Can’t Find the Value in your AI Initiatives?  You’re Measuring it wrong!

Most AI initiatives start with a quick win and a spike of optimism. A prototype that automates a task. A chatbot that actually works. Teams get excited, executives buy-in, and then reality chimes in: It’s not adding any value.

According to McKinsey's 2025 State of AI, 88% of companies now use AI somewhere, but less than one-third have managed to scale beyond experiments. And only 39% report any measurable financial impact.

But here's what that stat doesn't tell you: The problem isn't that AI isn't working. It's simply being measured incorrectly.

Most organizations treat AI ROI like any other technology investment—looking for direct cost savings or revenue attribution. That works for some AI applications, but fails spectacularly for others. When you try to force every AI initiative through the same financial lens, you’ll either miss its value entirely or kill a promising program before its value begins to compound.

AI creates value in fundamentally different ways, and each requires its own measurement approach. As I touched upon in my previous post, AI value can be categorized into three distinct waves.

Personal Productivity – Value That's Real But Not Reducible

This is where individuals use AI to draft emails, summarize documents, capture meeting notes or analyze data instead of grinding through spreadsheets. People save time and raise the quality of their work.

Traditional ROI wants to see fewer FTEs or lower costs. You won't find that here.

What you will see though are fewer hours lost to low-value work, more time for meaningful collaboration, more engaged employees who spot opportunities they'd otherwise miss.

What to measure instead: Adoption rates. Weekly Active Users. Employee sentiment. Time to competency on new tasks. Watch for when someone says "We always use it for this"—those organic patterns are your signal that a process is ready to formalize.

Don't skip this foundation. When you empower employees to experiment with AI, they become your best source of insights for operational efficiency and revenue growth. They become the experts that will tell you what works and what won’t.

Operational Efficiency – Where Traditional ROI Actually Works

This is where AI becomes much easier to defend, but only if you're applying it to the right problems.

If your process follows clear, repeatable rules with predictable inputs and outputs, traditional automation is probably your answer. AI shines when the problem requires assessment, analysis, judgment, or creativity.

Think about the difference: Routing a support ticket based on keywords? That's automation. Understanding the customer's underlying issue, assessing urgency based on context, and drafting a thoughtful response? That's where AI excels.

What to measure: Process-level outcomes. Cycle time. Error rates. Throughput. Customer satisfaction. Cost per transaction. These are the metrics finance understands.

The best operational initiatives either eliminate the mundane (freeing people for more valuable work) or turbocharge human capability (making people better at their existing jobs).

This is where you prove you can deploy AI, measure its impact, and support it operationally. For many organizations, this is the true first phase of enterprise AI: boring, measurable, and highly defensible.

Revenue Growth – Different Metrics, Longer Horizons

AI-powered products. Hyper-personalized experiences. Smarter forecasting. Done well, these create new revenue streams and real competitive advantage.

But trying to measure these the same way you'd measure operational efficiency is a recipe for killing them prematurely.

What to measure: Primary revenue levers tied to specific initiatives. Conversion rates. Deal size. Churn. Expansion. Customer lifetime value. Time to close.

Treat these as a portfolio of experiments with staged gates, not single bets. The feedback loop is slower and noisier than operational efficiency. You're building custom solutions, integrating with legacy systems, and taking on organizational risk.

And if your team hasn't lived with AI in their daily work and seen it transform their processes, asking them to trust it with revenue-critical decisions is a much harder sell.

The Real Lesson from McKinsey's Data

That 39% reporting measurable financial impact? Many of them are probably creating tremendous value in personal productivity or early-stage revenue experiments—but measuring it with the wrong ruler.

McKinsey's AI Rewired study confirms that value only scales when organizations adopt a defined set of capabilities in sequence, not at random. First the data plumbing. Then the workflow redesign. Then the talent shifts. Then the operating model changes.

You can't skip steps. But you also can't measure every step the same way.

In my final post of this series, I'll share a practical path forward—how to know where to start, what to track, and when you're ready to scale.

Which category is your organization struggling to measure right now? I'd love to hear what metrics you're tracking (or wish you were).

[Part 2 of 3]