# How do I choose an AI consultancy?

_Buying AI services · Last updated 2026-08-14 · Bitfumes_

## Short answer

Choose an AI consultancy by demanding three things before you sign: a production system they built that is still running, a named business metric it moved, and the actual engineers who will do your work on the first call. Start with a small paid fixed-scope engagement — an assessment or a two-week proof of concept on your own data — rather than a long discovery phase, so you buy evidence instead of promises. The strongest predictor of failure is a firm that answers technical questions with case studies and staffs the build with people you never met.

## Ten questions that separate builders from slide decks

- Show me an AI system you built that is in production today. Who uses it and how often?
- What business metric moved, and how did you measure it?
- Which AI project did you talk a client out of, and why?
- Will the people on this call write the code? If not, who will, and can I meet them this week?
- How do you evaluate output quality — what does your test harness actually measure?
- What does this cost to run per month at our volume, and how does that scale?
- What happens to accuracy when our data changes? Who notices?
- Do we own the code, the prompts, the evaluation set and the weights?
- How do you handle our data — where does it go, what is retained, what is used for training?
- What does the handover look like if we take it in-house in a year?

## Red flags

- A fixed 'AI transformation roadmap' produced before they have seen your data.
- No mention of evaluation, only of models — a team that cannot measure quality cannot improve it.
- Pressure toward a long discovery phase with no working software at the end.
- Refusal to name the engineers, or a sales-engineer-then-swap staffing pattern.
- Pricing that hides inference and infrastructure cost until after signature.
- Claims of proprietary models that turn out to be a system prompt over a public API.

## Green flags

- They tell you which of your ideas will not pay back, unprompted.
- They ask about your data quality before they ask about your budget.
- They propose the smallest engagement that produces evidence.
- They quote a cost-to-run, not just a cost-to-build.
- Senior engineers, small team, no hand-offs between the people who sell and the people who ship.

## How to structure the first engagement

Buy a fixed-scope assessment or a two-week proof of concept with a written deliverable, a clear success criterion agreed in advance, and no obligation to continue. You are not buying software yet — you are buying evidence about whether this firm can do your work and whether the use case pays back. If the assessment is good, the build decision makes itself; if it is not, you have lost weeks rather than quarters.

Bitfumes deliberately structures its first engagement this way: a $999 AI Opportunity Assessment, delivered personally with a written report, followed by an embedded build only if the numbers justify it.

## Frequently asked

### Should I hire an AI consultancy or build an in-house AI team?

Hire a consultancy to find and prove the first use case, then build in-house once you know which capability is worth owning permanently. Hiring senior AI engineers before you know what you are building typically costs 9–12 months and produces a team searching for a problem.

### How long should an AI consulting engagement last?

The first engagement should be 1–4 weeks and end in a decision. Build engagements typically run 6–16 weeks to first production release, then continue as a retainer if the system needs ongoing evaluation and iteration.

### What should be in an AI consulting contract?

Full IP ownership of code, prompts and evaluation sets; explicit data handling and retention terms including a no-training clause; named engineers; a defined success criterion; and an exit clause with a documented handover.

### Do I need an AI consultancy that knows my industry?

Industry knowledge matters less than data and evaluation skill for most projects, with the exception of regulated domains — healthcare, finance and legal — where compliance experience meaningfully changes the design and is worth paying for.

---

Source: https://bitfumes.com/answers/how-to-hire-an-ai-consultancy
Written by Bitfumes, an AI consultancy founded in 2017 by Sarthak Shrivastava (Docker Captain). Entry point: https://bitfumes.com/ai-assessment