Destination: Execution

June 29, 2026

The AI Metrics Look Good. The Transformation Doesn’t. Part One: The Traction Trap

This series was originally published on LinkedIn by ESG’s Enterprise AI Practice Leader, Richard Steele. Find it here.

Recently, my team was wrapping an exciting engagement, finalizing execution plans that included a multi-year AI implementation roadmap and governance framework designed to fundamentally reinvigorate our client’s longstanding market position.

It’s the kind of work that doesn’t look flashy, but it determines whether everything else will hold together afterwards. It also doesn’t sell elusive vanity metrics, but it establishes a durable, strategic, competitive edge.

As the engagement concluded, the CIO pulled me aside. He appreciated the work. He understood the logic. And then he said something I’ve heard in different forms across nearly every major technology transformation I’ve been a part of:

We need to get some wins on the board. The team needs to see progress. We’ll get to the foundational work, but right now we need to show the board we’re moving. We just can’t miss this opportunity with AI.

I get it. I’ve been involved in these conversations throughout my career, long before the priority became artificial intelligence. Navigating board pressure, customer perception, and technological advancements is no small feat. What I also understood, and what worries me about that conversation, is that the decision he was making wasn’t really about sequencing. It was about which priorities will get measured and reported. Once you optimize for what you can report, you’ve already made a much larger decision about what kind of transformation you’re actually building.

Unfortunately with AI, this is an all too common traction trap.

The Dashboard Doesn’t Lie. It Just Doesn’t Tell the Whole Truth.

I’ve watched this pattern play out across ERP implementations, M&A integrations, and major platform rollouts for the better part of three decades. The metrics looked clean. The transformation stalled.

The post-mortem almost always revealed the same thing: the organization measured what was easy to count (albeit not necessarily an important success indicator) and called it “evidence” of change.

When we look at the compounding gap between investment and effective execution, AI is following the same script, with one meaningful difference. The scale of investment is larger, the pace of change is faster, and the window to course-correct is shorter and less clear.

If you walk into almost any enterprise AI program “lessons learned” review today, you’ll find metrics that look like progress. New licenses got deployed and tools were activated. Several training sessions were held and the pilot programs launched. The numbers represent genuine organizational effort. In a board presentation, they’ll read as momentum.

What they don’t tell you is whether the organization has actually changed, whether the workflows are better, and whether the people using the tools have integrated them into how decisions get made. They don’t tell you whether this change was filed alongside every other system that got rolled out, celebrated, and quietly worked around until irrelevant.

Industry data reinforces what practitioners already know. Organizations that use their existing processes with new AI capabilities capture somewhere between 15 and 30 percent of available automation potential. Organizations that redesign their processes to fully leverage AI capture 40 to 60 percent. The gap between those two outcomes isn’t technology. It’s the foundational work that happens, or doesn’t, before the technology arrives.

Introducing “The Traction Trap”

There’s a specific dynamic at work here that I’ve started calling the “traction trap.” A leadership team makes board-level commitments on AI. The pressure to demonstrate near-term progress is real and legitimate. So the organization responds by prioritizing what’s visible: pilots that can be showcased, tools that can be counted, and training completions that can be reported. These are genuine wins. They show momentum. And they consume exactly the time, budget, and organizational attention that foundational work requires, while quietly discouraging the harder investments that don’t show up in a quarterly update.

The trap isn’t that the wins are wrong. It’s that they’re being treated as evidence of transformation when they’re actually evidence of activation. Activation and transformation are not the same thing, and right now most organizations don’t have a reliable way to tell them apart.

If you haven’t built the extra lanes before bumper-to-bumper traffic jams become the reality of your morning commute, freeway traffic will be a problem. The new lanes have to be there before the heavy traffic arrives, or construction delays may derail the anticipated benefits. Every organization deploying AI right now is managing an on-ramp. The drivers are coming, and expecting to operate in their designated lanes. The question is whether the infrastructure will be ready when they get there, or if the organization will spend the next several years managing congestion (and incidents) it could have prevented.

The CIO in the example above is a good leader. The board pressure he was responding to was important. The decision he made was rational given the incentives in front of him. That’s what makes the traction trap so insidious. It doesn’t require bad judgment. It only requires the tendency to optimize for what’s visible and defensible right now over what’s foundational and consequential later.

The Wrong Framework for the Wrong Problem

That CIO wasn’t making a mistake. He was operating inside a system designed to produce exactly that decision. And that is a difficult problem to solve.

Most organizations are reporting AI “progress” through frameworks built for a different kind of change. Project completion rates. Adoption percentages. Cost per training hour. These are project management metrics applied to a transformation problem. An outdated transformation problem. They answer one question: are we doing the things we said we’d do?

They don’t answer the question that actually matters in AI-related transformation: is the organization amplifying its capabilities?

The distinction sounds academic until you’re sitting in a board meeting eighteen months into an AI program, the metrics still look clean, and someone asks why the expected business outcomes haven’t materialized. Or what improvements the investments created. At that point, the window to course-correct has already narrowed considerably. The early wins are spent. The foundational work still hasn’t happened. And the gap between where the organization thought it would be and where it actually is becomes visible all at once.

This isn’t a measurement problem you can solve by adding “better” metrics to the existing framework. The metrics are a symptom. The underlying issue is structural.

Most organizations are managing AI transformation through change frameworks and accountability structures that were never designed for a workforce that is part human and part digital.

The way change gets owned, progress gets defined, and accountability gets distributed were all built for a different era. That structural problem is what makes the traction trap so persistent.

Early wins matter. The question worth sitting with is whether they’re building something or spending something. Those are not the same investment.

Read Part Two: The Visibility Gap and Part Three: The Legacy Map here.