Seven questions that tell you more than any sales conversation will, and what a good answer to each actually sounds like. They work on anyone you are considering. We have answered them about ourselves at the end.
Choosing an AI consultant in Scotland is harder than it ought to be. The work is new enough that there is no settled idea of what good looks like, and the language comes out much the same whoever is speaking. So what helps is a list of questions, and a sense of what a real answer sounds like next to a comfortable one.
These work on any supplier, including us. A good competitor would pass them — a guide only we could pass would not be worth your time.
Before you get to the interview stage, it helps to know roughly where the hours in your own week actually go and what you would want automated first — that shapes which of these seven questions matter most for your situation. And however you get there, this sits alongside what we do and how the work itself actually happens.
What will I own at the end, and can somebody else maintain it? A good answer is specific: where the thing lives, whose accounts it runs in, who holds the passwords, and what another supplier would need to take it over. A weak answer grants ownership in principle while the working parts stay in the supplier’s accounts, so the only way to keep it running is to keep paying them. Ask it directly — if we stopped tomorrow, what would I still have?
Is the price fixed before you start? A good answer is a number, in writing, with the scope beside it and what would count as a change. A weak answer is a day rate with an estimate attached. Day rates are not dishonest, and some work cannot be scoped until somebody has looked at it, but ask what the ceiling is. You should never be in a position where the cost of finishing is unknown.
What happens if it breaks in month two? Things break. A good answer tells you who to contact, how quickly they respond, what is covered and what would be charged — before you sign, not after. A weak answer is a reassurance that it will not break. Ask what happens when the underlying tools change, because they do.
Will you tell me if the honest answer is that I do not need this? Often the right answer is no, or not yet, or you already pay for something that does this. A good answer comes with an example of when that happened. A weak answer agrees warmly that of course they would, then never finds an occasion. Test it: describe something you suspect is a bad idea and see whether they talk you out of it.
What will you not recommend? Everybody has a list of things they will not do, and hearing it tells you what somebody actually believes. A good answer is immediate and specific. A weak answer is that it depends. If every problem you raise sounds like a good fit for the thing they sell, ask more questions.
Can you show me it working on a bad week rather than a demonstration? Demonstrations are built from tidy examples, and real work is not tidy. A good answer offers to run the thing against your own awkward cases, and will show you what it gets wrong as well as right. A weak answer is a polished video. Ask what it does when it is unsure, because something that admits it does not know is worth more than something that guesses.
A good answer is specific. A weak answer is a reassurance.
How is my information handled? You should get a plain answer with no legal vocabulary in it: what goes where, which company ends up holding it, whether it is used to train anything, how long it is kept, and who can see it. A good answer is written down and specific to you. A weak answer points at a policy page. Ask what happens to it when the work ends, and expect that in writing.
If the work touches anything about a person — hiring, pricing, anything with a legal consequence — ask who makes the final decision. The answer you want is a person, with the machine doing the preparation.
It would be poor form to publish that and not answer it. So, plainly.
Ownership. You own what is built. It runs in your accounts, under your logins, with written instructions in plain English so somebody else could take it over without ringing us first.
Price. Fixed and agreed in writing before the work starts, with the scope beside it.
Support. We agree what happens when something breaks before you commit, and it goes in writing rather than being left to goodwill.
Honesty. We publish an article showing people how to check whether the software they already pay for does the job. It costs us work, and we would rather lose it than sell a build somebody did not need.
Limits. We will not recommend anything that leaves a decision about a person to a machine, anything that sends your information somewhere we cannot explain to you, or anything that only keeps working while one particular person stays available.
Proof, and this is the part worth reading. We cannot show you a system built for somebody else, because this is a new business with no client list yet. No case studies, no testimonials. Saying so is more use to you than implying otherwise. What we offer instead is to work on your own examples early, so you judge the thing itself rather than a story about it.
Information. We agree what goes where before anything is set up, we prefer approaches that keep your information inside tools you already control, and we write it down.
Who you would be dealing with. Twenty years in financial services, including experience with AI inside a large corporation, and years of using these tools in his own businesses long before selling any of it. Based in Glasgow, working across Scotland.
There are good people doing this work here, and a number of them will answer all seven well. Any AI consultant in Scotland worth hiring should be comfortable being asked. If one of them suits you better than we do, that is a good outcome.
If you would rather see these answered against your own situation, that is what the audit is for.
Twenty minutes on the phone, free, no obligation. Bring the list above and ask them in any order — that is a good use of the call.