed: [list]”
- 6 hours remaining: “6 hours. [Name] and [Name] haven’t submitted sketches yet.”
- 1 hour remaining: “Final hour. Anything not submitted will be excluded.”
- Deadline: “Phase 2 is closed. [X] sketches received. Opening Phase 3 now.”
Sync Windows (Optional but Valuable)
Not everything has to be async. Some moments benefit enormously from real-time interaction. The async sprint designates optional sync windows — short, scheduled sessions where the team meets live.
Recommended Sync Points
| When | What | Duration | Priority |
|---|---|---|---|
| Phase 1, near the end | Decider picks the target (after seeing all async input) | 30 min | High |
| Phase 3, during critique | Live speed-critique of top sketches | 60 min | Medium |
| Phase 3, end | Decider supervote + storyboard kickoff | 30 min | High |
| Phase 4, end | Trial run of prototype | 30 min | High |
| Phase 5, end | Team debrief after all interviews | 60 min | High |
Scheduling Sync Windows
- At sprint setup, identify the 2–3 hour overlap window across all time zones
- Schedule sync sessions within that overlap
- If NO overlap exists (e.g., Tokyo + São Paulo), fall back to fully async with recorded walkthroughs
- Never require more than 3 sync sessions across the whole sprint
Activity Types
Every activity in the async sprint falls into one of three types:
1. Solo Work (Async, Individual)
Things each person does independently on their own schedule within the phase window. The AI assigns these and tracks completion.
Examples: writing HMW notes, sketching solutions, submitting votes, writing observation notes.
Key property: No one sees anyone else’s work until the AI reveals it.
2. React Work (Async, Sequential)
Things where people respond to shared material at their own pace. Order doesn’t matter as long as everyone contributes within the window.
Examples: reviewing the map and suggesting edits, reading expert Q&As and adding HMW notes, commenting on sketches during critique.
Key property: Material is visible to all, but contributions are additive (not destructive).
3. Sync Moments (Real-Time, Together)
Short, high-stakes moments where real-time interaction is dramatically more effective.
Examples: Decider choosing the target, speed critique of top sketches, trial run, debrief.
Key property: Scheduled in advance, time-boxed, facilitated by the AI.
How Urgency Works Without a Room
The original sprint creates urgency through physical presence and a visible countdown. The async sprint uses five mechanisms:
1. Hard Deadlines (AI-Enforced)
The phase closes. Late work is excluded. No extensions. No exceptions except by Decider override.
2. Public Progress Dashboards
The AI maintains a visible progress tracker showing who has completed what. Social accountability — no one wants to be the one holding up the team.
Phase 2: Diverge ━━━━━━━━━━━━━━━━━━━━ Deadline: Wed 23:59 UTC
✅ Alex — Sketch submitted ✅ Jordan — Sketch submitted ⏳ Sam — In progress ❌ Riley — Not started ✅ Morgan — Sketch submitted ⏳ Casey — In progress
4/6 submitted · 18 hours remaining
3. AI Nudges
The AI privately messages people who are falling behind. Not nagging — just clear, factual reminders.
4. Phase Gates
You can’t move forward until the current phase is complete. This creates natural bottleneck pressure — everyone knows the team is waiting on the stragglers.
5. Sprint Duration Cap
14 calendar days max. If you’re not done, you’re not sprinting — you’re just having a long meeting. The AI will call it.
Cognitive Load Management
In a room, the Facilitator watches for tired faces and calls breaks. Async doesn’t have this luxury. Instead, the design manages cognitive load structurally:
One Thing at a Time
Each phase has a primary activity. The AI doesn’t dump everything on the team at once. Within a phase, activities are sequenced — you can’t start voting until everyone has read the sketches.
Clear “Start Here” Moments
When a new phase opens, the AI sends one message with exactly what to do first. Not a wall of instructions — just the next step.
Separation of Thinking and Reacting
Heavy thinking (goal-setting, sketching, pattern-finding) happens in solo work blocks. Reacting to others’ work (voting, critiquing, commenting) happens in separate windows. The two are never mixed.
Reasonable Per-Phase Effort
Target per-person time investment per phase:
| Phase | Estimated Hours | What It Covers |
|---|---|---|
| Phase 1 | 3–4 hours | Read brief, draft goals, write questions, review map, HMW notes, vote |
| Phase 2 | 3–4 hours | Review inspiration, take notes, ideate, sketch |
| Phase 3 | 3–4 hours | Review all sketches, vote, critique, contribute to storyboard |
| Phase 4 | 4–8 hours | Build assigned prototype section (varies by role) |
| Phase 5 | 3–4 hours | Watch interviews, take notes, find patterns, debrief |
Total: 16–24 hours per person (vs. ~35 in the original). Async is actually more time-efficient because there’s less idle waiting and ceremony.