The Great AI Stagnation
Why most corporate AI projects are doomed to fail, and how to escape Pilot Purgatory.
There is a quiet, well-lit corner of the corporate world where innovation goes to die. It is a place of endless presentations, promising demos, and perpetually deferred value. It is the state known as "Pilot Purgatory."
For the better part of a decade, Pilot Purgatory has become the final resting place for countless digital transformation initiatives. Today, its population is booming, fueled by the immense promise and paralyzing complexity of AI.
The story of AI in the modern enterprise is a study in this paradox. The ambition (and hype) surrounding AI is boundless, but the results are stubbornly terrestrial. While executives speak of revolution, their teams are often tasked with delivering contained, low-risk experiments. The logic seems sound: before we attempt to rewire the business, let's first prove the technology works. Let's run a pilot.
The problem is that a successful pilot and a successful business are two entirely different things. A recent CIO Dive report found that 46% of AI proofs-of-concept never make it to production. They work, but they don’t matter. They demonstrate technical feasibility but fail to create the gravitational pull of a compelling business outcome. They answer the question "Can we do this?" but leave the more important question — "Why should we do this?" — unaddressed.
The Seduction of the Wrong Metric
Consider the recent case of the Swedish fintech company Klarna. In early 2025, the company announced its AI assistant had handled two-thirds of its customer service chats in its first month—the equivalent work of 700 full-time agents. The news was hailed as a landmark achievement in AI-driven efficiency. On the surface, the metric is intoxicating. It is a clean, quantifiable measure of displaced work, a story of cost-cutting so profound it seems to redefine the economics of customer service.
But this is a dangerous and seductive narrative. As reported by Fortune, the initial triumph of automation quickly soured. Customers began to report that the AI's responses were generic and unhelpful, particularly for complex or emotionally charged issues. The system that was so efficient at closing conversations was failing at the more crucial task of resolving problems. The backlash was significant enough that the company, after championing its AI-first approach, found itself scrambling to rehire human support staff to restore service quality and customer trust.
Klarna is a case study in mistaking a technical output for a strategic outcome. The number of human agents an AI can replace is a measure of automation, not necessarily a measure of value. It tells us nothing of the quality of the customer's experience, the complexity of the problems solved, or the long-term impact on customer loyalty and trust. An AI can handle ten thousand conversations poorly, and the metric of "work handled" will still look impressive. This focus on legible, but ultimately superficial, efficiency metrics is a cautionary tale. It is the siren song that lures companies toward an "AI-first" mindset, where the goal becomes the deployment of the technology itself, rather than the reimagination of the service it is meant to improve.
The Original Sin of the Pilot
This brings us to the original sin of Pilot Purgatory. We design pilots to mitigate risk. They are exercises in containment, built to test a hypothesis within a carefully controlled environment. Can the AI integrate with our legacy system? Can the model achieve the required accuracy on our proprietary data? These are important technical and operational questions. A successful pilot answers them, providing a valuable "output"—a working demo, a validated technical architecture.
But a pilot, by its very nature, is a poor vehicle for generating demand for its own succession. The demand for a production-ready, enterprise-scale solution is not driven by a successful demo. It is driven by a business case so compelling that the risk of inaction becomes greater than the risk of implementation. A pilot is designed to prove something is safe; a business case is designed to prove something is necessary.
Herein lies the trap. In our quest to design a pilot that cannot fail, we often design one that cannot possibly matter. We aim so low that even on a direct hit, the impact is negligible. The incremental improvement it delivers is not enough to justify the cost, complexity, and organizational pain of a full-scale rollout. The project is declared a "successful pilot," and the team moves on to the next experiment, leaving the organization no more transformed than it was before. The output was a success, but the intended outcome—meaningful business momentum—is entirely absent.
The Fallacy of Incremental Transformation
This reveals a deeper, more uncomfortable truth: true transformation, the kind that rewires a business process or creates a new form of customer value, can rarely be achieved in cautious increments. It is a step-change, not a gradual slope. It requires a fundamental rethinking of the work itself, not just the automation of existing steps.
A pilot that automates three steps of a broken, twenty-step process is not a stepping stone to transformation; it is a monument to the limits of incrementalism. It makes a bad process slightly more efficient. It does not create the elegant, intelligent, and resilient workflow that a business-first approach would demand. It is a tactical success that reinforces a strategic failure.
The escape from Pilot Purgatory, therefore, does not lie in building a better pilot. It lies in writing a better business case before the first line of code is written. It requires leaders to define the strategic intent with such clarity and to quantify the value of the desired outcome so persuasively that the pilot is reframed. It ceases to be an experiment to see if something is possible and becomes the first, deliberate step in a journey toward a destination that the organization has already agreed is worth reaching.
The challenge for leaders, then, is to have the courage to aim higher. It is to resist the allure of the easily-won pilot and instead embrace the difficult, cross-functional work of building a consensus around a vision for true transformation. It is to understand that the purpose of innovation is not to produce successful demos, but to build a more successful business.
Looking to move past the pilot? Book a discovery session or follow me on LinkedIn for regular insights on AI in business.
