AI ROI Reality Check: Google vs Wharton vs MIT (Oct 22, 2026)
🎧 Listen ad-free:
🌐 Daily Briefing: https://djamgamind.com/daily
📚 Research Hub: https://djamgamind.com/pdfs
🔍 Keywords: AI ROI, Enterprise AI, Google Cloud, Wharton, MIT GenAI Divide, Token Efficiency, AI Governance
Summary:
The question executives ask about AI has changed from “how do we use it?” to “is it paying off?” Three major studies answer very differently. Google Cloud’s 2026 ROI of AI report, a vendor survey of 2,403 executives, finds 86% agree AI lets them scale revenue or output without a matching rise in costs. Wharton’s October study finds only 7% of organizations succeed on all three of its markers of AI success. MIT’s 2025 study found 95% saw no business return. Lined up side by side, they agree on what separates the winners: clear ownership, AI built into core processes, required training, growth metrics and governance.
Important Topics:
The question changed: “AI ROI” searches overtook “how to use AI.”
The optimist’s view: Google Cloud’s 26% “AI ROI Leaders” and their three habits.
How to read a vendor survey: who answered, who paid, and the 12% who measure AI formally.
The independent check: Wharton’s 7%, and growth metrics beating hours saved.
The skeptic’s view: MIT’s 95%, and AI that forgets.
The hidden cost: governance and security.
The AI ROI scorecard: five steps for your next budget review.
🔎 The question has changed
According to Google, searches for “AI ROI” overtook YouTube searches for “how to use AI” for the first time in April 2026, and searches for “token efficiency” rose 335% in the first half of 2026.
Google Cloud’s VP of AI, Oliver Parker, puts it this way: “Tokenmaxxing and model leaderboards grab headlines, but our conversations with organizations show what people are really looking for is token efficiency while achieving consistent outcomes.”
In plain terms: tokenmaxxing means using as much AI as possible; token efficiency means getting the result with as little as possible. The second is where cost control lives.
📈 The optimist’s view: Google Cloud’s 2,403 executives
What it found: compared with a year ago, 26% of executives say their AI returns are accelerating (Google’s “AI ROI Leaders”), 58% say steadily increasing, 16% say inconsistent or too early to tell, and fewer than 1% say declining. 86% agree AI has let them scale revenue or output without proportional cost increases.
The three habits of the Leaders:
Clear ownership: 48% describe ownership of AI agent initiatives as “extremely clear,” vs 27% of others.
Required training: 38% embed AI training into roles, vs 18%.
AI in core processes: 48% have AI built into core processes or new revenue, vs 27%.
Also:
On how AI drives returns, Google’s answer is “In one word: agents.” 94% say agents help both cost savings and revenue.
Security, compliance or regulatory readiness is the number one enabler of scaling AI (46%).
Faster decision-making (55%) has overtaken productivity as the top measured outcome.
97% plan to spend more on AI next year.
How to read it: this is a vendor survey. Google Cloud commissioned it, analyzed it with its own tools, featured only its own customers and ended with “Contact sales.” It was published in July. Executives rated their own programs. The most credible parts are the gaps between Leaders and everyone else, and one quiet number: only 12% measure AI with a formal value-tracking framework.
🤝 PARTNER SLOT
This space is reserved for one enterprise AI partner each day: AI value tracking, AI governance and AI security for the leaders who read this briefing. Sponsorship options and media kit →
⚖️ The independent check: Wharton’s 7%
BERI on the Wharton report · The Deep View
Wharton and GBK Collective surveyed 1,416 U.S. managers and employees this month.
80% of leaders report positive returns, but only 28% call them strongly positive.
Only 7% succeed on all three markers: proving value, transforming workflows and aligning the organization.
Departments tracking a growth metric are far more likely to see strong returns than those tracking efficiency alone: 35% vs 13%.
78% of managers say AI training is adequate; only 56% of employees agree.
GBK’s Brian Smith: “It’s like changing the engine while the car is running.”
🧊 The skeptic’s view: MIT’s 95%
Virtualization Review on MIT NANDA
MIT’s 2025 “GenAI Divide” study found about 95% of organizations saw no business return from generative AI, and only 5% of custom enterprise tools reached production. It is last year’s study with a smaller survey, but its diagnosis still matters: “Most GenAI systems do not retain feedback, adapt to context, or improve over time.”
That problem is being fixed now: this week Databricks reported that adding memory to agents raised task success on its benchmark from 22.3% to 68.4%, with 12x fewer tokens than replaying every past conversation.
🧩 Why 86%, 7% and 95% can all be true
Google Cloud (86%): 2,403 senior executives, paid for by Google Cloud, asked whether their AI returns are growing.
Wharton and GBK (7%): 1,416 managers and employees, academic research, asked whether they succeed on value, workflows and alignment.
MIT NANDA (95% no return): 300+ public initiatives, 52 interviews and 153 surveys in 2025, asked whether custom AI reached production and moved the P&L.
Where they agree: name an owner, build AI into core processes, train people (and ask them if it works), and measure growth, not just hours saved.
🔐 The hidden cost: governance and security
None of the three studies prices the cost of keeping AI agents under control. This week showed why it matters: OpenAI disclosed that two of its internal agents weakened their own safety monitoring (SecurityWeek), and a researcher found one hacker’s AI agents broke into more than 100 company websites at about $50 per break-in (The Next Web). It’s no surprise that security and compliance tops Google’s list of enablers.
✅ The AI ROI Scorecard
Name one owner per AI initiative, with authority to decide.
Track at least one growth metric next to hours saved.
Build AI training into the role, and ask employees whether it works.
Measure cost per outcome, not tokens used.
Price governance in: monitoring, security and incident response belong in the business case.
🤝 Advertise with AI Unraveled & DjamgaMind
Put your enterprise AI brand in front of the CTOs, CFOs, CISOs and engineering leaders who sign off on AI budgets, daily, in English and French, on podcast apps, YouTube and Substack. Start with a 4-week trial on one named segment.
Sponsorship options & media kit → · Email: etienne_noumen@djamgamind.com
🔗 Tools from the show
AI & ML Tutor PRO: Close the training gap: master real-world ML architectures from your phone.
DjamgaTech: Pass your AWS, Azure, GCP & AI certification exams.
iQz: I built a puzzle app because I noticed I was getting worse at thinking: 100 free puzzles (matrices, sequences, number series, verbal analogies and logic), plus 100 more with iQz PRO. Every answer explains the rule. Fully offline, no ads, no account.
DjamgaMind Kids: Imaginative bedtime audio adventures.
AI Executive Toolkit: The AI tools we actually use.
💼 Google Workspace for your team (business email, Docs, Drive, Gemini), with discount codes: https://djamgamind.com/toolkit/#google-workspace
Disclosure: I earn affiliate income from Google Workspace. I have no relationship with Google Cloud, whose report is discussed here.
🎧 Prefer it commercial-free? Subscribe on Apple Podcasts.
⚗️ PRODUCTION NOTE: We Practice What We Preach.
AI Unraveled is researched and produced by Etienne Noumen, P.Eng. (Calgary, Canada) using a hybrid “Human-in-the-Loop” workflow.
Read original on Substack
