A navigation map, not a playbook — adopt selectively, not as a fixed progression.
Facilitate & Capture Events
Clarify Work & Value
Sense Flow, Risk & Impediments
Coach & Improve the Team
Align Stakeholders & Organization
Suggest techniques and questions to improve participation or event design.
Generate questions that expose ambiguity, assumptions, dependencies, missing examples, or weak goal alignment.
Explain plausible causes of aging work, uneven flow, or rising unplanned work from metrics I provide.
Offer coaching questions and alternative interpretations for a recurring team pattern.
Help frame a stakeholder conversation, decision request, or organizational impediment.
Build an agenda and activity plan from the goal, participants, timebox, and issues I identify, revising it as I guide.
Review selected work with me, surface gaps, and refine the questions as I add team context.
Pull current board data and analyze the specific flow signals, risks, or exceptions I request.
Turn observations into a coaching conversation or experiment plan, iterating as I challenge its assumptions.
Draft an audience-specific update, decision request, or escalation from the facts and outcome I define.
Convert notes or a transcript into decisions, unresolved questions, blockers, owners, and follow-up actions.
Review upcoming work and return ambiguity, assumptions, dependencies, missing examples, and connection to the goal.
Compare the current period with the team’s recent baseline and identify material changes plus likely system causes.
Analyze recent team history and propose evidence-backed coaching hypotheses and small experiments.
Synthesize cross-team dependencies, unresolved decisions, and the briefing each stakeholder group needs.
Before each event, assemble relevant decisions, blockers, unanswered questions, and overdue follow-ups.
Alert the team ahead of refinement when agreed ambiguity, dependency, or goal-alignment conditions appear.
Notify when aging WIP, blocked work, queue growth, or unplanned work crosses team-set thresholds.
Surface recurring patterns or stalled improvement experiments when enough evidence accumulates.
Surface unresolved dependencies and decision delays to the right people before they threaten the intended outcome.
Learn which pre-event signals and formats help the team act, then improve prioritization while preserving required information.
Learn which kinds of gaps correlate with rework and reprioritize checks without becoming a gatekeeper.
Tune signal priority against team feedback and outcomes while preserving mandatory alerts and auditing suppressed signals.
Track team-selected experiments and refine coaching hypotheses from observable change, never individual performance scoring.
Learn which stakeholder formats and escalation paths produce timely decisions, then adapt while preserving transparency.
Authority is a separate choice from autonomy
A proactive system can stay read-only; a manually directed one might still change records or send messages. Decide how independently the AI works and, separately, what it is allowed to do.
ReadObserve, retrieve, summarize, compare.
RecommendPropose questions, experiments, alerts, or actions.
ActChange records, send messages, assign work, or alter priorities — only with explicit permission.
Guardrails Across Every Level
AI should increase the team’s capacity to see and learn, not become a hidden system for control or individual surveillance.
Team transparency & consentThe team knows what the AI does and agrees to it.
Least-privilege accessIt can reach only the data the task actually needs.
No individual scoringInsights describe the system, never rank people.
Evidence & traceabilityEvery signal can be traced back to its source.
Human override & auditA person can stop, change, or review any action.
This is a navigation map, not a playbook. It shows how AI can be applied across the Scrum Master role, from basic assistance to self-adapting systems. Teams should adopt selectively based on context, maturity, and need, not as a fixed progression. Use the lowest autonomy level that achieves the outcome safely. The right level depends on the task’s ambiguity, reversibility, consequence, and need for human judgment.