A ground-up redesign of the Sprint methodology for distributed teams working asynchronously across time zones, mediated by AI through a chat platform.
Why a Redesign, Not an Adaptation
The original Sprint was designed around a room. Whiteboards ARE the shared brain. Silence IS the voting mechanism. The ticking clock on the wall IS the urgency. A Friday deadline with real customers in the building IS the forcing function.
Strip the room away and you don’t have a Sprint with video calls — you have a broken process. Every core mechanic relies on co-presence:
| Original Mechanic | Why It Breaks Async |
|---|---|
| ”Everyone votes silently” | Without simultaneous presence, early votes anchor late votes |
| Whiteboard as shared brain | No persistent, always-visible shared surface everyone glances at |
| Time Timer creating urgency | No shared clock. People work at different hours. |
| Energy management (breaks, snacks) | You can’t manage energy you can’t see |
| ”Leave the room to check your phone” | There is no room. Distraction is the default state. |
| Facilitator reading body language | No bodies to read |
| Speed critique with a 3-min timer | People are in different time zones. You can’t gather everyone for 3-minute rounds. |
| Friday’s looming deadline | Without a fixed calendar, urgency evaporates |
This spec rebuilds the Sprint from its principles — not its physical rituals.
The Principles That Survive
These are what make sprints work. They’re independent of a room:
- Individual thinking beats group brainstorming — Actually gets BETTER async. No one can shout over anyone.
- Concrete over abstract — Sketches and prototypes, not slide decks and debates. Medium-independent.
- Structured decision-making with a Decider — Works in any medium.
- Time pressure as a feature — Needs a new mechanism, but the principle holds.
- Test with real people — Requires scheduled sessions, but doesn’t require co-location.
- Surface first — Prototype the facade, not the system. Medium-independent.
Structural Changes
Days → Phases
The 5-day structure becomes 5 phases, each with a time window instead of clock hours. The total sprint is capped to prevent drift.
Clock Time → Deadlines
Instead of “10 AM – 5 PM,” each phase has a deadline. Work happens whenever people are available within that window.
Whiteboards → Living Documents
The AI maintains a persistent, always-current Sprint State — the digital equivalent of the whiteboard-covered room. Every team member can check it at any time.
Silent Exercises → Blind Submissions
Instead of “everyone votes at the same time,” everyone submits privately to the AI before a deadline. The AI reveals all submissions simultaneously when the deadline hits. This eliminates anchoring bias more effectively than the original.
Facilitator → AI Mediator
The AI replaces the human Facilitator entirely. It manages phase transitions, enforces deadlines, sequences reveals, synthesizes contributions, and protects the process.
Breaks → Cognitive Load Management
Instead of scheduled breaks, the async sprint manages cognitive load by limiting how many activities are open at once and by separating heavy-thinking phases from light-review phases.
Document Index
| File | Contents |
|---|---|
| async-00-structure.md | Phase structure, timing, and tempo |
| async-01-setup.md | Pre-sprint: challenge, team, platform setup |
| async-02-phase1-understand.md | Phase 1: Goal, map, expert input, target |
| async-03-phase2-diverge.md | Phase 2: Inspiration, individual sketching |
| async-04-phase3-decide.md | Phase 3: Critique, vote, storyboard |
| async-05-phase4-prototype.md | Phase 4: Build the prototype |
| async-06-phase5-test.md | Phase 5: Customer interviews and debrief |
| async-07-ai-mechanics.md | AI mediator behavior, state management, message patterns |
| async-08-exercises.md | Redesigned exercises for async |
| async-09-anti-patterns.md | What goes wrong and how to prevent it |
Async Sprint at a Glance
SETUP (before sprint starts) ├── Challenge selected ├── Team assembled (Decider identified) ├── Platform configured (channels, AI mediator active) └── Customer recruiting begins
PHASE 1: UNDERSTAND (48-hour window) ├── AI opens with challenge brief ├── Team async: draft goals and questions ├── AI synthesizes → team reacts ├── Expert Q&A threads ├── Map drafted by AI from inputs → team refines ├── HMW notes collected (blind) ├── Target voted (blind) → Decider decides └── GATE: AI verifies all outputs before Phase 2 opens
PHASE 2: DIVERGE (48-hour window) ├── Lightning Demo submissions (links + commentary) ├── AI compiles inspiration board ├── Individual sketching (private, deadline-driven) ├── Blind submission to AI └── GATE: All sketches received before Phase 3 opens
PHASE 3: DECIDE (48-hour window) ├── AI reveals all sketches simultaneously (anonymous) ├── Heat map voting (async, blind) ├── AI-narrated critique threads (one per sketch) ├── Straw poll (blind, revealed together) ├── Decider supervote ├── Rumble or all-in-one decision ├── Storyboard built collaboratively └── GATE: Storyboard approved by Decider
PHASE 4: PROTOTYPE (72-hour window) ├── Roles assigned ├── Parallel building with AI coordination ├── Stitcher reviews ├── Trial run (short sync session or recorded walkthrough) ├── Interview script written └── GATE: Prototype complete, customers confirmed
PHASE 5: TEST (48-hour window) ├── 5 customer interviews (scheduled video calls) ├── Team watches recordings and posts observations ├── AI compiles observation grid ├── Pattern identification (individual then shared) ├── Sprint questions reviewed ├── Decider decides next steps └── SPRINT COMPLETE
Total elapsed time: 10–14 calendar days (vs. 5 consecutive days in the original).