The short answer
Before hiring an AI agency, ask twelve things: what problem they would solve first, who does the work, where AI is used, what examples they can show, how success is measured, what data they need, who owns the output, what it costs and what is included, how long it takes, what they do not do, what happens when something breaks, and how you can leave. Clear, specific answers matter more than polished ones.
- Ask about scope, people, tools, data, ownership, cost, timing and exit.
- Specific answers beat confident general ones.
- Ask for a past client you can speak with.
- Ask what they will not do.
Questions about the work
1. What would you tackle first, and why? 2. Who will actually do the work, and who is my contact? 3. Where do you use AI and where do people step in? 4. Can you show working examples and put me in touch with a past client? The best agencies answer these without hesitation and name real details. Listen for whether the person pitching can describe a real project step by step, including what went wrong and how it was fixed. Real experience shows up in the unglamorous detail.
Questions about results
5. How will we know it is working? 6. What would make you recommend stopping? An agency that can describe failure honestly is more believable than one that only talks about wins. Agree the measures in writing before the work starts. Measures should connect to something you care about, such as enquiries, hours saved or errors avoided. Activity numbers like posts published or tasks automated only tell you what was done, not what it achieved.
Questions about your data
7. What data do you need and where will it be stored? 8. Which AI providers will see my information? The OAIC’s guidance on commercially available AI products is a good reference point for personal information, and the National AI Centre’s guidance for AI adoption covers governance for businesses. If the answer includes a long list of tools you have never heard of, ask for a one-page summary of what data goes where, so you can review it with anyone responsible for privacy in your business. Sources: OAIC: privacy and commercially available AI products, National AI Centre: Guidance for AI adoption
Questions about ownership
9. Who owns the code, content and accounts when we finish? You want your name on the accounts and a clear right to the work. Ask for documentation so another provider can pick it up if needed. Put this in the contract as a clear sentence, not a verbal promise. Ask also whether any templates or tools the agency owns are licensed to you, and for how long after the engagement ends.
Questions about money and time
10. What does this cost, what is included, and what could add to it? 11. How long will the first version take? Ask for a written scope. Check the official pricing pages for any software subscriptions that sit outside the agency’s fee. Ask what happens if the project takes longer than planned, and who bears that cost. Fixed-price work should say what counts as a change in scope, so the boundary is clear before anyone is frustrated.
Questions about risk and exit
12. What happens if it breaks, and how do I leave? Support terms, response times and notice periods should be clear. As a bonus question, ask what they do not do. A confident agency will tell you which services and channels it leaves to others. Test the answers by asking for the support terms in writing. If the agency is reluctant to put response times or notice periods on paper, that tells you how the relationship may feel under pressure.
Questions people ask
Which question matters most?
Ownership and exit. Most disputes come from unclear terms, not poor work.
Should I ask for references?
Yes. One call with a past client tells you more than a case study.
How many agencies should I interview?
Two or three. More than that adds noise.
tools.ceo is published by Kamand, which is mentioned on this page. No affiliate links on this page at publication. Product names link to official sites so you can check current pricing. How we work.
