KEIRIN
Public Review Edition 0.9

Keirin Framework

The AI-Augmented Agile Delivery Framework
The machine sets the pace.
Your team wins the race.

Keirin embeds governed AI into the delivery cadence, accelerating performance while humans remain in control. It turns scattered, unmeasured AI use into a governed, context-aware and measurable delivery capability.

Download the framework (PDF) View on Zenodo
DOI10.5281/zenodo.21758038 Amr Abulnaga · August 2026 · CC BY 4.0
01

What Keirin is

Governed AI inside the cadence your teams already run

Keirin is an amplification layer, not a replacement. It doesn’t ask teams to abandon the agile framework they already run. It adds governed AI to every ceremony, with a human gate on each, so teams gain speed without giving up accountability.

AI drafts, analyzes and summarizes. People decide. Every AI use case earns its place on a one-page canvas before it enters a ceremony, and every gain is reported alongside its guardrails.

It plugs into
ScrumSAFeLeSSScrum@Scale
In keirin track cycling, a pacing bike, the derny, brings riders up to a speed they could not reach alone, then pulls off the track. The riders sprint for the line themselves. AI is the derny. It never finishes the race.
02

Six layers, two configurations

Core stays teachable in two days

5Measurement & LearningFive result classes, reported with stability guardrails
4Governance & AssuranceGreen / Amber / Red data, review gates, incidents, change control
3Cadence IntegrationAI woven into every ceremony, with a human gate on each
2Use-Case & Work DesignDay-0 Canvas, AI Use-Case Register, risk tiering
1Readiness & ContextAgile hygiene, Team Context Repository, tooling, AI literacy
0Outcomes & ValueObjectives, AI opportunity portfolio, capacity reinvestment
Core · Layers 1–5 at team scope. Every pilot starts here.
Enterprise · Adds Layer 0: outcomes, portfolio and reinvestment.
Signature artifact

Day-0 Value & Assurance Canvas. One page per AI use case: outcome, work, suitability, context, data, accountability, validation, evidence, impact, fallback, technology. Every use case earns its place before it enters a ceremony.

03

Who does what

AI drafts. People decide.

AI assistantPaces, never races

Drafts, analyzes and summarizes across every ceremony, grounded in the Team Context Repository. A context-aware assistant, not a colleague. Unattended runs fail closed when context is missing.

RoleKeeps human, never delegated
Product OwnerPrioritization, acceptance, stakeholder relationships
Scrum Master / Delivery LeadFacilitation, conflict, coaching; curates the prompt library
Developers / MakersDesign decisions, review sign-off, understanding what ships
QA / QualityThe call on “good enough to release”
Business Analyst / DesignerElicitation conversations, design judgment
Engineering / Functional LeadArchitecture, standards, people decisions
Sponsor / LeadershipTooling budget, data-policy sign-off; never sets AI usage quotas
AI-Agile ChampionNEWPart-time local onboarding and library care, 2–4 hrs a week
04

Rollout, middle-out

Top-down enablement, bottom-up proof

Phase 0WK 1–2

Diagnose

Baseline, opportunity heat map, draft Canvases, signed reinvestment agreement

Phase 1WK 3–8

Pilot

One volunteer team, about three sprints, before/after report

Phase 2WK 9–20

Scale

Waves of 2–3 teams, each funded by the last wave’s numbers

Phase 3WK 21–26

Institutionalize

Champion network owns the practice; governance moves to risk and quality

Executives make four decisions up front: sponsorship, budget and tooling, governance sign-off, and capacity reinvestment. Not a rollout mandate. Scaling waits for the pilot’s numbers.

05

Outcomes

Five result classes, reported together

Business valuebenefit vs hypothesis
Customer outcomesreached customers
Flow & deliverycycle time, throughput
AI assurancefactuality, incidents
Human sustainabilityload, learning, trust
Guardrails travel with every gain. A throughput gain with rising instability is reported as a problem, not a success.
Reclaimed capacity is reinvested, never silently harvested. The sponsor signs it at Phase 0.
Day-0 CanvasOne page per AI use case
Prompt & Pattern Library48 prompts across 12 series
Interaction patterns9 named patterns
Foundations Workshop2-day delivery kit
06

Read & cite

Open for public review

Abulnaga, A. (2026). Keirin Framework: The AI-Augmented Agile Delivery Framework (Version 0.9 — Public Review Edition). Zenodo. https://doi.org/10.5281/zenodo.21758038

Review it. Pilot it. Tell me what breaks.

Feedback on the Public Review Edition shapes Keirin 1.0.

feedback@keirinframework.com