TL;DR: Measure AI ROI where it shows up in the business, because task-level gains often fail to reach the income statement. Write the KPI before the spend and name a person who owns the link between task gains and financial results, then track whether those gains reach the books.
Why does AI ROI go missing?
AI ROI often goes missing in the gap between a faster task and a better business result. McKinsey’s The state of AI in 2026: On the road to ROI, published August 25, 2026, found that 37 percent of respondents attribute at least some EBIT impact to AI use, about the same share as the year before. Eighty percent reported that AI had improved their individual productivity, and the report describes the pattern plainly: “These individual gains have yet to translate into broad financial impact for organizations.”
The wait is also longer than many technology budgets assume. Deloitte’s October 2025 survey of 1,854 executives across Europe and the Middle East reported that “most respondents reported achieving satisfactory ROI on a typical AI use case within two to four years,” which “is significantly longer than the typical payback period of seven to 12 months expected for technology investments.” Only six percent reported payback in under a year.
In his analysis of enterprise AI ROI across seven landmark reports, Basil C. Puglisi, MPA, named the mechanism those reports left out. Organizations measure at the use-case level while the revealing data sits at the firm level, and “a process can run faster without the enterprise P&L moving if the time saved is reabsorbed elsewhere rather than converted into revenue growth or cost reduction.” In his reading, that gap “is not a rounding error but the place where value is structurally disappearing.”
Timing explains part of the gap. In The Minds That Bend the Machine, Chapter 13 reviews Erik Brynjolfsson, whose Productivity J-Curve describes why measured productivity can dip while organizations invest in data, process redesign, and training that accounting systems do not capture, before gains arrive later.
What should you measure?
The arithmetic is simple: net financial benefit minus total AI cost, divided by total AI cost. The hard part is proving the benefit reached the books, so measure whether task-level gains reach financial results and assign someone to answer for that connection. The ROI analysis recommends naming “a named person at the function level to own the translation between use-case efficiency and firm-level P&L movement,” someone “not responsible for the technology itself, but for connecting the efficiency gain to a line on the income statement and explaining quarterly why the connection held or did not.”
It pairs that owner with a signal leadership can track: the “percentage of use-case efficiency gains that appear as corresponding movement in firm-level P&L within two quarters.” A gap that persists “is a measurement architecture problem, not a technology problem, and it requires a governance response, not a technology upgrade.”
Two further checks keep the number honest:
- The Authorship Test. If the person presenting a productivity number “cannot defend it line by line against the question” of what rework consumed, then “the number is not yet authored but forwarded.”
- Where the money goes. The same analysis notes that budgets cluster in high-visibility applications while “The highest and most consistent ROI comes from back-office automation,” where the cost base is controllable and gains are directly measurable.
People belong in the measure too. Basil’s book Governing AI: When Capability Exceeds Control describes a workforce measurement crisis in Chapter 9, “Operational Governance Failures,” where the same AI can be deployed to replace human labor or to extend human capability, and organizations make that choice without validated instruments. An instrument for measuring how people govern AI, which the author calls HEQ and AIS, is meant to address that gap. It is “a deployed, research-grounded instrument at diagnostic stage, not yet validated on an independent cohort.”
How do you set the KPI before the spend?
Write the measure down before the money moves, so the result can’t rewrite the goal. A method that pairs every fact with a tactic and a measurable outcome, which the author calls Factics, makes the case in two sentences: “A KPI written after results arrive becomes a narrative, because whatever happened turns into the stated intent. A KPI written first is a commitment that can fail.”
The method is honest about its cost. Factics “slows decisions by requiring a defined measurement before approval,” and it “cannot say what an organization should want.” It also “sidelines claims that cannot be measured, without forbidding action on them,” which keeps hopeful projections out of the evidence column.
Measurement also determines what outsiders will credit. Basil’s analysis of insurance, The AI Risk Economy, puts it in one line: “An insurer cannot price governance it cannot verify.” The paper also states that “No actuarial evidence validates governance as a pricing factor,” so a measured program gives an insurer better evidence without guaranteeing a discount.
The frameworks carry their own evidence gap. In The Minds That Bend the Machine, Chapter 4 reviews Andrew Ng and acknowledges that no controlled study across multiple enterprises yet shows dollar-value improvement from these governance methods, a study that remains on the research agenda.
An outside reviewer can test those numbers. Puglisi Consulting, the author’s advisory practice, describes its AI ROI and Value Review as one that “ties AI spending to outcomes the business already tracks, so the return shows up in numbers leadership trusts.” It is one of several ways to get that outside read.
Questions Readers Ask
What is a realistic payback period for AI?
Deloitte’s 2025 survey found most respondents reached satisfactory ROI on a typical AI use case in two to four years, far longer than the seven to 12 months typically expected for technology investments. Only six percent reported payback in under a year, so set budgets and board expectations from your own payback estimate rather than a one-year assumption.
Why do AI productivity gains not show up in profits?
Time saved on a task often gets reabsorbed elsewhere instead of becoming revenue growth or lower costs. Rework spent checking AI output can also consume part of the gain. Without someone accountable for converting task gains into financial results, use-case reports and the income statement keep telling different stories.
Who should own AI ROI in a company?
Basil’s ROI analysis recommends a named person at the function level who owns the translation from use-case efficiency to firm-level P&L movement. That person owns the connection between the efficiency gain and a line on the income statement, and explains each quarter whether the gain reached it.
What KPIs should we use for AI?
Choose KPIs the business already tracks, such as cost per transaction or cycle time, and write them before the spend. Track net gains after rework, and add a signal for how much of the use-case gain appears in firm-level results within two quarters.
Should employee capability be part of AI ROI?
Yes. The same AI can extend people’s capability or replace tasks, and the return can depend on how well people direct and verify the output. Measuring that capability alongside financial results may help explain why similar tools produce different returns across teams.
Sources
Deloitte. (2025, October 22). AI ROI: The paradox of rising investment and elusive returns. https://www.deloitte.com/global/en/what-we-do/capabilities/ai/research/ai-roi-the-paradox-of-rising-investment-and-elusive-returns.html
McKinsey & Company. (2026, August 25). The state of AI in 2026: On the road to ROI. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
Puglisi, B. C. (2025). Governing AI: When capability exceeds control (2nd ed.). https://basilpuglisi.com/governing-ai-when-capability-exceeds-control/
Puglisi, B. C. (2026, April 2). Enterprise AI ROI: What seven landmark reports found, what they missed, and five decisions worth making now. basilpuglisi.com. https://basilpuglisi.com/enterprise-ai-roi-seven-landmark-reports-five-decisions/
Puglisi, B. C. (2026, May 24). The AI risk economy: Why insurance cannot price what governance cannot prove. basilpuglisi.com. https://basilpuglisi.com/ai-risk-economy-insurance-governance/
Puglisi, B. C. (2026, September 13). Measuring the governance competence at the center of AI literacy. basilpuglisi.com. https://basilpuglisi.com/measuring-governance-of-ai-literacy/
Puglisi, B. C. (2026). The minds that bend the machine. https://basilpuglisi.com/minds-that-bend-the-machine/
Puglisi, B. C. (n.d.). Factics: The method behind valuable content and trusted AI. basilpuglisi.com. https://basilpuglisi.com/factics/
Puglisi Consulting. (n.d.). Advisory reviews. https://puglisiconsulting.com/advisory-reviews/
#AIgenerated from #AIassisted content under HAIA Ecosystem
Free for personal, educational, and noncommercial research use with attribution. Commercial exploitation, paid productization, and enterprise commercialization require separate permission and licensing.
Suggested citation: Puglisi, B. C. (2026). How do I measure the ROI of AI in my business? basilpuglisi.com.
[…] How do I measure the ROI of AI in my business? […]