First-step checklist
Find your first safe AI step
The company-wide AI plan will take time. Your first step does not have to. Pick one journey you have in mind, then answer twelve questions.
Twelve questions, one journey
Check the ones you can answer yes to with confidence. If you are not sure, leave it unchecked.
Value: will it matter?
- Can you name one user journey where a faster or smarter flow would be noticed by customers or staff within a quarter?
- Can you describe success in one number, such as time saved, tickets avoided, conversion, or app rating?
- Is there one business owner who wants this outcome and can decide quickly?
Scope: can it stay small?
- Could a small team of two to four people deliver a first version in weeks, not quarters?
- Can it run on the APIs and data you already have, without a new platform first?
- Can you limit it to a small set of actions, with a clear way to switch it off?
Control: is it safe to try?
- Do you know which personal data it touches, and can you keep that to a minimum?
- Can risky actions wait for a human approval?
- Can every action be logged, so you can show who did what and why?
Team: will the learning stay with you?
- Will your own engineers work on it, so the know-how stays in the company?
- Are there review rules and tests that AI-written code must pass before release?
- Is there a decision date to build further, change course, or stop?
Whatever your score
- Should assistants or agents act inside your app? See how the Agent Gateway adds consent, scoped permissions, approval flows, and audit logs.
- Already have AI-built code? Take the 12-question readiness check for that code first.
- No review rules or tests for AI-written code yet? That becomes part of the project, not a blocker.
What a good first step looks like
We call it a Lighthouse Project: one valuable journey, a small set of safe actions, and existing APIs where possible. A control layer adds consent, scoped permissions, approval flows, rate limits, and audit logs.
It can be an internal MVP, a technical proof point, or a customer-facing pilot. Either way you get controlled risk, a small blast radius, and a reviewable first step.
We build it with AI as a force multiplier for senior engineers, not a replacement. Every change goes through manual code review, automated tests, and security and privacy checks.
Frequently asked questions
What is a good first AI project?
One user journey that matters, a small set of actions, and data you already have. It should show a real result within weeks, with a human approval step for anything risky and a log of every action.
How small should a first AI project be?
Small enough for two to four people to deliver a first version in weeks. Big enough that customers or staff notice the difference. If it needs a new platform first, it is too big for a first step.
Do we need a company-wide AI strategy first?
No. The company-wide plan will take time, for good reasons. A focused first project with controlled risk gives that plan real evidence instead of assumptions.