Companies have discovered that AI costs money

Dear readers, I have monumental news to share: after three years of investing in Generative AI like drunken sailors in a port, companies have discovered something revolutionary, something that will change the corporate world forever.

They have discovered... mathematics.

Yes, folks. In the 2025 Wharton–GBK report, 72% of respondents said their organisations formally measured AI ROI. The survey covered about 800 US senior decision-makers at commercial enterprises with at least 1,000 employees and more than $50 million in revenue. These are reported practices, not an audit proving returns for every company. The figures below come from that report; the imaginary boardroom dialogue is my satire.

The Three Eras of Corporate Enlightenment

Allow me to guide you through this epic journey of corporate self-discovery:

2023 - The Era of Innocent Fascination

Also known as: “ChatGPT can write emails!”

• 37% used AI weekly
• General sentiment: “This is MAGIC ✨”
• Most common question: “What is a prompt?”
• Metrics used: None, why bother?

2024 - The Age of Carefree Spending

Also known as: “Let's experiment with the IT budget”

• 72% used AI weekly (+35pp)
• Spending increased by 130%
• Sentiment: “I'm still excited! (but less amazed)”
• Most common question: “Does anyone know if this works?”

2025 - The Age of “Oh Shit, We Need Results”

Also known as: “Responsible Acceleration” (because it sounds better than panic)

• 82% use AI weekly
• 72% now formally measure ROI
• Sentiment: “I need to justify this to the CFO”
• Most common question: “How much did you say we spent last year?”

The Welcome to Reality Committee

The best part of the study is how companies are “pivoting” (corporate speak for “doing what they should have done from the beginning”):

Before:

• “Let's buy all the AI licenses!”
• “Pilot in all departments!”
• “No matter how much it costs, it's INNOVATION!”

Now:

• “Um... does anyone have an Excel sheet with the metrics?”
• “What does ‘productivity increase’ mean?”
• “I need three studies that prove this is worth it.”

The Great Paradox of 2025

Here comes the juicy part: 88% of companies plan to increase their investment in AI over the next 12 months. But wait, there's more:

• 11% are now cutting budgets in other areas to pay for AI (+7pp vs. 2024)
• The favorite victims: “Legacy IT” and “HR Programs.”
• Translation: “We're going to fire the tech support guy to pay for more chatbots.”

It's like watching someone in a casino saying, “This time I'm going to be responsible with my money... after betting everything on red.”

The Winners (and Losers) Club

Sectors reporting positive ROI:

• Tech/Telecom: Obviously, they were selling the drug
• Banking/Finance: They have money to burn... sorry, “invest”
• Professional Services: Basically automatic invoices with AI

Sectors still looking for ROI:

• Retail: “Do customers buy more if a chatbot serves them?” Spoiler: No
• Manufacturing: Turns out AI can't weld physical parts (yet)

The Human Capital Shift

But here's the satirical gem: while companies celebrate that AI “increases employee skills” (89% agree), simultaneously:

• Investment in training DECREASED (-8pp)
• Confidence in training as a solution DECREASED (-14pp)
• 43% warn of a “decline in skill competence”

It's like saying, “This hammer will make you a better carpenter... but we're not going to teach you how to use it, and in fact, you'll probably forget how to use your previous tools.”

Investment in Internal R&D: The Plot Twist

30% of AI budgets are being allocated to internal R&D.

Let me translate this: after spending millions on commercial AI solutions for three years, companies have decided that “maybe we should build our own tools.”

It's the business equivalent of buying all your furniture from IKEA for your home, and after three years deciding, “You know what? I'm going to learn carpentry.”

The Verdict: ROI or Death?

Three out of four business leaders are already seeing positive returns. Or so they say. Or so they need to say to keep their jobs.

The reality is more nuanced: Tier 1 companies (over $2B in revenue) say it's “too early to measure” (translation: “we have no idea”). Tier 2 and 3 companies report faster ROI (translation: “we made up the metrics first”).

Conclusion: The Self-Fulfilling Prophecy

The irony of it all is that the “Responsible Acceleration” of 2025 basically means doing what any first-year MBA would have told you in 2022:

1. Define clear objectives
2. Measure results
3. Adjust based on data
4. Don't spend money like it's Monopoly

But of course, where's the fun in being prudent from the start, right?

The real ROI here is the friends... sorry, the “lessons learned” we picked up along the way.

Moral for 2026: If you hear an executive say, “This time it's different, we have a clear AI strategy,” ask them to show you the Excel spreadsheet with the metrics. If they hesitate for more than three seconds, run.