Laveeka / Guides / Choosing a partner

The questions to ask before you sign with an AI consultancy

We wrote these for our own discovery meetings, then realised the list is more useful to the buyer than to us. Ask them of any firm you are considering, this one included.

The short answer

Ask who owns the intellectual property, what happens to the system if the firm disappears, and what the run cost looks like in year three. Most engagements that go wrong go wrong on one of those three, and all three are answerable before you sign.

Below are twelve questions, each with why it matters and what a worrying answer sounds like.

Ownership and exit

Who owns what gets built, in writing?

If the answer involves a licence, a platform fee or anything that renews, you are buying access rather than an asset. Both are legitimate, but they are very different purchases and the difference should be explicit before you sign, not discovered at renewal.

If you disappeared tomorrow, could another team run this?

Ask to see the documentation standard they hand over, not a promise that documentation will exist. A worrying answer is that everything lives in a platform only they can access.

Where does the code live during the build?

In your repository, under your account, from the first week, is the answer you want. If it lives on their machines until the end, you are relying on goodwill at exactly the point where a relationship is most likely to have soured.

The work itself

What did you measure before you proposed this?

A proposal that arrives before any measurement is a proposal about their capability rather than your business. Ask what the number was and where it came from.

Which parts of this should stay with a person, and why?

A firm that says everything can be automated has not looked closely. Judgement, relationship and risk usually should not move. A good answer names specific steps they deliberately left alone.

What happens when the system is wrong?

Every system is wrong sometimes. The question is whether that is designed for. Ask where a person approves things, what gets logged and how you would find out afterwards which decision was made on what basis.

Show me something you built that is still running two years later.

A demonstration proves a firm can build. Something still in daily use proves somebody wanted to keep using it, which is a much harder test and the only one that matters.

Money over time

What does this cost to run in year three?

Build cost is the easy number. Running cost is where the surprise lives, and it usually grows with usage. Ask for it modelled at three times today's volume.

What is the smallest version of this that would prove the idea?

A firm that cannot describe a smaller version is selling a programme rather than solving a problem. There is nearly always a smaller version.

What would make you tell me not to do this?

If nothing would, be careful. Every honest practitioner has a list of conditions under which their own service is the wrong purchase, and they should be able to say it out loud.

The uncomfortable two

Who inside my business owns this after handover?

This is a question about you, not about them, and it is the single best predictor of whether anything survives. If the answer is nobody, the honest move is to fix that first or to price a support arrangement deliberately.

What are you certified for, and what are you only working towards?

Certifications are often described loosely. Ask which are issued and which are in progress, and expect a straight answer. For our part, Laveeka's SOC 2 Type 2 and ISO/IEC 27001 programmes are underway with Oneleet and are not complete, so we do not describe ourselves as certified in either.


The one to ask yourself

Can you name, in one sentence, the decision that gets made better because this exists? If nobody in your business can, the project has no target, and no amount of good engineering will give it one.