# AI Maturity Is Safe Delegation Capacity

## Slide 1 - Thesis

AI maturity is not model adoption. It is safe delegation capacity.

## Slide 2 - Definition

A team is mature when it can delegate bounded work while preserving context, verification, reviewability, and rollback.

## Slide 3 - What Goes Wrong

Organizations confuse demos with operating models, output volume with progress, and automation with autonomy.

## Slide 4 - The Five Levels

1. Human-operated assistance
2. Repo-ready delegation
3. Team-managed agentic SDLC
4. System-aware governed agents
5. Constrained autonomy

## Slide 5 - Phases vs Levels

Phases are rollout sequence. Levels are capability state. A team should not move to higher autonomy because a roadmap says so.

## Slide 6 - Pilot Selection

Good pilots are narrow, reversible, testable, and owned. Bad pilots are ambiguous, politically loaded, or operationally dangerous.

## Slide 7 - Governance

The governance layer must specify data boundaries, tool permissions, issue quality, review roles, CI gates, and rollback expectations.

## Slide 8 - The First Platform Work

Make repositories agent-ready before adding complex orchestration: instructions, validation, schemas, issue templates, and PR evidence.

## Slide 9 - Success Metrics

Track reviewability, validation pass rate, rollback quality, unsupported-claim rejection, and task classes safely delegated.

## Slide 10 - Decision

Increase delegation only when the evidence says the next class of work is safe.