What should you ask before hiring an AI automation company?
Eight questions: who writes the code, whether the price is fixed, who owns the repository on day one, what happens when it breaks after handover, whether you can see something running, which numbers were measured rather than designed for, what would make them turn the work down, and where your data goes.
5 min read
Ask these of us too, and compare the answers. Each one takes a minute to answer honestly and is difficult to fake under a follow-up question.
- Who writes the code — the person on this call? If the answer involves a team you will meet later, ask when, and ask who writes the scope.
- Is the price fixed before work starts, and what happens when scope changes? Time and materials means the estimation risk is yours. A fixed price means it is theirs.
- Who owns the repository on day one? Not at the end, not on final payment. Day one. If the answer is anything else, you are renting.
- What happens if it breaks six months after handover? If the only answer is a retainer, the handover was never real.
- Can I see something running? Not a deck, not a recorded demo. A live URL belonging to a named client who agreed to be named.
- What do you measure, and was it measured or designed for? Every studio shows numbers. Ask which came out of a client's system and which are targets. Both are legitimate to publish; conflating them is not.
- What would make you turn this project down? A studio that has never said no either has no standards or has not told you about the projects that went badly.
- Where does my data go, and who can see it? Which provider, which region, what gets logged, and what happens to it when the engagement ends.
If an answer is long, vague, or arrives later as a follow-up email, that is your answer.
Why these eight
They are the questions whose answers are expensive to change later. Ownership, pricing model and data residency are all settled before a line is written, and all three are painful to renegotiate once a system is live and you depend on it.
The two about proof — seeing something running, and separating measured numbers from designed targets — are there because this is where marketing material is weakest. A studio with real work answers both in a sentence. One without produces a case study with no client name attached to it.
Related questions
- Should an AI automation company give a fixed price?
- Yes, for a defined scope. A fixed price means the studio carries the estimation risk rather than you. Ask specifically what happens when scope changes mid-build: the honest answer is that it is requoted before the work happens, not invoiced after.
- Who should own the code for a custom AI build?
- You, from day one — the repository, the infrastructure and the data. If ownership transfers only at the end or on final payment, you are renting the system you paid to have built, and the leverage sits with whoever holds the repository.
- How can I tell whether a case study is real?
- Ask for a live URL and a named client who agreed to be named. Then ask which figures were measured in that client's system and which are targets the build was designed against. Both are fine to publish; presenting the second as the first is not.
Still deciding?
A 30-minute call, then your scope and your price within one business day. If we are not the right people for it, we will say so.
Book a free call