• Skip to primary navigation
  • Skip to main content

Puglisi Consulting for Artificial Intelligence

Independent AI Reviews: Strategy, Governance, and ROI, Before and After You Deploy

  • Home
  • Advisory Reviews
  • The Advisor
  • Marketing Across Verticals
  • Track Record
  • AI & Digital Marketing Blog
    • AIO Prompts
  • Basil Puglisi’s Blog

AI Studies and ROI: 240 Executives Reported AI Impact to Prophet, and Reported Is Not Measured #AIg

November 21, 2026 by HAIA Agents

Prophet logo
Image: Prophet (EZ Newswire), via Fortune Press Releases (2026).

Two studies published in November 2026 both ask people what AI did for their business. Prophet’s report came out on November 20, 2026, and EY’s private markets survey offers a contrast. Neither study measured anything inside your company. For a business owner or a finance and operations leader, that gap decides whether a return figure belongs in a budget meeting or only in a slide.

What Prophet asked, and what its release leaves out

Prophet’s survey polled 240 executives at global companies. Prophet sorted respondents by growth: high-growth organizations are those with revenue gains of 10% or more in the last 12 months. The headline finding is that 72% of high-growth organizations report significant positive business impact from AI, compared with just 20% of moderate- or no-growth organizations.

Two more figures matter to anyone who owns a budget. Prophet reports that unclear ROI was cited as a major barrier to scaling AI pilots by 100% of no-growth companies and 83% of moderate-growth companies, but by only 39% of high-growth companies. It also reports that 83% of high-growth organizations use AI-linked leadership KPIs, compared with 48% of moderate-growth organizations and 0% of no-growth organizations, and that among companies with AI-linked KPIs, 57% report significant positive AI impact, compared with just 2% of companies without them.

Read those as survey answers, and label them as Prophet’s. The impact figures are what executives reported, not what an outside party measured. The release text does not state how many respondents sit in each growth group or how significant positive impact was defined. The copy of the release on Fortune’s site is also labeled Paid Content, which is a reason to go to Prophet’s full report before quoting any number.

Basil’s caution is about direction. The groups are defined by revenue growth, and the impact is self-reported, so the release cannot say whether AI produced the growth or whether growing companies had the money and appetite to adopt AI. The KPI finding has the same problem. A company that already sees results may be the one that bothers to write KPIs. Prophet’s data points to a link between having AI-linked KPIs and reporting impact. It does not show that setting KPIs creates the impact.

What EY found when investors and managers were asked about AI returns

EY’s 2026 Global Private Markets Survey is a useful contrast because it asked two groups about AI. The research comprised two parallel surveys, one of institutional investors and one of alternative fund managers, conducted with the online panel of the Gerson Lehrman Group.

The two groups did not agree. EY reports that 46% of institutions believe AI has already improved their investment returns, while only 19% of managers say AI has improved investment performance. EY also reports that managers view AI spending as generating only modestly positive ROI to date, and that 24% cite operational cost reductions as a significant factor. On adoption, EY reports that managers’ efforts have largely focused on increasing personal productivity rather than automating entire processes.

EY also reports that Aaru simulated investor responses gave a more conservative assessment of AI’s impact on investment performance to date. Basil’s reading: a simulated respondent is a model’s estimate. It is a fair extra data point, and Basil classes it as machine checking machine, which is Responsible AI. A model’s estimate cannot stand in for a number your own team measured.

One more limit applies. EY surveyed private markets professionals. Whether a small marketing agency or a regional distributor would see the same split is Basil’s judgment call, not an EY finding.

A return figure is a measurement claim

Both studies report opinions about returns. Basil’s method for turning opinion into something a business owns pairs a fact with a tactic and a KPI that shows whether the tactic worked; he calls it Factics. Applied to AI spending, the fact is something your own team measured, the tactic is the specific change someone made, and the KPI is the number that should move if the tactic works. A return claimed without a fact you measured is a borrowed number.

Dashboards, vendor ROI calculators, and automated scoring are useful, and they are Responsible AI: machine checking machine. They can run all day without anyone being accountable for what they say. AI Governance is different, because a named person holds a checkpoint and decides to accept, modify, or reject before a figure goes to a board, a lender, or a client.

The fact, tactic and KPI a named person sets before claiming a return

These steps are Basil’s recommendations, not findings from either study.

  1. Fact: measure the baseline before the tool. For one workflow, record hours per finished unit, cost per finished unit, and the share of work redone, with dates and the name of the person who measured. Without a dated baseline, any later improvement is hard to defend.
  2. Tactic: name one owner for each workflow. That person accepts, modifies, or rejects the AI output and the change to the process, and the decision is written down.
  3. KPI: pick two or three numbers tied to the baseline. Cost per finished unit, rework rate, and cycle time are practical starting points. Basil recommends a review at 90 days, and recommends counting only the change that traces back to the baseline.
  4. Count the full cost. Include subscriptions, setup time, and the hours reviewers spend checking output. A saving that ignores review time overstates the return.
  5. Keep the record. A short log of what changed, who decided and why, which Basil calls CARCS, lets a finance leader reconstruct the number later.

When a vendor, consultant, or industry study quotes a return, ask the same three questions of it: who measured, what was the baseline, and who signed the result.

Where the boundary stays open

A clean baseline shows what changed in a workflow. It does not show why. Pricing moves, seasonality, a new hire, and an AI tool can all land in the same quarter, and neither Prophet’s survey nor EY’s can settle attribution for a single company. Deciding how much of a gain to credit to AI remains a judgment that a named person has to make and defend.

Sources

  • – EY. (2026). 2026 EY Global Private Markets Survey. https://www.ey.com/content/dam/ey-unified-site/ey-com/en-gl/insights/wealth-asset-management/documents/ey-gl-2026-global-private-markets-survey-11-2026.pdf
  • – Prophet. (2026, November 20). Prophet report finds AI’s ‘ROI risk gap’ separates high-growth companies from slower competitors [Press release]. Fortune Press Releases (EZ Newswire). https://fortune.com/press-releases/prophet-ai-roi-risk-gap-high-growth-companies-2026-11-20/
  • – Puglisi, B. C. (2026). Factics: The method behind valuable content and trusted AI. basilpuglisi.com. https://basilpuglisi.com/factics/
  • – Puglisi, B. C. (2026). HAIA-CARCS: Compliance accountability record and case study. basilpuglisi.com. https://basilpuglisi.com/haia-carcs-compliance-accountability-record-case-study/

FAQ

Why can’t a business just use the ROI percentage from a study?

A study’s percentage describes the people it surveyed, in their conditions, using their own definitions. Your return depends on your costs, your baseline, and your workflow. The study can suggest where to look, and your own dated measurements have to supply the number.

What does self-reported mean for an AI ROI survey?

It means respondents described the results themselves, rather than an outside party measuring them from financial records. Self-reported figures are still informative about sentiment and practice. They are weaker evidence of actual returns.

What baseline should a small business record before adopting an AI tool?

Pick one workflow and record the hours and cost per finished unit and the share of work redone, with dates and the name of the person who measured. Basil recommends doing this before the tool goes live, because it is hard to reconstruct reliably afterward.

Who should sign off on an AI return figure?

A named person who can explain how the number was measured and who accepts, modifies, or rejects it before it is shared. That is the checkpoint that separates AI Governance from automated scoring alone.

Is a simulated survey the same as a survey of people?

No. A simulation models respondents rather than asking people. It can add a comparison point, but it does not replace people who answered or a number your team measured.

#AIg

Share this:

  • Share on Facebook (Opens in new window) Facebook
  • Share on X (Opens in new window) X

Like this:

Like Loading…

Filed Under: AI Consulting, Digital & Internet Marketing, Mobile & Technology

Reader Interactions

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

© 2005 to 2026 Puglisi Consulting | basil@puglisiconsulting.com | basilpuglisi.com

%d