What is AI-native development?

A way of working where AI agents write a large share of the code and the team has rules, checks and accountability set up so that it is safe. It is not a tool. It is rules in the repository, gates in CI and a clear answer to who is accountable for what.

What are gates and why are rules not enough?

A gate is a check that stops a change: CI, a test, a hook or a setting. An agent can work around a rule in text, not around a gate. That is why items 3, 5 and 6 ask about enforcement, not whether the rule is written down somewhere.

Which items should we fix first?

Items 5, 9 and 10: CI blocks the merge, the PR author answers for AI output and agents have no production access. While any of them is missing, bigger tasks for agents make no sense, whatever the score.

Can the team fix the missing items on its own?

Mostly, yes. A missing item is a work list: the three mandatory items first, then rules into the repository and gates into CI. After that, add new rules from review when the same mistake happens twice.

A test for your dev team

How far are you from AI-native SDLC? 12 questions

An AI-native dev team has shared rules for AI agents in the repository, CI gates that stop a bad change, and a clear answer to who is accountable for AI output. Answer yes or no to 12 questions in three areas and see where your team is today. Items 5, 9 and 10 are mandatory either way.

What you fill in and tick stays in this browser window only. Nothing is sent unless you ask for your result by email. It is gone when you close the page, so print your result.

For CTOs and tech leads whose team already uses Claude Code, Codex, Cursor or Copilot. Answer for the whole team, not for yourself. Where you are not sure, leave it unticked.

Today AI writes all of our code and a senior answers for the result. Since 2023 we have delivered 15+ production projects. We ask these questions with every team we start with.

A. Rules

1. The team has shared rules for AI agents. Not every developer in their own way.
2. The rules live in the repository, in CLAUDE.md or AGENTS.md. Check: you can find them without asking anyone.
3. A setting or a hook enforces the important rules. Not just a sentence the agent can work around.
4. A new rule comes from a mistake that happened twice. Check: you can tie the latest rule to a specific review comment.

B. Checks

5. CI checks run on AI code too and block the merge. Not just report.
6. The main flows have tests that really fail when something breaks. Check: break one line and see whether a test fails.
7. You have rules for reviewing AI changes. What the reviewer must look at and how big a change can still pass.
8. A new developer learns the rules from the repository. Not by word of mouth in their first meeting.

C. Accountability and security

9. It is clear who answers for AI output. The PR author, not the agent.
10. Agents have no access to production or production data. Check: try whether an agent in the dev environment can reach the production database.
11. You have a list of approved AI tools and what must never go into them. Secrets, personal data, client code.
12. You can answer management's question "what is our plan with AI". In one sentence, with what you measure.

Compare your team

Got result links from colleagues? Paste them here, one per line, up to 12.

Result links

The answers sit in the link after the # sign, with no names or emails. Browsers never send that part of a link to the server and we do not store it.

Scoring

Count the ticked items. But look at items 5, 9 and 10 first. If even one of them is missing, fix it first, whatever the score.

Ticked	Verdict	What next
0 to 4	Getting started	You use AI, everyone in their own way. AI runs in the company, but nobody steers it. Start with items 1, 5 and 10.
5 to 8	Halfway there	You have rules, not gates. Add the missing checks before you give agents bigger tasks.
9 to 12	Almost there	You are close to an AI-native SDLC. You can give agents bigger tasks. Keep adding rules from review.

A missing item is not a team failure. It is a work list, and you can do most of it yourselves.

More tools for you

How many of 10 gates stop bad AI code in your CI?

Tick the gates you have and see which are missing. Each one with a GitHub Actions snippet.

AI in the SDLC in 30 days: a week-by-week plan

Rules in AGENTS.md, CI gates and hooks, review and measurement. Four weeks with snippets from our public repository.

Frequently asked questions

What is AI-native development?: A way of working where AI agents write a large share of the code and the team has rules, checks and accountability set up so that it is safe. It is not a tool. It is rules in the repository, gates in CI and a clear answer to who is accountable for what.
What are gates and why are rules not enough?: A gate is a check that stops a change: CI, a test, a hook or a setting. An agent can work around a rule in text, not around a gate. That is why items 3, 5 and 6 ask about enforcement, not whether the rule is written down somewhere.
Which items should we fix first?: Items 5, 9 and 10: CI blocks the merge, the PR author answers for AI output and agents have no production access. While any of them is missing, bigger tasks for agents make no sense, whatever the score.
Can the team fix the missing items on its own?: Mostly, yes. A missing item is a work list: the three mandatory items first, then rules into the repository and gates into CI. After that, add new rules from review when the same mistake happens twice.

Where to next

AI-native SDLC

Your team has AI, but no rules or checks

Want to know what to fix first?

Tell us which items you are missing. We get back to you within two business days with a few questions about your work. Then we agree on what we do first.

Tell us what your team keeps postponing Or a 15-minute intro call

Who we help

Services

Resources

What is an AI agent? Plain answer, business examples

When is Excel no longer enough for a business?