Your AI Readiness Action Plan

This page accompanies the AI Readiness Assessment. Find your level below for a focused 30-day plan. Each plan is opinionated, free, and assumes no budget. If you want help executing, book a call.

Level 1 — Starting Line (0–20%)

You’re early. Treat the next 30 days as a controlled experiment, not a transformation.

Week 1 — Pick one tool, one team.

• Choose Cursor or GitHub Copilot. Don’t try both.
• Pilot with 3–5 engineers, not the whole team. Volunteers, not draftees.
• Agree what “success” looks like: an honest weekly retro, not a metric.

Week 2 — Write down the rules.

• A one-page acceptable-use note: what code can the AI see, what it cannot, where the data goes. Tape it to the wall (figuratively).
• Decide who reviews AI-generated code. Same standard as human code — nothing less.

Week 3 — Make the team’s standards legible to the AI.

• Create a single AGENTS.md (or CLAUDE.md) in your most-touched repository. List: coding style, test conventions, what “done” means, and 3–5 example patterns you actually use.
• This single file is the highest-leverage 2-hour investment you’ll make this year.

Week 4 — Honest retro.

• Are people faster? More tired? Shipping more bugs? Ask, write it down, share with the team.
• Decide: expand to the next 3 engineers, or pause and fix the rough edges first.

30-day check-in: Can a new engineer use AI in their first week without asking what’s allowed? If no, the policy isn’t real yet.

Level 2 — AI Experimenting (21–40%)

You have AI usage. You don’t have AI practice. The gap is standardisation.

Week 1 — Measure what’s actually happening.

• Survey the team: who uses what, how often, for what kinds of tasks.
• Pull tool-side metrics if you have them (Copilot/Cursor admin dashboards).
• You’re not optimising yet — you’re getting a baseline.

Week 2 — Pick one workflow to systematise.

• Three good candidates: AI-assisted PR reviews, AI-generated unit tests, or AI-written commit messages. Pick the one with the most pain today.
• Write the standard: when to use it, what to check, what to never accept.

Week 3 — Make context portable.

• Roll out AGENTS.md to your top 3–5 repos. Same conventions across all of them.
• Add the standard from Week 2 to those files so the AI knows your team’s expectation.

Week 4 — Name an owner.

• One person (not “everyone”) owns AI practice for the team. 2–4 hours/week.
• They run a monthly internal show-and-tell: what’s working, what’s not, what to try next.

30-day check-in: If your best AI user left tomorrow, would the team’s productivity drop? If yes, your practice is still personal, not team-level.

Level 3 — AI Adopting (41–60%)

You’ve crossed the personal-use threshold. The next gains come from workflow integration and governance — not better tool choices.

Week 1 — Audit your CI feedback loop.

• Time how long it takes from PR open to test results. Anything over 15 minutes is killing AI-agent productivity.
• Pick the slowest 2 jobs. Parallelise, cache, or kill them.

Week 2 — Bring AI into the merge pipeline.

• Add an AI PR-review step (Coderabbit, GitHub Copilot autoreview, or a custom one). It comments alongside humans — it doesn’t replace them.
• Set the expectation: humans review the comments, not the code.

Week 3 — Write the review standard.

• A 1-page “How we evaluate AI-generated code” doc. What gets a fast-track. What gets extra scrutiny. What never ships.
• Add it to your contributing guide. Share it with new hires day one.

Week 4 — Start measuring outcomes.

• Pick 2 metrics that matter to leadership: deployment frequency, change-failure rate, or PR cycle time.
• Establish the baseline this week. Don’t try to move it yet.

30-day check-in: Can you show your CEO one chart that proves AI is making the team better? If no, you have adoption without proof — and adoption without proof gets cut in the next budget review.

Level 4 — AI Integrating (61–80%)

The fundamentals are working. The next moves are last-mile: evals, agents, cross-team consistency.

Week 1 — Build your first eval suite.

• Pick one AI-assisted workflow (e.g., automated test generation). Define 10 cases where you know the right answer.
• Run the workflow on those cases weekly. Watch the trend, not the snapshot.

Week 2 — Audit cross-team variance.

• Compare AI adoption, tool choice, and AGENTS.md quality across your sub-teams.
• The gap between your strongest and weakest team is your real opportunity — not pushing the leaders further.

Week 3 — Pilot one agentic workflow.

• Pick something with clear input/output: bug triage, dependency upgrades, or doc generation.
• Run it with a human in the loop for two weeks. Document where it fails.

Week 4 — Codify what’s working.

• Promote your best team’s practices to engineering-wide standards.
• Schedule a quarterly review — what’s still working, what’s now table stakes, what to retire.

30-day check-in: Are your weakest sub-teams now at the level your strongest team was 6 months ago? If yes, compounding is real. If no, you have champions, not a culture.

Level 5 — AI Native (81–100%)

You’re at the frontier. The risk now isn’t falling behind — it’s getting comfortable. The model and tool landscape will shift again in 90 days.

This quarter — Watch these four things:

1. Agentic workflows. Specifically: long-running agents with their own evals, memory, and tool access. The teams that nail this in 2026 will pull away from the rest.
2. Cost-per-task. As model usage scales, tooling spend becomes a P&L line. Have a clear unit-economics view of AI cost per engineer, per feature, per customer.
3. Compliance posture. DPDPA, EU AI Act, client MSA terms — the regulatory surface is growing. A 1-pager mapping your AI usage to obligations is now a board-level artifact.
4. Talent strategy. Your hiring bar, onboarding, and career ladder all need an update. “Senior engineer in 2026” means something different from 2024.

30-day check-in: Could your top competitor copy your AI practice in 6 months by hiring 2 of your engineers? If yes, your moat is people, not practice — and you need to invest in practice that survives turnover.

What next?

If any of these plans landed and you want help executing — book a free 30-min call. No pitch, no slides. We’ll talk about what’s working and what isn’t.

If you haven’t taken the assessment yet, start here.