How to Measure AI ROI (And Why the Usual Metrics Fall Short)

According to a fresh MIT study, 95% of generative AI projects stall before delivering results.

The report, titled “The GenAI Divide,” suggests that despite a rush of $30 to $40 billion in enterprise spending, the vast majority of companies remain stuck, unable to extract real value from their AI initiatives.

The finding quickly went viral, fueling a bunch of doom-and-gloom posts that we’re all in an “AI bubble” and the technology is massively overhyped.

But is that the real story?

I don’t think so. The problem we see time and time again isn’t technology. It’s the planning. It’s the way we measure. And it’s the impatience.

You can’t stick AI on top of broken processes or shaky strategy and expect results.

Right now, so many marketing teams are pouring money into shiny tools without building the foundations to make them useful.

It’s the equivalent of giving your team the keys to a Ferrari, but asking them to drive it on a pot-hole riddled road without a map. The engine’s powerful, but if the road isn’t fit, you’re not going anywhere fast.

That “95% of AI pilots fail” headline makes it sound like AI doesn’t work. It makes all the nay-sayers and sceptics very happy, and gives everyone an excuse to stay working as we do today.

What it really shows is our ongoing obsession with short-term ROI: hours saved, quick wins, cost cuts. The kind of metrics that are easy to track and look good on the surface.

But, like trying to measure a Brand campaign against click-through-rate – it’s kind of missing the point.

The real ROI of AI works exactly like Binet & Field’s long vs short of marketing.

The Short > Performance marketing → immediate impact, easy to track quick wins; think hours saved, productivity gains, faster but not necessarily better.

The Long > Brand building → harder, slower to track and measure – but the compounding impact that actually shifts the needle and grows the business.

Just like the original study – short term impact will only take you so far – if you don’t fix the backbone strategy behind it, you’re going to come unstuck when you try to scale.

AI ROI is no different. The short-term gains matter, but they’re only the tip of the iceberg. The compounding value – less hamster-wheel busy work, more time on strategy, better insights, smarter decisions, new ways of working, happier teams – currently sits below the waterline.

The 3 Levers of AI Value

If you want to measure AI properly, you need to look across three levers. Each matters on its own, but the real transformation happens when you connect all three.

1. Efficiency (short-term, easy to see)

Doing the same work, faster and with less effort.

This is where most teams start – leaning on AI for repeatable tasks, saving hours, speeding up workflows.

Examples: research and reporting in minutes, first drafts on demand, faster campaign launches.

Metrics: time saved, reduction in manual effort, cost per output, speed to market.

Watchout: Don’t expect instant results. Early AI projects need time to train, adopt, and iterate. Track leading indicators (adoption, accuracy, process improvements), not just cost savings.

2. Effectiveness (mid-term, where the quality lifts)

Doing the work better – improving precision, decision-making, and impact.

Examples: deeper consumer insights, more time on research, more strategic thinking, consistent briefs, stronger brand voice.

Metrics: engagement, accuracy, speed and depth of insights, conversion lifts.

Watchout: AI amplifies your strengths and exposes your weaknesses. If your data is messy, your strategy shaky or your teams aren’t aligned, you’ll feel it.

3. Innovation (long-term, hardest to measure but highest value)

Doing new work that wasn’t possible before.

Examples: real-time interactive content, AI-native workflows, new product ideas.

Metrics: speed to market with new ideas, earned media uplift, new revenue streams.

Watchout: This is the “brand-building” side of AI ROI. It’s harder to prove on a spreadsheet, but it’s what convinces boards to invest long-term.

So What Does “Good ROI” Look Like?

The teams in the 5% club winning with AI don’t just chase immediate impact.. They focus on longer term capability building, creativity, and competitive edge.
They measure both:

Leading indicators: adoption, process speed, model accuracy, cultural uptake.

Lagging indicators: revenue, margin, market growth.

It’s not just: “Did this save us money?”

It’s: “Did this help us work in ways we couldn’t before – and how it that shift delivering value”

The message is simple: build your road before you start driving.

The teams in the 5% club get this.

They apply AI with focus, measure what’s working, and scale when it makes sense.

And when they report back to the board on AI “ROI”, they’re not just talking about hours saved.

They’re showing value created — faster time to market, sharper insights, new growth levers.

Ready to learn more and lead your AI strategy?

We cover exactly how super-users in teams are doing this today; from learning to speak “boardroom” to ROI calculators and scaling your impact – it’s all inside our Leaders Program.