Chasing the AI Illusion

There is surely nothing quite so useless as doing with great efficiency what should not be done at all. — Peter Drucker

We have a French Bulldog named Bossman Rossman. He is bold. He is confident. He is certain he is in charge. And on sunny afternoons, when the light hits one of the shiny decoys in the backyard at just the right angle, Bossman will chase the reflection convinced he will catch it. He chases it hard. He corners it. And then he stands there, panting, confused — because there is nothing there. It was a reflection. An illusion. And the crazy thing is, he keeps going at it. He does not learn that it is an illusion.

That is exactly what most companies are doing with AI right now.

The pressure to appear innovative has overridden the discipline to be transformative. Boards want AI announcements. CEO’s want AI roadmaps. So they deploy. They pilot. They procure. And then they get there, panting, confused, with nothing to show for it.

The data is brutal. Global organizations are forecast to spend $1.5 trillion on AI by the end of 2025.1 RAND Corporation documents that over 80 percent of AI projects fail to meet their stated objectives. A failure rate twice as high as traditional IT implementations.2 Gartner predicted that 30 percent of generative AI initiatives would be entirely abandoned after the proof-of-concept stage.3 McKinsey found that more than 80 percent of organizations report no tangible impact on enterprise EBIT.4 And BCG reveals that 60 percent of companies generate no material value from their AI investments — with only 5 percent creating substantial value at scale.5

The root cause is complex. Executives are captivated by a vendor demo or a macro trend and instruct IT to deploy — with no specific, quantifiable business problem attached. No outcome defined. Just: do AI.

We know how to run companies. We are far less skilled at reinventing them. Clayton Christensen documented why in The Innovator's Dilemma — the very capabilities that make a company excellent at executing today's model become liabilities when the model needs to change. Processes, incentives, and governance structures optimized for today actively destroy the conditions required for tomorrow. AI is exposing that structural flaw at speed.

We focus on solely on the “productivity” play and not solving the secrets of company where asymmetric lie. And then we wonder why employees aren’t behind the AI initiative while reading about AI driven layoffs in every media outlet.

An overwhelming 84.4 percent of implementation experts agree that successful AI strategy must be driven strictly by business pain points, not software feature sets.6⁷ Yet executive pressure to demonstrate relevance overrides this logic, resulting in AI deployed for the sake of AI. This is what I call innovation theater — leadership treating AI as a plug-and-play technology initiative while neglecting organizational change management, rigorous problem definition, and workflow redesign.7

In Big Bet Leadership, I argue that the Hyper-Digital Era has transformed strategic tension from a theoretical risk into an existential threat.8 But there is a critical distinction between making a Big Bet and chasing a shiny object. One is disciplined. The other is dangerous.

Share

Just as companies are discovering that basic generative AI has not delivered, the vendor community is already pushing the next shiny object: agentic AI, semi-autonomous programs capable of reasoning through multi-step logic and executing actions across enterprise software with minimal human supervision.9 By early 2025, 62 percent of organizations were already experimenting with AI agents and nearly a quarter had begun scaling them.10 The pressure to adopt is intense. The readiness to succeed is, guess what, low.

Agentic AI is reliant on your underlying data architecture. Companies failed at generative AI because they skipped the foundational data work. Now they are jumping to agentic AI — which requires even more foundational work — without fixing what broke the first time.

The Right Framework: Structure and Operating System Together

In my recent white paper with Andy Forti, Big Bets in Zone to Win, we make the case that transformation failure is not a technology problem. It is a management system problem. Companies apply the metrics, governance, and cultural expectations of their core business to disruptive AI initiatives — and it crushes them. Between 70 and 90 percent of all major corporate transformations fail. Only 0.5 percent of megaprojects meet their cost, time, and benefit targets. The answer is not a better vendor. It is a better management system.

The framework requires two things working together. First, the organizational architecture — the what. Geoffrey Moore’s Zone to Win creates protected spaces where transformation can be nurtured without being suffocated by the core business. The Incubation Zone operates with venture capital governance. The Transformation Zone activates with CEO-direct command for one initiative at a time.

Second, the initiative operating system. This is “the how” of how leadership needs to lead. Big Bet Leadership provides the disciplined methodology including Thinking in Outcomes (future-back, risk-forward), the Big Bet Experiment Planner (stack-ranked hypotheses, not business plans), and Think Big, But Bet Small.

The leaders who win share three habits.

They Create Clarity. They start by asking: What sucks? What is the specific, vexing pain point we are eliminating? They define one killer feature — one decisive capability — validated before scale. Not a feature list. Not a pilot. One hypothesis, precisely defined.

They Maintain Velocity. They reduce internal friction so experiments run in weeks, not quarters. Organizations that validate or kill AI hypotheses quickly avoid the massive sunk costs that plague slow movers.

They Prioritize Risk and Value. They sequence experiments by highest-risk, highest-value hypotheses first. They Think Big, But Bet Small — maintaining bold vision while making small, reversible bets before irreversible commitments.

And the CEO is directly involved. CEO active involvement of AI transformation is the single most impactful element influencing EBIT attributable to generative AI. When AI strategy is delegated to the CIO or siloed in an innovation lab, it fails to achieve the cross-functional traction required for enterprise-wide transformation.

What separates these “future-built” companies? According to BCG, they realize that value doesn’t just come from the technology but from how they empower their people to capitalize on it. About 10% of value from AI comes from the algorithms themselves and another 20% comes from the technology required to implement them. The remaining 70% comes from rethinking the people component. This is the 10-20-70 rule.

Bossman Rossman eventually comes back inside. He is not embarrassed. He also does not adapt or learn. He is fully certain that next time, he will catch whatever that was. He will do exactly the same thing again.

Regrettably, too many companies act in the same way. They are running the same failed pilots, making the same governance mistakes, making the same leadership and change management miscues, skipping the same foundational work, and proud they can check the box on their AI strategy. The tragedy is not the wasted money. It is that nothing is being learned.

So stop chasing the light. Learn your lessons. Start chasing something real.

Bossman will be back chasing the reflections tomorrow. Watching. Waiting. Ready to sprint the moment something shiny catches the light.

Do not be Bossman.

Onward,

John

John Rossman

Managing Partner, Rossman Partners | Author, Big Bet Leadership, The Amazon Way & Think Like Amazon.

Your guide to making winning high-stakes decisions in the AI era.

www.johnrossman.com

See John in Action