The 10 questions
1. What comparable work can you prove?
Ask what the supplier built, its role, when it was used and how results were measured. Ask which projects were for outside clients and which were for related companies. Ask which results were measured and which are product claims. Kin-etiq.ai has one related-company example, Staffy, and says so on every page that mentions it.
2. Who will do the work?
Find out who owns the build, the security review, the testing and the delivery. Ask about subcontractors and substitutions.
3. What is in the price, and what is not?
Ask for deliverables, exclusions, assumptions, payment milestones and how changes are handled. Kin-etiq.ai quotes each custom project separately and in writing, as the pricing page explains.
4. When will we see working software?
Agree milestones that fit the scope. Ask what you will see at each review and which of your own tasks could delay delivery.
5. Where does our data go?
Ask for a list of every provider, where processing happens, how long data is kept, whether it is used for training, who can access it and how it is deleted. A security credential on its own does not show how one system handles your data.
6. What will it cost to run?
Separate hosting, charges from the AI provider, monitoring, maintenance and support. Ask for the usage assumptions behind each number.
7. How will we know it works?
Agree on real test cases, how many errors are acceptable, when the system hands off to a person, and what counts as done. Decide who can pause or undo the system if something goes wrong.
8. When would you recommend something else?
A good supplier compares a build against changing settings, changing the process and buying a product, and tells you when not to build. Our guide on build vs buy for AI agents sets out that comparison.
9. What will we own?
Get it in writing: who owns the code, who can get at it, how you take your data with you, and what you receive at handover. List outside parts and any continuing fees.
10. Who is responsible after launch?
Agree support hours, response targets, who owns incidents, and cancellation terms. Ask what happens if the original developer is not available.
How do you make proposals comparable?
Give every shortlisted supplier the same workflow, the same systems to connect, the same workload numbers and the same acceptance test. Ask each one to name what is still uncertain. Compare the full delivery and running commitment before you compare headline prices.
Where do AI tools look when you ask them for an agency?
When we asked ChatGPT and Claude "Best AI agency Toronto?" on 20 October 2026, both built their answers from Clutch's Toronto rankings. Treat a ranking as a starting list. The ten questions above are how you get from that list to a choice.