Executive Resources · for UK SME leaders
How to Sanity-Check AI Vendor ROI Claims
Every AI vendor will tell you their product saves time, reduces cost and delivers measurable returns. Most quote generic industry benchmarks, not results from organisations like yours. Five questions separate real ROI from sales fiction: comparable case studies, full implementation costs, who measures the result, what happens if the projection misses, and realistic break-even by use-case type. Then apply the sanity-check formula - minus 30% on benefits, plus 20% on costs, plus six months on the timeline - and see whether the business case still stands up.
Why vendor ROI numbers are usually wrong for you
The gap between the headline ROI figure in a sales deck and what your organisation will actually see is almost always wider than the deck suggests. The benchmarks come from somewhere - but rarely from a UK SME of your size, in your sector, doing your specific use case, in the last two years.
This is not necessarily dishonesty. It is the maths of marketing material. The case studies that get into the deck are the strongest. The implementation work that did not go to plan stays out. The five questions below are designed to drag the realistic picture into view before any commercial commitment is made.
Question 1: Show me ROI for organisations similar to mine
Generic claims like 'customers save an average of 40% of processing time' are meaningless without context. Push for the definition of similar: same industry, comparable headcount and revenue, same use case, and case studies from the last two years.
A red flag is an averaged figure with no named customer behind it: 'we have saved clients an average of £50,000 per year.' That tells you nothing about whether you will see those savings. A green flag is a named company of similar size, a specific use case, year-on-year savings figures, and a contact you can phone to verify the story.
Question 2: What are the full implementation costs?
Vendors quote software costs. They frequently omit everything else. When building a realistic year-one picture for a UK SME, the hidden costs typically include professional services to configure the system (£15,000–£50,000), staff training (£5,000–£15,000), integration work with your existing tools (£10,000–£30,000), any custom development required (£20,000–£100,000), and ongoing support contracts at 10–20% of software cost annually.
Request a full year-one P&L before any commercial decision. If the vendor resists producing one, that is the answer.
Question 3: Who measures the ROI - and how?
If the vendor measures ROI on your behalf, they have an obvious incentive to show positive numbers. The right answer is that you own the measurement, the vendor suggests the metrics, and your finance team validates the results. If ROI measurement will happen through the vendor's platform, make sure you can export the underlying data for independent verification.
- Hours saved per month from time-tracking data, not estimates.
- Error rate reduction, measured before and after deployment against the same definition.
- Cost per transaction tracked over time, not at a single point.
- Staff satisfaction scores from the team actually using the tool.
- Customer satisfaction indicators where the AI touches a customer-facing process.
Question 4: What happens if the projected ROI does not materialise?
The honest answer is that some organisations are poor fits for some AI tools. A good vendor knows this and will have a structured approach to handling it: a 90-day success guarantee with defined exit or remediation terms is a sign of a vendor confident in their product.
'Everyone sees ROI with this tool' is a sign of a vendor that has not thought it through, or hopes you have not. Push specifically on what the contract says about year-one underperformance, what data you can take with you on exit, and how the support relationship is structured if early results are below the projection.
Question 5: How long before break-even?
Realistic break-even varies sharply by use-case type. Use the table below as a sanity check on any vendor projection. Anyone claiming month-one ROI on anything more complex than simple automation is either misrepresenting the implementation timeline or defining ROI in a way that does not map to your actual business outcomes.
| Use-case type | Realistic break-even |
|---|---|
| Simple automation (document processing, scheduling) | 6–12 months |
| Complex AI (predictive analytics, customer insight) | 12–24 months |
| Strategic AI (market entry, product innovation) | 18 months or more |
The sanity-check formula
When a vendor gives you a projected ROI figure, apply three adjustments before treating it as credible. They are blunt on purpose; their job is to expose business cases that only work in optimistic conditions before you sign for them.
- Reduce the benefit estimate by 30% - vendors are optimistic and best-case scenarios rarely survive contact with operations.
- Add 20% to the cost estimate - hidden fees and scope creep are real and almost universal in year one.
- Extend the timeline by six months - implementation always runs longer than the Gantt chart in the proposal.
What the formula is really doing
If the adjusted numbers still support the business case, you have something worth pursuing. If they do not, the vendor's original claim was not credible enough to sign on.
Vendor ROI claims are a starting point for analysis, not a conclusion. The five questions and the formula together turn a sales presentation into a structured business case the finance director can actually sign off - and the board can defend later, when the actuals come in.
Take the next step
Want help applying this to your organisation? Use the resource below or book a 30 minute strategy call with Simon — no pitch, just practical advice.
Frequently asked questions
It is deliberately conservative. The point is that any business case which only works under best-case assumptions is fragile, and AI projects in particular tend to consume their margin of safety in implementation. A vendor whose ROI story collapses under a 30% benefit haircut, a 20% cost uplift and a six-month timeline slip was probably going to disappoint you anyway. A vendor whose story still stands up to those adjustments is worth a commercial conversation, and the formula has earned its keep.
Treat that as a serious red flag. Reference customers are how you verify that the case studies in the deck describe real outcomes rather than rounded marketing figures. A confident vendor will arrange one or two short calls without resistance. A reluctant vendor is either too small to have happy customers willing to talk, or has had outcomes they would prefer you did not hear about. Either way, you have learned something useful before signing.
Productivity gains are real but slippery. Convert them into a number your finance team will accept: hours saved per week multiplied by fully-loaded staff cost, or transactions processed per FTE before and after. If the vendor cannot help you make that conversion for your specific use case, the productivity number is a slogan, not a financial argument. Insist on a measurable definition before treating any productivity figure as part of the business case.
Only if you can export the underlying data and validate the result independently. The vendor has an unavoidable incentive to present favourable numbers, even with the best intentions. Your finance team should own the ROI calculation, the vendor can suggest sensible metrics, and the data should sit somewhere both sides can audit. If the vendor's platform is the only source of truth, the ROI conversation has already been lost before it has started.
It scales down sensibly. For a £50-per-user subscription, a five-question conversation and a quick adjusted-payback calculation is enough; you do not need a board paper. For anything that requires a configuration project, integration work or a multi-year commitment, run the full formula. The principle is the same at every price point: do not buy ROI claims you have not personally adjusted for optimism, hidden costs and timeline slip.
Related resources
Executive Resources
Compare AI Vendors Properly
A five-step method to compare AI vendors on your data, your cost model and your operational risk - not on demo accuracy percentages designed to be won.
Executive Resources
AI ROI Measurement
How do you actually measure what AI delivered? This framework runs through four steps, gives you a KPI library, and produces a monthly board report that finance directors can interrogate rather than just accept.
Executive Resources
AI Business Case Template
Boards reject AI cases built on vendor productivity claims. This template builds the case on your own baseline: a financial model, a risk register, and a governance section the audit committee can scrutinise.
Executive Resources
AI Tool Evaluation Criteria
Before committing budget to an AI tool, eight criteria decide whether it's worth it. Data security, integration depth, total cost of ownership, vendor stability - and the exit terms buried at the back of the contract.
Find Out Where AI Can Save or Generate Money in Your Organisation
Book a free 30-minute call with Simon. Bring a real problem - staff time, governance worry, vendor proposal, failing pilot - and leave with a concrete first step you can take next week.