Quantum Beetle

GuidesChoosing a partner

Fifteen questions to ask before hiring an AI company

Questions about data, method, ownership, people and proof that separate an AI company that will deliver from one that will demonstrate. With the answers to listen for.

Choosing a partner · 3 min read · 10 October 2026

Every AI company has a good demonstration. Demonstrations are built to succeed. What you need to know is how the company behaves with your data, your constraints and the awkward cases, and that comes out in the answers to plain questions.

Ask these of anyone you are considering, and listen for specifics.

01About your data

  1. Where exactly will our data go? Draw it.
  2. Is any of it used to train models that other customers use?
  3. Can the system run entirely on our own infrastructure?
  4. What happens to our data when the contract ends?

Listen for a diagram and a straight answer. "It is all secure" is not an answer.

02About the method

  1. Which decisions are made by rules, which by search and which by a model?
  2. What does the system do when it is unsure?
  3. Can you show one decision and exactly why it was made?
  4. How will we measure whether it worked, and what is the number today?

03About ownership

  1. Who owns the code, the configuration and the trained components?
  2. If you stopped trading, would our system keep running?
  3. What would our own team need to know to run this without you?

04About people and proof

  1. Who will actually do the work, and can we meet them?
  2. What is the smallest first project you would recommend?
  3. What would make you say this is not a good fit for you?
  4. What will this cost, in writing, before any work starts?

05Answers that should worry you

  • A promise of accuracy before they have seen your data
  • No interest in how your process really works
  • Everything solved by one large model
  • A long contract before a small first result
  • Reluctance to say what they are not good at

06Answers that should reassure you

  • Questions about your data that you find hard to answer
  • A first project small enough to judge
  • A clear statement of what you will own
  • A plan for handing over
  • A price and a scope in writing

In short

  • Ask where your data goes, how decisions are made, what you will own and who does the work.
  • A good company asks hard questions back and proposes a small first project.
  • Be wary of accuracy promises made before anyone has seen your data.
  • Get the scope and the price in writing before work starts.

Questions

How many companies should we talk to?

Three is usually enough to see the differences. More than five and the comparison becomes its own project.

Should we ask for references?

Yes, while knowing that much enterprise work is confidential and cannot be named. Where references are restricted, ask for a small paid first stage that you can judge yourself.

Is the cheapest proposal a warning sign?

Not by itself. Check whether it covers the same scope. A low price usually means something has been left out.

How would Quantum Beetle answer these?

Deployments run on the client's own infrastructure with local models by default; systems are built rules first, meaning next, AI reasoning last, with each decision explained; the client owns what is built; and the scope and price are agreed before work starts.

Sounds like your problem?

Tell us about it. We'll say honestly whether the swarm can help, and what it would take.

Talk to ushello@quantumbeetle.ai

Read next