AI Consulting Firms: How to Choose
Four kinds of firm will pitch you, and they are good at genuinely different things. We are one of them, so read this knowing that, and note that we describe the cases where we are the wrong choice.
Global consultancies
Accenture, Deloitte, McKinsey, EY, BCG
Strong at: Board-level credibility, genuine change-management capability, and the capacity to run a programme across many business units at once.
Weaker at: Expensive, and the people who sold the work are often not the people who do it. Strategy and delivery are frequently separate teams with a handover between them.
Platform partners
Cloud and vendor implementation partners
Strong at: Deep expertise in one stack and the fastest route to a working deployment on it, often with vendor funding attached.
Weaker at: The recommendation is constrained by the platform. A partner earning margin on a vendor is not a neutral party on whether that vendor is right.
Product companies
AI SaaS vendors with services attached
Strong at: Fastest to value where their product fits your problem closely. Little bespoke work needed.
Weaker at: The answer is always their product. Where the fit is partial, you adapt your process to the software.
Specialist engineering firms
Senteras and firms like us
Strong at: Build what fits the constraints, including on your own infrastructure. Vendor-neutral because there is no vendor relationship to protect. Same people from scoping to delivery.
Weaker at: Smaller. We cannot staff a fifty-person programme, and we do not have a global brand for a board that needs one.
The questions that actually separate them
Case studies all look similar and reference calls are curated. These four questions are not, and the answers tend to be revealing:
- Who owns the output? Code, model weights, fine-tuned artefacts, evaluation sets. If the answer is not you, understand exactly what happens when the relationship ends.
- Do you take vendor commissions? Not disqualifying, but you should know before weighing a technology recommendation.
- Who does the work? Whether the people in the room are the people who will build it, and if not, when the handover happens.
- What would you refuse to build? A firm with no answer either has not thought about it or will build whatever is asked, and both are a problem.
Where we fit
We are a specialist engineering firm. We are the right choice when the constraint is technical or regulatory: data that cannot leave your perimeter, an accreditation boundary, a cost curve that has stopped working, or a use case that needs building rather than buying.
We are the wrong choice for a multi-year enterprise transformation programme across a large organisation, and for a board that needs a global brand on the cover. Those are real requirements, and a large consultancy serves them better than we would.
Related
Common questions
When should we hire a large consultancy instead?
When the problem is organisational rather than technical: a multi-year transformation across many business units, or a board that needs an established name attached to the recommendation. Those are real reasons and we will say so.
When is a platform partner the right choice?
When you have already committed to a platform and want the fastest path to using it well. The trade is that a partner paid by a vendor is not the right party to ask whether that vendor is the right choice.
Should we just hire in-house?
If AI is going to be core to your product, yes, eventually. Senior ML engineers are expensive and hard to hire, and a first hire with no existing team around them often struggles. A common sensible pattern is to use a firm for the first deployment and hire against a system that already works.
How do we evaluate any AI consultancy?
Ask what they would build, then ask what they would refuse to build. Ask who owns the code and the model weights at the end. Ask whether they take commissions from any vendor they might recommend. The answers are more revealing than a case study deck.
Find out what this looks like for your organisation
A 30-minute call. We will tell you plainly whether AI is the right tool for the problem you have, including when it is not.
Book a free consultation