A lightweight self-assessment for teams moving from random AI usage to disciplined AI-assisted delivery.
Many teams are already using AI coding assistants. That does not automatically mean they have mature AI-assisted engineering practices.
This maturity map helps teams ask a better question:
Are we using AI as a personal productivity trick, or have we built enough engineering discipline around it?
It is designed for engineering managers, tech leads, architects, CTOs, and software teams who want a practical team-level view of AI-assisted engineering maturity.
This is not a certification, audit, or individual productivity score. It is a conversation tool for improving the engineering system around AI-assisted work.
- Open assessment-scorecard.md.
- Gather recent evidence from PRs, tickets, review comments, docs, tests, incidents, and team discussion.
- Score each dimension from 0 to 5.
- Calculate the average score.
- Identify the lowest two dimensions.
- Pick one improvement action for the next 30 days.
Use the spreadsheet version if that is easier: spreadsheet/maturity-map-scorecard.csv.
| Level | Name | Meaning |
|---|---|---|
| Level 0 | No Practice | AI use is absent, blocked, unclear, or unmanaged. |
| Level 1 | Individual Experimentation | Developers use AI personally, but practices vary widely. |
| Level 2 | Team Awareness | The team discusses AI usage and basic review expectations. |
| Level 3 | Context-Aware Usage | Repos, tickets, docs, tests, and PRs start carrying better context. |
| Level 4 | Review-Governed Usage | AI-assisted changes use passports, review checks, boundaries, and evidence. |
| Level 5 | Learning Engineering System | Review findings, incidents, and AI mistakes improve docs, tests, ADRs, prompts, and workflow. |
Read the full level guide in maturity-map.md.
| Dimension | Core question |
|---|---|
| AI Usage Clarity | Does the team know where AI is encouraged, optional, risky, or restricted? |
| Context Readiness | Can AI and humans find requirements, rules, docs, ADRs, and examples? |
| Repository Readiness | Does the repo contain README, assistant guidance, architecture notes, tests, and standards? |
| Review Discipline | Are AI-assisted outputs reviewed for assumptions, not just syntax? |
| Business Rule Visibility | Are important rules captured in tickets, tests, docs, or examples? |
| Architecture Boundaries | Are module, service, data, and API boundaries visible and reviewable? |
| Evidence & Ownership | Do important changes preserve intent, validation, rollout, and owner? |
| Team Learning Loop | Do repeated AI mistakes improve the engineering system? |
Use scoring-guide.md for score-by-score criteria.
Team AI-Assisted Engineering Maturity
Overall Score: 2.8 / 5
Overall Level: Level 3 - Context-Aware Usage
Strongest Areas:
- Individual AI usage
- Repository readiness
- Basic team awareness
Weakest Areas:
- Business rule visibility
- Evidence and ownership
- Architecture boundaries
Recommended Next Actions:
1. Add AI Change Passport to PRs for high-risk changes.
2. Create AGENTS.md with repository guidance and boundaries.
3. Convert 3 repeated review comments into repo guidance, tests, or examples.
- Level 1: Individual Experimentation
- Level 3: Context-Aware Team
- Level 5: Learning Engineering System
| Weak dimension | Suggested next artifact |
|---|---|
| AI Usage Clarity | AI Use Case Fit Matrix / Agent Permission Matrix |
| Context Readiness | AI-Assisted Engineering Starter Kit |
| Repository Readiness | AI-Assisted Engineering Starter Kit / AGENTS.md template |
| Review Discipline | AI Change Passport |
| Business Rule Visibility | Business Rule Capture Card |
| Architecture Boundaries | Architecture Drift Radar |
| Evidence & Ownership | AI Change Passport / Evidence Chain Template |
| Team Learning Loop | AI Delivery Health Radar / Prompt-to-Process Converter |
| Need Automation | Software Signal Gate |
No. Score team practices, not people.
No. A team can use AI heavily and still have weak context, review, evidence, ownership, and learning loops.
No. The score is a conversation starter. The weakest dimensions usually matter more than the exact average.
Every 4 to 6 weeks is usually enough for a team-level improvement loop.
AI-Assisted Engineering Maturity Map is a lightweight self-assessment for teams moving from random AI usage to disciplined AI-assisted delivery.
Find your current level, identify weak areas, and choose the next practical artifact to improve your practice.