AI Transformation: ’Twas Ever Thus

I miss thinking about these kinds of things.

Earlier in my career, I worked as a software engineer and as a technology adoption and change-management consultant, primarily in the Enterprise Resource Planning (ERP) world. More recently, I’ve been applying what I learned about human behavior in the context of my therapy and consulting practice.

So I find myself especially intrigued watching, listening to, and reading about companies struggling to leverage AI.

A note before going further: I am neither praising nor damning AI here. I sincerely believe it has the potential to help us enormously or harm us profoundly. Those are important questions, but they are questions for another discussion.

What interests me here is something so familiar.

My observation—and perhaps my prognostication—is that the same human factors that have always fueled returns from technology transformation when they are present, and torpedoed those returns when they are absent, are becoming more important than ever.

AI implementations are moving faster, carrying more leverage, and unfolding in a more competitive environment than most technology transformations that came before them. Yet it is fascinating how often we appear to be ignoring lessons organizations have learned repeatedly over the past several decades.

Once again, many are throwing technology at the wall and waiting to see what “sticks.” And fortunes and careers will rise and fall as a consequence.

A familiar matrix

Remember the Boston Consulting Group’s once-revolutionary product portfolio matrix?

A similar stratification is useful when thinking about technology implementation portfolios, particularly where AI is concerned:

	Weak Technology Systems	Strong Technology Systems
Weak Human Factors	AI theater — pilots, demos, and low ROI that eventually erode investor, innovator, and user enthusiasm	Underutilized infrastructure — powerful and expensive technology that never translates into meaningful business returns
Strong Human Factors	Grassroots experimentation — energetic early adoption that eventually runs out of gas as skilled users become frustrated, disengage, or are poached by competitors	Transformation — AI becomes embedded in the organization’s processes and applied where it has the greatest impact

The technology may be different. The organizational challenge is not.

Human factors still determine whether transformation takes hold

For transformation to succeed, people have to:

• Want to use the technology.This includes incentives, reward/recognition systems, and that the tech provides wins for the users in the form of career upside,not that they become perceived as dispensable.

• Be empowered to change how work gets done, rather than simply layering AI on top of old processes.

• Experience leaders as guides and designers of measurement systems, keeping attention on outcomes rather than activity.

• Feel psychologically safe enough to experiment, knowing that some experiments will fail and that thoughtful failure will be treated as learning rather than punished.

• Receive training that helps them rethink the use of their subject-matter expertise—framing problems, judging outputs, recognizing exceptions, and designing AI-enabled workflows rather than merely completing tasks the old way with a faster tool.

The more I think about it, the more I come back to a simple formulation:

AI ROI = f(valuable use cases, adoption, workflow integration, data/system readiness, execution discipline)

If any one of those variables approaches zero, ROI tends to follow it.

We have learned this before

Quality Circles. TQM. Six Sigma. Kaizen.

Those of us who spent time with the work of Deming—and other thinkers such as Senge, Kanter, Feigenbaum, Kleiner, Feynman, and others—have seen versions of this movie before.

One enduring lesson is that the people closest to the work often understand the work best.

Frontline employees know where friction lives. They know which handoffs are absurd, which reports no one reads, which steps and approvals add no value, where errors repeatedly arise, and where knowledge gets lost between systems.  They know where complexity is important, and where it just gets in the way.

Engaging those people in redesigning workflows is therefore not merely a matter of employee engagement. It is a mechanism for discovering where AI can genuinely save effort, improve decisions, increase quality, or accelerate throughput.

Conversely, when “solutions” are instead imposed on employees, they frequently fail to match the actual work.

And another old lesson remains stubbornly true:

You cannot improve what you do not meaningfully measure.

Not AI adoption.

Not tokens consumed.

Not number of pilots.

Measure what matters: time, cost, quality, cycle time, revenue, throughput, customer outcomes, and error rates.

Here's what will distinguish 2036's corporate ghost campuses from the sparkling towers.

Start with business problems, not AI

When technology is applied to clearly articulated business problems—expensive, slow, error-prone, or revenue-constraining processes—we are much more likely to see meaningful returns. When teams deploy AI because AI itself has become the objective, we waste resources. A particularly perverse example is rewarding the consumption of AI—tokens used, tools activated, pilots launched, seats purchased—as though usage itself represented transformation.

It does not.

And the cost of these failures is greater than the money spent. Failed or poorly conceived initiatives leave behind a residue. The “stink” of failure attaches itself to subsequent AI efforts. Enthusiasm dulls. Skepticism increases. We end up going slower by trying to go fast in an undisciplined way. Employees become less willing to experiment the next time.

That organizational memory matters.

Ownership matters too

AI’s potential begins to break free when the line of business owns the initiative.

Executives closest to the business problem should set priorities, fund integration, remove barriers, hold teams accountable for outcomes, and reward behavior that produces those outcomes.

Without that ownership, AI easily becomes trapped as a succession of pilots run by IT, innovation groups, or technology enthusiasts—interesting enough to demonstrate, but disconnected from the systems where economic value is actually created.

One other thing concerns me

There is another dynamic I have been noticing, and perhaps others are seeing it as well.

Many of the actors shaping the AI conversation are extraordinarily young, smart, and technologically fluent. Those are enormous strengths.

What they cannot yet possess, through no fault of their own, is the crystallized intelligence that comes from decades of watching organizations change—and watching change initiatives fail.

Experience teaches things that technical fluency alone does not.

It teaches why safety, security, and governance have to be designed into systems rather than bolted on afterward. Experience teaches that changing technology also changes roles, incentives, status, identity, culture, training needs, communication patterns, and informal power structures. Accrued wisdom understands that people rarely resist “change” in the abstract, but they respond to what they believe the change means for them.

Finally, and maybe most importantly, it teaches humility: organizational systems have a remarkable ability to produce unintended, unanticipated consequences.

That wisdom should not compete with the intelligence and energy of the AI-native generation. The two need each other.

The encouraging part

There is much more to say about all of this.

But this fact that we are not entering entirely unknown territory profoundly reassuring. find one aspect of the current moment reassuring.

The models are new. The capabilities are extraordinary. The pace is unprecedented.

But many of the questions organizations now face—How do we build trust? How do we redesign work? How do we measure value? How do we align incentives? How do we involve the people closest to the work? How do we govern powerful technology without suffocating useful experimentation?—are questions we have encountered before.

We do know a great deal about how to navigate transformation.

The question is whether, in our excitement about what is new, we will remember to use what we already know.

-Disclosure: These thoughts come from my experience.  ChatGPT helped me organize them.