RUMBLINGS
Transformation Myopia: Why visible progress can obscure what transformation requires
The most-cited statistic in the transformation literature has been 70%, the share of large enterprise transformation programs that fail to achieve their original ambitions. The number has been reproduced in McKinsey research, in Kotter’s foundational work, in Prosci’s benchmarking studies, in Boston Consulting Group analyses, and in countless press releases announcing yet another methodology designed to address the pattern. The 70% number has functioned as warning and as alibi for two decades. It has shaped how enterprises plan, how vendors sell, and how consulting firms position the expertise they offer.
It is no longer the right number. Bain’s database of more than 24,000 transformation initiatives found that only 12% of business transformations achieve their original ambitions. 88% fall short of their intended goal. The trajectory of two decades of measurement has not been a story of improvement under the weight of better methodology. It has been a story of slow deterioration as conditions have grown more demanding and the organizational muscle to meet them has not been built.
This is not a story of poor diagnosis. The literature on why transformations fail is substantial, accurate, and stable. The reasons have been documented across waves and methodologies. Process debt, the silos that prevent cross-functional work, the friction between strategic intent and operational execution, the failure to embed new ways of working into business-as-usual, the leadership patterns that reward visible progress over durable capability. Each pattern has been named, measured, and discussed in print for at least fifteen years. Recent contributions to the field, including a thoughtful piece by Karim Lakhani, Jared Spataro, and Jen Stave in HBR earlier this year naming seven structural frictions slowing AI transformation, continue to add precision to a picture that has been substantially clear for a long time.
So the pertinent question, and the one this piece is concerned with, is not what causes transformations to fail. It is why enterprises have remained so consistently unsuccessful at addressing the known causes, and what that long pattern of unaddressed failure means for an AI era less tolerant of those gaps than any preceding technology wave. The cognitive error worth naming, because so much of the current AI conversation is committing it, is the one I will call transformation myopia. It is the practice of seeing the friction at the last mile while missing the system of upstream conditions that produces it, wave after wave.
It is easier to identify and act on implementation challenges than to identify and confront the organizational dynamics that contributed to them. That, in turn, directs disproportionate attention toward visible progress and performative measures of success while the less visible organizational conditions required to create durable value remain unresolved or underinvested.
Why diagnosis has not closed the gap
The reason naming the failure patterns has not changed the failure rate is that the patterns are not the problem. The patterns are the predictable output of a system of conditions that enterprises operate within, and the conditions have remained stable across waves. The conditions interact, reinforce each other, then are individually documented and collectively unaddressed at scale.
Incentives in most enterprises reward visible activity rather than durable value. The executive who approves a scope override in the first quarter is promoted before the operational disruption appears in the third. The product owner who stops defending strategic logic under sprint pressure has moved to the next program by the time adoption stalls. The CIO literature has recently begun calling this pattern incentive drift, and the mechanism it describes is precise. The temporal misalignment between decision-makers and the consequences of their decisions means the feedback loop required for organizational learning never closes. People absorb a vague sense that transformation is hard, but they rarely absorb the specific correction tied to the specific decision a specific leader made. This is one reason two decades of accurate diagnosis has produced two decades of repeated failure. The decision-makers receiving the diagnosis are not the decision-makers who produced the problem, and by the time the problem surfaces, no one is left who needs the lesson.
Incentives in most enterprises reward visible activity rather than durable value.
Most enterprises are also wired vertically. People are organized by function, rewarded for working within their function, measured against function-specific metrics, and held accountable for function-specific outcomes. Transformation requires horizontal work across functions. The shared outcomes that successful transformation requires are rarely produced because the structure does not reward them, the measurement system does not capture them, and the incentive system does not reinforce them. Enterprises talk about shared outcomes routinely. Very few are organized to produce them.
The measurement system itself contributes to the pattern. Most organizations track what is countable rather than what is consequential. Deployment milestones, project completion, headcount changes, and budget adherence are easy to measure. Cultural change, cross-functional capability, sustained adoption, and organizational readiness for the next wave are harder to quantify, and they are therefore quietly deprioritized in the planning process. Optimization follows the metric. Durable capability rarely gets built because the system does not measure it, which means it does not appear in any review that anyone is held accountable for achieving.
Then there is the pattern John Kotter named Error 7 in 1995, the practice of declaring victory too soon. Quick wins generate the visible progress that leadership needs. They also generate a release of executive attention, often within weeks of the declared win, while the cultural and structural changes that the program was meant to embed have not yet taken hold. The organization reverts to legacy behaviors under pressure. The structures that produced the quick win are dismantled because they are seen as program scaffolding rather than standing capability. The temporary becomes permanent in everything except the capability that needed to be sustained.
The leadership accountability gap reinforces every other condition. Bain’s database of more than 24,000 transformation initiatives shows that large-scale change efforts achieve 24% more of their planned value when a dedicated Chief Transformation Officer oversees them. That difference is evidence that dedicated transformation leadership matters, not simply that a transformation role exists on an organization chart. The value of the role comes from having the capacity and authority to make the structural decisions the transformation requires.
These conditions together produce the seventh and most consequential pattern, the one that contains all the others. Transformation has been treated, almost universally and across every preceding wave, as an episodic state with a beginning and an end. A program is launched, structures are stood up, the work is executed, and the program is closed. The next strategic technology shift arrives, and the enterprise begins from approximately the same position as before. The structures stood up for the prior program were scaffolding, not capability. The institutional knowledge accumulated during the work disperses with the team that built it. The organizational muscle that would have made the next wave easier was not built, because the system that delivered the program was not designed to build it. That 88% fall short of their intended goal is the predictable output of this system working as designed. The recurring failures of cloud transformation, of digital, of ERP before it, all have the same shape because the underlying system that produces them has not changed.
Why AI raises the stakes
The conditions producing the failure rate are not new. What is new is the cost of leaving them unaddressed and the speed at which the cost is now coming due.
McKinsey’s 2025 global survey found that 88% of respondents said their organizations regularly use AI in at least one business function, while only 39% reported enterprise-level EBIT impact. Its 2026 survey found that AI high performers still represented only about 6% of respondents. The pattern is familiar: adoption is moving far faster than enterprise value realization.
McKinsey found that while AI use is widespread, only 21% of organizations using generative AI have fundamentally redesigned at least some workflows. That reflects a familiar pattern across technology waves. Organizations are far more likely to introduce new technology into existing processes and operating models than to redesign the work around the business outcome they are trying to achieve. Redesign is difficult, disruptive, and rarely contained within the boundaries of a technology implementation.
With AI, that tradeoff becomes more consequential. Reinvention is not incidental to the opportunity. It is a significant part of the opportunity. Applying AI to an existing process may make that process faster or more scalable, but it does not necessarily make the process better suited to the business need. At scale, AI can automate and accelerate processes that were never designed to take full advantage of the technology, increasing the speed and reach of the very limitations the organization should be reconsidering.
AI inherits the organizational state that preceded it.
AI is also arriving in enterprises carrying unresolved process, data, and operating-model challenges from prior waves of transformation. Many organizations are layering AI onto processes earlier technologies left largely intact, training models on fragmented data environments, and deploying AI into operating models that never fully evolved with the technologies that preceded it.
The unfinished work of prior waves becomes a constraint on AI value capture.
Speed is the other variable. Many prior technology waves offered a reasonably clear future state, even when getting there was difficult. A cloud transformation, for example, could be organized around an ambition to move applications and data to the cloud, exit data centers, or operate in a cloud-native model. AI offers no similarly clear destination. The technology is still evolving, the business models around it are still taking shape, and organizations are being pushed to move quickly before there is broad agreement on what the future state should look like. Everyone is accelerating, but toward a destination that remains unsettled.
There is no version 1.0 of an AI strategy that can be handed off to operations and declared complete.
That makes the unresolved organizational conditions more consequential. AI compresses the time enterprises have to recognize and recover from structural mistakes, while increasing the speed and scale at which those mistakes can propagate.
When AI becomes a workaround for transformation
That creates a familiar temptation. The hardest parts of transformation are rarely the technology itself. They are the organizational and cultural challenges that take longer to address, are harder to quantify, and are more difficult to demonstrate as visible progress. AI can make it appear possible to bypass some of that work rather than confront it.
If process debt is the obstacle, automate the broken processes. If institutional knowledge is the obstacle, capture the knowledge in AI systems. If cross-functional coordination is the obstacle, deploy agents that span the boundaries. If labor cost is the obstacle, reduce the workforce and redirect the savings toward AI investment. If people or roles are perceived as difficult, duplicative, or no longer necessary, remove them before fully understanding the work they perform or the gaps they have been compensating for.
The appeal is obvious because automation, headcount reduction, and AI investment are all easier to quantify and present as evidence of action than the slower work of redesigning an operating model, changing decision rights, addressing cultural friction, or rebuilding a process around the business outcome it is meant to serve.
But replacing people, processes, or functions before understanding the system underneath them risks scaling the very dysfunction the transformation was supposed to resolve.
That mindset is being reinforced by a broader public conversation about AI, productivity, and workforce substitution. Mustafa Suleyman, the CEO of Microsoft AI, has predicted that AI will reach human-level performance across many professional tasks. Dario Amodei, the CEO of Anthropic, has warned that AI could eliminate a significant share of entry-level white-collar jobs, while Ford CEO Jim Farley has predicted substantial reductions in white-collar employment. These forecasts are different in substance and intent, but together they contribute to an environment in which replacing human effort can begin to look like a strategy in itself rather than one decision within a much larger redesign of how the enterprise creates value.
That logic underestimates what organizations are asking AI to absorb and overlooks the conditions under which automation has failed to deliver previously. Experienced people often understand the inefficiencies in an operating model better than anyone because they have been compensating for them. In large enterprises, they have been doing invisible work to absorb data fragmentation, integrate disconnected systems, route around broken processes, and translate between functions that the structure prevented from communicating. They have been making the architecture work despite the architecture’s flaws. Removing them in pursuit of AI-driven efficiency does not fix the architecture. It exposes the gaps to AI without the human judgment that had been holding the system together.
Early evidence should at least challenge the assumption that deploying AI or reducing human effort automatically translates into greater productivity or enterprise value. Research on AI use across professional work is producing uneven results, and in some cases the technology has made experienced workers slower rather than faster. The point is not that AI cannot create substantial value. It is that the value does not appear automatically simply because more work has been automated or fewer people are doing it.
A randomized controlled trial from METR found that 16 experienced open-source developers completing 246 real tasks in repositories they knew well took 19% longer when using early-2025 AI tools. The authors explicitly caution against generalizing that result to software development broadly, but it is a useful counterweight to assumptions that AI use necessarily improves productivity in every context. Gartner’s 2026 survey of 350 executives at large enterprises found a similar disconnect at the organizational level: about 80% of organizations piloting or deploying autonomous capabilities reported workforce reductions, but reduction rates were nearly the same among organizations reporting higher ROI and those reporting modest or negative outcomes.
The response to two decades of transformation failure cannot be to skip transformation. It has to be to finally do it differently.
How leadership has to change in the AI era
How leaders are rewarded shapes how they lead, and for decades the rewards have pulled in one direction. Among the 250 largest U.S. public companies, total shareholder return was used by 73% of companies as a long-term incentive measure in 2024, and 90% of performance-based awards were measured over a three-year period. Current board research reflects the same pressure: in NACD’s 2025 survey, 41% of directors at publicly traded companies identified capital-market or investor expectations as a key source of short-term pressure, while 47% said the board itself was a source of that pressure. The result is a predictable set of priorities: optimization over adaptation, cost reduction over capability investment, and the near-term wins that keep stakeholders confident over the harder work of earning their confidence in a longer bet. Sustaining belief in a long-horizon investment is more demanding than delivering a visible quarterly result, and it asks more of a leader. The reward system favors the more expedient path, and the pressure to keep one’s standing and collect on those incentives pulls most leaders toward it.
AI requires leadership that can hold direction without performing false confidence about the destination. As the technology evolves, the regulatory environment continues to take shape, and competitive dynamics keep shifting, AI strategy will have to remain more adaptive than traditional planning cycles typically allow. The leadership demands of the AI era are different in kind from those of the prior era, not merely different in degree.
Russell Reynolds’ 2026 Leadership Through Uncertainty study found that 52% of CEOs identified strategic thinking as the skill with the greatest influence on organizational health, while 87% said they were comfortable being imperfect. The study describes the leaders who stand out as those able to operate effectively amid ambiguity, make thoughtful decisions without complete information, and adjust course as events unfold.
AI requires leadership that can hold direction without performing false confidence about the destination.
The alternative exists, though it is rare, and the reason it is rare is the same reward structure. In the late 1990s and through the dot-com collapse, when e-commerce still represented less than 1% of U.S. retail sales and Amazon’s share price had fallen dramatically, Amazon continued to prioritize long-term market leadership over short-term profitability. Sustaining that course meant persuading shareholders to judge the company on its long-term value rather than its near-term price, and it was possible in large part because the company’s governance gave its leadership the room to absorb short-term pain without being forced to reverse course. Most enterprises do not have that protection. Their leaders sit inside the reward system described above, where pressure to produce visible near-term results comes from their own boards and investors. Two things have to be true at once. The reward and governance structure has to change to make long-term, adaptive leadership the rule rather than the rare exception, and leaders themselves have to stop reaching for the convenient, well-compensated path when that choice comes at the expense of their people and the durable value of the companies they run. They are not powerless inside the system. If the people with the most influence will not show that conviction, it is hard to see who will.
What makes the performance of certainty so striking is how little supports it. AI’s future is still being written, enterprise value remains uneven, and the economics are not yet behaving the way many cost-reduction narratives imply. Organizations are reducing headcount in anticipation of AI-driven efficiency while continuing to increase AI investment, even though those workforce reductions are not consistently translating into higher returns and many enterprises are still struggling to demonstrate value at scale. Rather than slowing investment while the economics become clearer, many organizations are doing both at once: removing labor and committing more capital to AI. The result is not necessarily lower cost. In many cases, it is a shift in where the organization is spending money before it has established that the new model creates greater business value.
At the same time, transformation has become so ubiquitous in executive and consulting language that the term itself is becoming a weak signal of expertise. It appears across titles, profiles, role descriptions, and service offerings to the point that it can be difficult to discern who has led complex transformation, carried accountability for the outcome, or learned from failure. If everyone is a transformation expert, the label stops telling us very much. That raises a harder question: if enterprises have accumulated decades of transformation research, methods, case studies, data, and claimed expertise, why are they still so consistently poor at it?
The AI era requires a harder and less performative form of leadership. The leaders who will navigate this wave are not the ones who move fastest to look certain or to shed headcount. They are the ones who can set a direction worth committing to before the destination is fully clear, hold it steady while the particulars keep shifting, and change course in the open without treating the change as failure. They build the conditions that let an organization keep learning, keep adapting, and keep its people willing to tell the truth about what is and isn’t working, while preserving human agency by design.
What reinvention requires
Continuous reinvention can feel like a mystery because most organizations have never treated it as an organizational discipline that can be deliberately built and nurtured. Instead, the capabilities that make reinvention possible are typically assembled around a specific initiative, managed as separate workstreams, and then retired or dispersed once that initiative ends. What is rare is integrating those capabilities into something the organization can sustain and draw on repeatedly.
These are not the only capabilities transformation requires. They are the ones most often separated from implementation, underinvested in, or treated as temporary, despite being critical to whether transformation produces durable value.
Organizational conditions
Organizational conditions make any of the rest possible. Organizational culture matters because it is expressed in the behaviors that determine what happens in practice, not simply in the values or beliefs leaders say should guide the organization. Trust and psychological safety, change aptitude as an organizational capability rather than a project skill, a learning orientation, cross-functional collaboration as the default rather than the exception, and a leadership posture that absorbs the friction real change generates rather than displacing it onto the organization all shape whether transformation can take hold. These conditions are often separated from implementation or treated as secondary to it, even though they materially influence whether new capabilities become institutionalized.
Cross-functional capability with real authority
The mechanisms most enterprises run are widely practiced and individually appropriate: program structures, integration forums, working sessions, and Centers of Excellence with cross-functional representation. The mechanisms are rarely the problem. The capability is whether those mechanisms are treated as the standing way the enterprise does strategic work, with real decision rights, financial accountability that converges at the outcome level, and operating rhythms that survive any single program. The difference is whether the authority is real and the capability is durable, or whether both exist primarily in name.
Systems thinking applied end-to-end
The work of transformation is not the sum of its parts. The value emerges from how the parts integrate. An enterprise can have a strong strategy, a competent data platform, a capable change function, and a defensible governance framework, and still produce no value, because the integration between them is treated as a coordination problem rather than as the work itself. Systems thinking is the orientation that lets leaders see the integration as the work.
Leadership for the AI era
Leadership itself has to evolve with the AI era. The technology is changing the nature of work and forcing decisions with long-term societal, cultural, and ethical implications, often before leaders can know the full consequences. That requires a leadership profile able to make decisions under uncertainty, hold competing interests and time horizons at once, preserve human agency where it matters, and make consequential tradeoffs visible rather than burying them inside an implementation plan. Building that kind of leadership also requires changing what gets promoted, rewarded, and modeled by the people in the most visible roles. None of the other three components survive without it.
What needs to finally change
Recent AI research offers a useful signal. HFS Research, in partnership with Infosys, found that only 12% of enterprises had reached what it describes as Purposeful AI, where AI is embedded into core strategy, culture, operations, and decision-making. The remaining 88% were still in earlier maturity phases, constrained by gaps in areas such as data, governance, process, talent, and leadership.
That pattern matters because the organizations furthest along in AI are not using the technology to bypass the underlying work of transformation. They are redesigning workflows, reshaping operating models, strengthening data and governance, and changing how people and technology work together in order to create value from AI at scale.
The evidence emerging around enterprise AI does not support the assumption that reducing labor and increasing AI investment will automatically produce greater value. AI use is widespread, but enterprise-level financial impact remains limited, and Gartner’s research suggests that workforce reductions associated with autonomous technologies are no more likely to produce higher returns. The stronger signal is coming from organizations that redesign workflows and operating models around what the technology makes possible rather than treating automation or headcount reduction as the outcome itself.
What has been missing for two decades is not the diagnosis. It has been the institutional discipline to treat reinvention as a standing and durable capability, and the leadership willingness to invest in the less visible conditions that produce durable value rather than only the activity that produces visible credit.
What has to change is the way enterprises build and retain the capacity to transform. The capabilities, leadership, integration, organizational conditions, and learning required to produce durable value cannot be assembled for each initiative and then dispersed when it ends. They have to become part of how the organization operates, so experience accumulates, weaknesses are surfaced rather than reinforced, and each transformation leaves the enterprise more capable of adapting as conditions change.