How to judge an outside partner before they build your first AI workflow.
Use this when you are deciding whether an outside partner should build your first AI workflow. Check what is true. A partner worth hiring clears most of the green list and none of the red.
What to look for
Senior people do the actual work. Not a sales team out front and juniors behind.
A production track record they can show you. Real systems running in real operations, not just demos.
You own what they build, on your systems and open standards, with nothing you cannot export.
No new platform to buy. The AI goes into the stack you already run.
Real, measured results. Tied to a number the business already tracks.
It starts small and is priced to the return. A bounded first project, not a multi-quarter commitment.
They bring a battle-tested harness and tune it to your codebase, rules, and release process.
Red flags
They cannot tell you who actually does the work.
They lead with a platform you would rent, and lock-in you cannot leave.
They are vague about who owns the IP and the data.
They have demos but no production references.
Questions to ask in the room
Which models do you use, and how do you choose between Claude, GPT, Gemini, or a custom one?
How does our data reach the model, and where does it live? Does it ever train a model?
Which decisions are made by fixed rules we can read, versus the model?
How do people sign in, and is every action logged to a named user?
How do you validate before shipping? What is in your harness?
What do we own at the end, and what happens if we part ways?
Want a second read on this? Set up a conversation. We will map your highest-opportunity workflow with you, free.
Bob Klein, CEO · Digital Scientists [email protected] digitalscientists.com