Five Questions to Ask Before Buying an AI Tool
The demo is slick, the case studies are glowing, and everyone on the call is nodding. Here are the five questions to ask before you sign; including the two that make vendors visibly uncomfortable.
An AI tool lands in your inbox. The demo is slick, the sales deck is full of case studies, and everyone on the call is nodding. This is exactly the moment to slow down. The questions to ask before buying an AI tool are simple but almost nobody asks them, because the demo is designed to make you feel behind.
I sit on the buyer's side of these calls for clients. These are the five questions I ask every time, and what each answer tells you.
1. What specific problem does this solve, in one sentence?
The most common misstep in AI buying is starting with the technology instead of the business objective. If you can't articulate the exact problem this tool addresses before the demo, you're buying on excitement. Excitement does not show up in your ROI.
The test is brutal and useful: name the problem in one sentence, without using the word "AI". "Our team spends six hours a week answering the same customer questions" is a problem. "We need to leverage AI" is a budget waiting to be spent.
If you haven't identified the problem yet, the answer isn't a better demo. It's an audit of your own operations first.
2. Where does our data go, and who owns it?
This question makes vendors uncomfortable. Ask it anyway.
Specifically: what data does the tool collect, where is it stored, is our data used to train your models, and what happens to it when the contract ends? You want plain answers to all four, in writing. If the contract uses ambiguous language about data ownership or training rights, slow down. Ambiguity here is always expensive later and in the UK, it can be a commercial and GDPR problem.
This is also where your own house needs to be in order. If you don't yet have a policy stating what data can go into which tools, write that first.
3. Who will train our team and who owns adoption?
Here's the number that should reframe every buying decision: 88% of HR leaders say their organisations haven't realised significant business value from AI tools.
The tool is rarely the problem. The gap between deployment and adoption is. So ask the vendor what structured training is included, and ask yourself who internally owns embedding this into daily workflows. A named person, not "the team". If your implementation plan is "roll it out and see", you're not buying software. You're buying shelfware with a monthly fee.
4. What does success look like at 30, 60, and 90 days?
Thi should not sound like vague success. Specific, measurable success referring to hours recovered, response times cut, a number that moves.
A good vendor will help you define this, because they've seen their tool work and know what it changes. If they can't tell you what should be demonstrably different within 90 days, that's a red flag about how well they understand their own product.
Write the numbers into your decision before you sign. Then actually check them at day 90, that review meeting is where bad purchases get caught early instead of renewed annually.
5. What happens when it gets it wrong?
Because it will. AI tools produce confident, fluent output that is occasionally wrong and that's not a flaw of one product, it's the nature of the technology.
The question isn't whether that happens. It's whether your organisation has a human review process for anything high-stakes, and whether the vendor is transparent about where their tool's limits sit. A vendor who talks openly about failure modes is showing you maturity. A vendor who says "our accuracy is basically perfect" is showing you the door you should walk out of.
The questions to ask before buying an AI tool, as a checklist
Before your next demo, have these five written down: the one-sentence problem, the data trail, the adoption owner, the 90-day numbers, and the failure plan.
Any vendor worth signing with can answer all five without flinching. Most buyers never ask which is precisely why so many AI purchases become line items nobody can defend a year later. The organisations getting real value from AI aren't the ones who moved fastest. They're the ones who asked better questions before they moved at all.
Before the demo: know your own answer to question one
The hardest of the five questions is the first one and it's yours to answer as the business owner, manager or leader. If you're not certain which problem in your business AI should solve first, our free AI Opportunity Score identifies your top three in three minutes.
Walk into your next vendor call already knowing your problem, your data rules, and your success numbers.
FAQs
What questions should you ask an AI vendor before buying?
Five essentials: what specific problem the tool solves, where your data goes and who owns it (including whether it trains their models), what training and adoption support is included, what measurable success looks like at 30/60/90 days, and how failures are handled and reviewed.
How do you know if an AI tool is worth buying?
If you can name the problem it solves in one sentence, define measurable 90-day success criteria, and identify a named internal owner for adoption, it's worth evaluating seriously. If any of those three is missing, the purchase is being driven by excitement rather than need.
Why do most AI tool purchases fail to deliver value?
The gap is usually adoption, not technology. 88% of HR leaders report their organisations haven't realised significant value from AI tools, most often because deployment wasn't followed by structured training and workflow integration.
Should you ask AI vendors about data training rights?
Always. Ask whether your inputs are used to train their models, where data is stored, and what happens to it at contract end; in writing. Ambiguous data clauses create both commercial and GDPR exposure, and clear answers are a mark of a trustworthy vendor.