First test
What baseline?
ROI is a comparison. If the baseline is last year's chaos, any tool looks like a win.
Reports
How to Read an AI Vendor ROI Claim
Takeaway. A vendor ROI slide is a sales argument with a number on it. Separate measured operating results from modeled stories that have no owner after signature.
Most AI vendor decks open with a payback number. The number is rarely the problem. The problem is that the baseline, the omitted costs, and the person who will defend the figure after go-live are missing. If you cannot name those three things, you are not buying a return. You are buying a story that will be re-argued in the budget committee.
Second test
What is omitted?
Review time, integration, failure handling, and second-pass cleanup usually sit outside the slide.
Third test
Who owns it?
After signature, who will defend the number to finance when the workflow does not match the demo?
The slide is not a model
A payback figure feels finished because it is large and the demo is fluent. Fluency is a delivery format. It is not evidence. What procurement needs before signature is a named standard of evidence: which operating state is the baseline, which costs sit in scope, and what would count as a measured result on your workflow. You do not need to rebuild the vendor spreadsheet to ask those questions.
MIT Project NANDA, in The GenAI Divide: State of AI in Business 2025, reported that the vast majority of integrated AI pilots remained stuck with no measurable P&L impact. That is a finding about measured financial effect, not a licence to invent a different failure percent. Read a vendor ROI slide against that standard: is the number a measured operating result, or a modeled story?
Five reading tests
Write these on one page before procurement. If a row is blank, you do not have a return. You have a claim looking for a sponsor.
- Decision test: what decision does this number authorize, and who is allowed to act on it?
- Baseline test: compared with what operating state, on which process, over which period?
- Scope test: which costs sit outside the slide (review, integration, failure handling, second-pass cleanup)?
- Evidence test: measured on your workflow, or modeled on a vendor sample?
- Ownership test: who defends the number to finance after go-live, and which budget pays when it misses?
Uncertainty does not retire at signature
A confident interface does not remove incomplete information. If your approval process treats fluent prose as proof, you have not designed a control. You have purchased a story that finance will stress-test after the first miss.
McKinsey / QuantumBlack, The State of AI: Global Survey 2025, found that nearly two-thirds of respondents said their organisations had not yet begun scaling AI across the enterprise. A slide that treats scale as a month-two certainty is arguing with that survey, not reading it.
A licence pile is not a business model
Vendor ROI often assumes that access equals capability. The harder test is whether the organisation can price, govern, and repeat the workflow after go-live. A claim that cannot name an owner, a measured process, and a review path is a procurement slide, not an operating plan.
Many AI budgets die in committee because nobody writes down what would falsify the number before signature. The gap is not missing software. It is missing judgment on what the number is allowed to authorize.
FAQ
What should I check first on an ROI slide?
The baseline. ROI is a comparison. If the baseline is last year's chaos, any tool can look like a win.
What costs are usually omitted?
Review time, integration, failure handling, and second-pass cleanup often sit outside the slide.
Who must own the number after signature?
A named internal owner who can defend the claim to finance when the workflow does not match the demo.
Where can I practise this skill?
Start with the free AI Fluency course and Coach tools, then certification if you want a shared standard.
Sources cited
- The Decision Maker's Handbook to Data Science. Apress / Springer Nature, 2020 and 2024 editions, sole author. How executives own data and AI strategy and the standard of evidence.
- Predicting the Unknown: The History and Future of Data Science and Artificial Intelligence. Apress (Springer Nature), 2023, sole author. ISBN 9781484295045. Uncertainty stays after the interface becomes fluent.
- Business Models in Emerging Technologies: Data Science, AI, and Blockchain. Business Expert Press, 2022, lead author with Mourouzis, Cardoso, and Zinopoulou. ISBN 9781637423134. Value is captured when a capability can be governed and repeated.
- The GenAI Divide: State of AI in Business 2025. MIT Project NANDA. The vast majority of integrated AI pilots remained stuck with no measurable P&L impact.
- The State of AI: Global Survey 2025. McKinsey / QuantumBlack, also framed as The State of AI in 2025: Agents, innovation, and transformation. Nearly two-thirds of respondents said their organisations had not yet begun scaling AI across the enterprise.
Practice, then a shared standard
Start with the Free AI Fluency Course and the AI Coach inside it. AI Fluency Certification is a four-week, project-based credential when you want a shared standard. Advisory is for organisations that need named owners, not another licence pile.
Tesseract Academy · How to Read an AI Vendor ROI Claim · Planning estimates only · More reports