A Decision-Maker’s Checklist
A Bisque whitepaper for SME owners and decision-makers
Executive summary
You’ve read the case for AI. You can see where it might pay back in your business. Now comes the dangerous part: buying it. This is where money gets wasted, contracts get signed that shouldn’t, and “AI projects” quietly die eighteen months later having delivered nothing but invoices.
The technology is rarely the problem. The buying decision is. Vague promises, the wrong commercial model, lock-in that leaves you trapped, and “success” that was never defined — these are what burn SMEs, not the AI itself.
This paper is your protection. It covers how to spot snake oil, how to choose between building, buying, and a retainer, the questions to ask any provider before you sign, and how to define success so you can tell whether you got it. It’s the capstone of this series: everything we’ve argued for, with the guardrails on. The only thing we ask is that you don’t sign anything until you can answer the questions in Part 3.
Part 1 — How to spot AI snake oil
The tells are consistent once you know them. Be wary of any of these:
Vagueness about outcomes. A provider who talks endlessly about “transformation,” “leveraging AI,” and “future-proofing” but goes quiet when you ask “what number will move, and by when?” is selling a feeling. Real providers answer in enquiries, hours, costs, or errors.
No human in the loop. Anyone promising fully autonomous, no-oversight AI for anything that touches a customer, a price, or a legal obligation is either naive or dishonest. Good AI keeps your people in control of what matters.
The wrapper with a markup. A great deal of “AI” sold to SMEs is a thin layer over a tool you could broadly access yourself, resold at a healthy margin. Ask what it does that you couldn’t get from the underlying tool directly. Watch how vague the answer gets.
Solving a problem you don’t have. If a provider leads with their technology and then hunts for a use in your business, you’re being sold a product, not a solution. The right order is your problem first, technology second.
Pressure and hype. “Everyone’s doing it,” “you’ll be left behind,” “this offer ends Friday.” Good AI pays back on its own merits. It doesn’t need a countdown timer.
Part 2 — Build vs. buy vs. retainer
There are three ways to get AI into your business. Each is right for a different situation.
Buy off-the-shelf software. Best when your need is common and standardised — something thousands of businesses have identically. Cheapest and fastest to start. The trade-off: it does what it does, you bend to fit it, and you’re one of many customers. Good for generic needs; poor for anything that depends on how your business actually works.
Build custom. Best when the opportunity is specific to your business — your quoting logic, your workflow, your data — and worth solving properly. You get something that fits exactly and that you own. The trade-off: more upfront investment, and you need a builder who’ll do it right. This is where the biggest wins usually are, because the most valuable problems are rarely the generic ones.
Retainer / managed. Best when you want the outcome without owning the headache of running it — someone who builds, maintains, updates, and stays accountable to the result. The trade-off: an ongoing relationship, so the terms of that relationship matter enormously (see Part 4). Good when you value accountability and don’t want surprise invoices; risky if it becomes lock-in.
A simple way to choose: Is the problem generic or specific to you? Generic leans buy. Specific leans build. Do you want to own and run it, or have someone accountable for the outcome? Own-and-run leans build/buy; want-accountability leans retainer. There’s no universally right answer — only the right one for the problem in front of you.
Part 3 — Questions to ask any provider before you sign
Print these. Ask every one. A provider managing real outcomes answers with specifics; a provider selling activity reaches for adjectives.
On outcomes:
- What number will this move, and by how much, and by when?
- How will we measure it — and what happens if it doesn’t hit?
On substance: 3. What does this actually do that I couldn’t get from the underlying tool myself? 4. Where does a human stay in control, and where doesn’t one? 5. What happens when the AI gets something wrong?
On ownership — this is the big one: 6. If I leave, what do I keep? My system? My data? My customer relationships? 7. Is anything locked to you — can I take it elsewhere or run it myself? 8. Who owns the data this system touches and produces?
On the relationship: 9. What’s included in ongoing costs, and what triggers an extra invoice? 10. What does the first 90 days look like, concretely?
If a provider can’t or won’t answer these plainly — especially the ownership ones — that is your answer.
Part 4 — Ownership, lock-in, and data
This deserves its own section because it’s where SMEs get quietly trapped.
The bad deal looks fine at first: a provider builds you something useful, and you’re happy. Then you realise the system lives on their platform, your data is in their account, your customer relationships run through their tools — and the price to leave is starting over. That’s not a supplier. That’s a hostage situation with a monthly invoice.
Insist on the opposite, on principle:
- You own the system. What’s built for you is yours, not rented back to you.
- You own your data. Your customer data, your operational data — yours, exportable, not held in someone else’s account as leverage.
- You own the relationship with your customers. No supplier should sit between you and the people who buy from you in a way you can’t remove.
- No hostage-taking. You can leave, and when you do, you keep what matters. A provider confident in their work doesn’t need to trap you. They keep you by being worth keeping.
This is non-negotiable, and it’s a fast filter: a provider who resists it is telling you their plan is to make you dependent. Walk away.
Part 5 — Defining success so you can tell if you got it
The reason so many AI projects feel like a waste isn’t always that they failed. It’s that “success” was never defined, so no one can say either way — and ambiguity always gets read as disappointment.
Before you start, write down:
- The number. The specific metric this is meant to move — enquiries, hours saved, errors cut, costs reduced.
- The baseline. What that number is today. You can’t prove improvement without a starting point, and most businesses skip this.
- The target and the timeframe. What good looks like, by when.
- The review point. When you’ll honestly check, and what you’ll do if it’s not working — including stopping.
Every system should have a number attached. If it’s working, you scale it. If it isn’t, you cut it. That discipline — deciding in advance how you’ll judge it — is what separates an investment from a vanity purchase. It also makes the buying decision easier, because a provider who won’t sign up to a number is telling you something.
A note on how we work
Everything in this paper is how we run Bisque, not just what we advise. We start with your problem, not our technology. Every system has a number attached. You own what we build — your system, your data, your customers — with no lock-in and no hostage-taking. Ongoing costs are clear, with no surprise invoices. We hold ourselves to exactly the checklist above, because if we couldn’t pass it, we’d have no business asking you to.
What to do with this — and where the series leaves you
Over six papers we’ve made one argument from every angle: AI is genuinely valuable for SMEs in 2026, but only the specific, measurable parts, and only when someone understands your business before reaching for the technology. The rest is hype, activity dressed as outcome, and lock-in waiting to happen.
You now have the tools to tell the difference — the reality check, the opportunity audit, the case for speed over spend, the shift in how customers find you, and the guardrails for buying. Use them on us as readily as on anyone else.
Our free 30-minute audit is where it gets concrete for your business: where AI would actually pay back, where it wouldn’t, and what it would take — with a number attached. No obligation. No sales pitch. Just the maths.
Book a free audit at wearebisque.co.uk — or reply to the email this came with.
Bisque — AI automation and enquiry generation for UK businesses. We don’t sell activity. We build systems that produce results — and you own them.