What questions should I ask an AI automation agency before signing a contract?
Before signing with an AI automation agency, ask who will do the work, what will be true when it is done, where your data goes, who owns what gets built, what drives the price, and what happens in month six. The best agencies have specific, unhesitating answers to all of these. Vague answers on ownership, data handling or the definition of done are the clearest signal to keep looking.
By Cade Dannels, Founder, Daita Dynamics. Updated .
How to use this list
Ask these in the second conversation, once the agency has seen your workflows. Under each question is the answer you should hear. You are not looking for perfection; you are looking for specificity and for an agency willing to say no.
Scope and outcome
1. What exactly will be true when this project is finished?
You should hear a change to work your team does today, in one sentence. "Your controller closes client books in an afternoon instead of three days." Not "you will have an AI foundation."
2. Which of the things we asked for would you advise us not to build?
A good agency has an opinion. If everything you suggested is a great idea, the agency is agreeing with you rather than advising you.
3. Have you built something comparable, and can we see it running?
Look for a walk-through of a working system at a business roughly your size, not a slide. Ask what broke during that build and how they fixed it.
People
4. Who will scope the work, and will that person also build it?
The right answer names a person and says yes. If scoping and building are done by different people, ask how the handoff works and how much is typically lost.
5. Who will we talk to when something breaks?
You want a named, senior person, not a ticket queue.
Data and security
6. Where does our data go, and who can see it?
The agency should be able to describe, in plain language, which systems your data passes through, whether any of it is sent to third-party AI providers, what those providers do with it, and how access is controlled. If they cannot answer without a follow-up, they have not thought about it.
7. What happens to our data when the engagement ends?
Clear deletion or return terms, in writing.
8. How do you handle errors and edge cases in the automation?
Automations fail. The good ones fail loudly, route exceptions to a person, and log what happened. An agency that has not built error handling into its plan has not run automations in production.
Ownership and handover
9. Do we own everything you build, including code, prompts, and configurations?
Yes should be the whole answer. Anything else means you are renting your own systems.
10. If we stopped working with you tomorrow, could our team or another firm keep the systems running?
Look for documentation, access to every account in your name, and a training plan for an internal owner.
11. Who in our company will you train, and how?
A named role, hands-on sessions, and written runbooks. An agency that keeps you dependent is not on your side.
Pricing
12. What drives your price, and what would make it go up?
Scope, the state of your data, the number of systems to integrate, whether anything custom is needed, compliance requirements, and ongoing operation are reasonable drivers. A quote that arrives before anyone has looked at your workflows is a guess.
13. What is included after launch, and for how long?
A defined support window, what counts as a fix versus new work, and what ongoing operation costs if you want them to run it.
Judgement
14. What would you tell us not to automate?
The most revealing question on the list. An agency that has done this work will have examples of processes that looked automatable and were not, and will tell you when a simpler tool or a process change would do the job.
Red flags in the answers
- Every answer leads back to a single platform the agency sells or is certified in.
- Ownership of code or prompts is "shared" or stays with the agency.
- Nobody will commit to being on your project from scoping through handover.
- Data handling answers require a follow-up email.
- "Done" is described as a capability rather than a change to a task.
Where Daita Dynamics fits
Daita Dynamics is a Denver-based AI strategy and implementation firm founded in 2024, working remotely with growing B2B businesses across the United States. We expect to be asked all fourteen of these questions, and our answers are consistent: the senior people who scope the work build it, you own everything, and we train someone on your team to run it. The free AI Opportunity Assessment is where most of these conversations start.
Frequently asked questions
- Should the contract include a fixed price?
For a defined build, yes, or at least a fixed price per phase. For ongoing operation, a monthly arrangement is normal. Avoid open-ended hourly work with no deliverable attached.
- What should a good proposal contain?
The outcome in one sentence, the systems involved, what you must provide, the sequence and timeline, who does the work, what you own, the price and its drivers, and the support terms.
- Is it reasonable to ask for a paid pilot first?
Yes, if it has a defined outcome and a short timeline. One real task automated in a few weeks tells you more about an agency than any reference call.
- How many agencies should we talk to?
Two or three is enough. More than that and you are comparing proposals rather than choosing people, and it is the people who determine whether the systems work.