Why Data and Digital Transformations Fail: Evidence from 82 Senior Executives

Most data, digital, and now AI initiatives do not fail because the business case was wrong. They fail because the organization was never adjusted to deliver it. I surveyed 82 senior executives to find out how often that happens, and what separates the initiatives that deliver from the ones that quietly do not.

This is the research behind my Leadership Impact Project, the capstone of the Advanced Management Program at Harvard Business School, which focused on driving enterprise value from data, digital transformation, and AI. Responses are reported in aggregate. I designed and conducted the survey as part of the program. This independent research is my own; Harvard Business School did not sponsor, review or endorse the study.

What 82 executives reported

Survey of 82 senior executives, 2026. All figures are self-reported by respondents about their own organizations.

That last figure is the most telling. The five conditions behind it are organizational readiness assessed before capital was committed, dedicated ownership and resources, incentives redesigned to support the initiative, strong and clearly named executive ownership, and a method for measuring value defined in advance.

Key descriptive findings

Observed relationships

This is a transformation problem, not a technology problem

That reading is consistent with the wider literature. McKinsey’s 2025 State of AI survey found more than 80% of respondents were not seeing a tangible enterprise-level EBIT impact from generative AI, and that workflow redesign had the largest effect among the attributes it tested. Deloitte’s January 2025 State of Generative AI in the Enterprise research found more than two-thirds of respondents expected 30% or fewer of their experiments to be fully scaled within three to six months, while nearly three-quarters said their most advanced initiative was meeting or exceeding ROI expectations, a contrast that shows individual use cases can create value even while enterprise-wide scaling remains difficult. PwC’s 2026 Global CEO Survey found 56% of CEOs reported neither higher revenue nor lower costs from AI during the prior year, while 12% reported both. My survey adds a practitioner-level observation to that pattern: 65.9% of respondents said their organizations selected the technology before the commercial outcome had been defined, which is the point at which the rest becomes hard.

The failure is rarely dramatic. Resources get quietly withdrawn to cover business as usual. Two initiatives draw on the same finite talent pool with no stated first priority. The people who stand to lose control, budget, or status do not object in the meeting; they simply do not move. None of that shows up on a milestone dashboard until the return is already gone.

AI is repeating the pattern, faster

This survey asked about data and digital initiatives, because that is the wave most of these organizations have already been through. What is striking is how closely the AI conversation is now reproducing it. The technology is being selected before the commercial outcome is defined. Pilots are being added to teams who already have day jobs. Nobody will be paid differently if it works. Those are the same patterns that respondents associated with weaker delivery in the data era, and there is little reason to expect a different result from a different acronym.

If anything, AI compresses the timeline. The technology is easier to acquire than a data platform was, which removes the procurement friction that used to force at least some organizational conversation. It is now entirely possible to have AI running in a dozen places in the business without anyone having defined what commercial outcome it is meant to produce, who owns it, or how the return will be verified. Ease of adoption is not the same as readiness to benefit.

Methodology

This research used a voluntary convenience sample drawn from one senior-executive education cohort. The full AMP 210 cohort at Harvard Business School, approximately 148 participants though cohort membership changed during the program, was invited by direct email between March 16 and April 7, 2026, with at least two reminder messages. Eighty-two respondents completed the survey, an approximate participation rate of 55%. The findings describe the experiences reported by this sample and should not be generalized to all executives or organizations without further research.

Two follow-up executive interviews were conducted by video. Each lasted approximately 30 to 60 minutes and used a semi-structured format tailored to the participant’s survey responses and experience. One conversation explored an initiative viewed as successful and the other explored an initiative that did not produce the intended result. Interview insights are presented only as anonymous paraphrased themes; no company, industry, location or individual is identified.

Limitations

• The sample was not randomly selected and is not nationally or globally representative.
• Responses were self-reported and may reflect recall, perception and role-based bias.
• The survey was cross-sectional and cannot establish causality.
• Some subgroup analyses rely on small cell counts and are labeled exploratory.
• The meaning of success may vary by organization and initiative.
• Many initiatives were still in progress, which limits final-outcome interpretation.
• The two follow-up interviews were explanatory and not independently representative.
• The research did not validate a numerical risk score or universal success threshold.

What the follow-up interviews added

The two follow-up interviews highlighted different paths. The initiative described as successful had a clearer connection to the business objective, stronger accountability, and more disciplined attention to adoption and measurable outcomes. The initiative that underperformed reflected a gap between the intended strategy and the operating conditions required to execute it: a business case had been approved, but ownership, resourcing and the definition of value stayed unresolved after launch.

Because the conversations were tailored and the sample was limited, they should not be treated as independent proof. Their value is explanatory. They illustrate how the conditions identified in the survey can appear in practice, and why technically sound initiatives can still struggle to create commercial value.

Readiness, then strategy, then deployment

Readiness means diagnosing where today’s norms, power, and incentives will resist tomorrow’s strategy, and adjusting them before capital is committed. It is not a checklist run at kickoff. Alongside the conditions the survey measured, four structural requirements need to be in place before launch rather than after it:

1. An aligned senior team. If key leaders are not genuinely on board, resolve that friction before launch. It does not dissolve on contact with a roadmap.
2. Defined leadership. A clearly named leader, with the emotional intelligence to map where the cultural resistance sits and the standing to influence through inclusion rather than mandate.
3. Explicit resources and architecture. Protected capital, protected talent, and a structure insulated from business as usual, which will otherwise reclaim both.
4. Sustained support. Commitment has to outlast a single budget cycle; short-cycle support is how initiatives stall before they can deliver.

From Research Finding to Organizational Diagnosis

The research identifies conditions associated with stronger delivery. The Congruence Model, developed by David A. Nadler and Michael L. Tushman, provides a practical way to investigate those conditions before additional capital is committed. It is not a survey-derived scoring method. It is a structured way to determine whether the organization is configured to deliver the intended business outcome.

Technology investment is more likely to create value when the work, people, formal organization and informal organization are aligned with the objective being funded.

The checklist and diagnostic guidance translate the research findings into practical executive questions. They are not a validated numerical scoring model.

Nadler, David A., and Michael L. Tushman. “A Model for Diagnosing Organizational Behavior.” Organizational Dynamics, 1980.

What it looks like when it works

Transformations that work start from one of two kinds of gap, and the two are worth naming separately. Michael Tushman and Charles O’Reilly draw the distinction in Winning Through Innovation: a performance gap is something broken or inefficient in the business you already run; an opportunity gap is new value you could create. They argue the two need structurally separate teams, because the second will always lose to the first if they compete for the same people.

In one transformation I led, the performance gap was channel mix. Too much revenue was flowing through high-cost third-party channels. Shifting demand to direct channels meant the same customer could be acquired at lower cost and higher margin, improving the economics without changing the underlying product.

A related opportunity involved pricing. Combining customer willingness to pay with competitive market data by date and segment made it possible to present a more relevant offer at the right time. The work began with a clearly defined commercial outcome, a discovery phase before major technology commitments, and independent validation of the result.

Why I studied this problem

Earlier in my career, I was asked to help address complex technology initiatives that had defined objectives but lacked several conditions required for successful execution. Significant resources had been invested in engineering, development and product work, yet the initiatives were incomplete and the expected return had not materialized. Stakeholders were asking why the work was not finished and where the value was.

The existing presentations were polished and the high-level vision appeared logical. When I looked beneath the presentations, much of the rigorous work required to make the vision executable was missing. Process maps were incomplete. Dependencies were not fully understood. Business requirements lacked necessary detail. Manual workarounds and tribal knowledge shaped how the work was actually performed.

I met with the accountable stakeholders, reviewed the project documentation and financial objectives, assembled a cross-functional team spanning product management, architecture, engineering, business analysis and project management, and went to the front lines. We asked employees to show us how the work was performed, where their time went, what prevented them from accomplishing more, and which tasks or outcomes were not possible with the existing systems. Their knowledge revealed unnecessary workarounds, hidden dependencies and opportunities that were not visible in the executive-level materials.

The work also produced an important leadership lesson. Involving frontline employees was not enough. Existing project leaders needed to participate earlier so they could help shape the solution rather than experience the work as a judgment on their prior decisions. Transformation affects ownership, authority, professional identity and internal influence, and those dynamics have to be managed alongside process and technology decisions.

AI is the newest major transformation wave, following the internet, big data and broad digitalization. The technology is different, but many of the organizational risks are familiar. That is why I undertook this research.

What leaders should do before approving additional investment

Leaders should require a disciplined review of the objective, value hypothesis and operating conditions before committing additional capital. The purpose is not to slow innovation. It is to improve the probability that the investment can create measurable value.

• Define the enterprise objective and the commercial outcome before selecting the technology.
• Identify the accountable business owner and the executive sponsor with authority to resolve conflicts and secure resources.
• Determine whether the initiative requires a dedicated team, shared resources or a protected operating structure.
• Map the work, workflows, dependencies, systems and data before finalizing the solution.
• Assess whether the organization has the skills, structure, incentives and cultural support required for adoption.
• Establish milestones for delivery, adoption and business value.
• Define the evidence that will confirm or reject the business hypothesis.
• Agree in advance on how the business owner and transformation team will demonstrate value and how Finance will verify it.
• Set the conditions that would trigger continuation, correction, redesign, scaling or termination.

Proof of concept versus proof of value

A proof of concept establishes that a proposed solution can work. A proof of value establishes that it creates a measurable commercial or financial return.

The initial proof of value should be selected so that its measurable upside can justify or recover a substantial portion of the initial investment. Once that value is verified, the organization has greater confidence in funding and scaling additional use cases. The work is not complete when the technology is delivered; completion should be defined by the accountable business owner before the work begins, using measurable acceptance criteria and financial verification where appropriate.

When to pause, redesign or stop

A missed milestone or modest cost increase does not necessarily justify stopping an initiative. Complex transformation work frequently takes longer or costs more than initially expected. The greater concern is a pattern of missed milestones, escalating costs or weak adoption combined with unresolved conditions that were already identifiable, such as unclear ownership, inadequate resources, misaligned incentives, incomplete process understanding or weak executive sponsorship.

When delivery problems confirm previously visible readiness risks, leadership should pause additional investment and conduct a structured diagnosis before proceeding. The decision should be to continue because the hypothesis remains credible, correct specific execution or adoption issues, redesign the solution or operating model, or stop when the expected return no longer justifies additional investment.

How an underperforming initiative should be diagnosed

The diagnostic should begin with a comprehensive document review to establish what was approved, what was promised, how the solution was designed, and whether the documentation is sufficiently detailed to support execution.

• Business case, project charter and approved objectives
• Project plan, milestones, budget and financial model
• Board, investment committee and executive presentations
• Product requirements and business requirements documents
• Existing process maps and customer or employee workflows
• Architecture diagrams, architecture-review materials and technical specifications
• APIs, integrations and data flows
• Vendor proposals, contracts and statements of work
• Governance records, status reports, decision logs and RAID logs
• Adoption plans, performance measures and value-verification reports

The first interview should normally be with the executive sponsor or business owner who is financially accountable for the result. The next conversations should include the project manager and cross-functional project team, compared against the documentation to identify differences among executive expectations, formal reporting and actual execution. The investigation then follows the evidence, which may expand into frontline observation, process mapping, technology architecture, data, financial assumptions, governance, resources, incentives or adoption. It should not impose a predetermined explanation on every troubled initiative.

The final deliverable should include an executive summary, root causes, unsupported assumptions, financial exposure, remaining value opportunity, organizational barriers, prioritized corrective actions, a recommendation to continue, correct, redesign, scale or stop, and an implementation roadmap with accountable owners, milestones and value measures.

Governance as a sensing instrument

Most governance tracks milestones and spend. That catches problems after they have cost something. Governance that works surfaces resource withdrawal, misaligned incentives, and passive resistance early, which means it has to be looking at the organization and not only at the plan. The project team should maintain a current RAID log covering risks, assumptions, issues and dependencies, and at each governance review report progress against milestones, decisions required, scope or cost changes, emerging RAID items, adoption indicators, and movement toward the intended business outcome.

The accountable business owner should interpret what the evidence means commercially. Finance should verify financial assumptions and realized value. A monthly steering review comparing approved capital against independently verified results will find drift sooner than any status report. Governance is not a stop mechanism. It is a navigation instrument, and its real job is to protect the new behaviors long enough for them to take hold. The same principles that shaped data and digital transformation investments now apply to AI investment governance and value realization at the board level.

Technology Investment Risk Checklist

One missing condition does not guarantee failure, but every unresolved condition increases execution risk. Multiple warning signs indicate that leadership should diagnose the operating conditions before committing additional capital.

• ☐The business outcome was not clearly defined before the technology was selected.
• ☐The initiative lacks visible sponsorship from a senior leader with sufficient authority.
• ☐Leadership is not aligned that the initiative is an enterprise priority.
• ☐No executive is financially accountable for the value outcome.
• ☐The operating model, decision rights or ownership are ambiguous.
• ☐Employees are expected to deliver the transformation in addition to unchanged full-time responsibilities.
• ☐The initiative lacks dedicated or protected resources.
• ☐Existing incentives continue to reward legacy priorities and behaviors.
• ☐Workflows, dependencies, frontline requirements or data flows have not been mapped adequately.
• ☐Required capabilities are missing or unavailable.
• ☐Governance tracks activity and delivery but not adoption and value.
• ☐Progress is inconsistent, costs are escalating or the expected return cannot be verified.

What should a CEO or board ask?

• What business hypothesis are we testing?
• What evidence would confirm or reject it?
• What commercial or enterprise outcome are we funding?
• Did we define the outcome before selecting the technology?
• Who is financially accountable for the result?
• Does the initiative have a dedicated team or rely on existing resources?
• If it relies on existing resources, what work has been stopped, deferred or reassigned?
• Have we mapped the tasks, workflows, interdependencies, systems and data involved?
• Do we have the required skills and an accountable project leader?
• Do the formal structure, decision rights and governance process support the work?
• Are incentives and cultural signals aligned with the new priority?
• Does the team have the access and organizational authority required to investigate and act?
• What proof of value will justify further investment?
• What milestones should we expect for delivery, adoption and value?
• Which metrics measure delivery, which measure adoption and which measure business value?
• How will the business owner demonstrate value and how will Finance verify it?
• What assumptions are most likely to be wrong?
• What conditions would cause us to pause, redesign or stop?
• What must be true before the initiative is scaled?
• Who will own continuous optimization after implementation?

The practical version

Before the next initiative is funded, three questions are worth answering honestly. Can we describe exactly what work needs doing and who owns each piece? Do we have a dedicated team, or are we adding this to people who already have day jobs? Will anyone be paid differently if this succeeds?

If the answer to any of those is no, the business case is not the thing to revisit. The organization is.

Download the Executive Research Report

Get the complete executive report, including detailed findings, anonymous interview themes, external research comparison, the Technology Investment Risk Checklist and the CEO and Board Technology Investment Review.

Is a major technology investment underperforming?

Commission an independent diagnostic before approving more capital. If your organization is investing in data, digital transformation, technology or AI and is not seeing the expected progress or financial return, I can help determine what is preventing the initiative from creating value, right-size the investment and operating model, and lead the work through implementation and verified results. I am available for select full-time executive leadership, interim leadership and advisory mandates.

Start a conversation

For organization-wide engagements, explore executive and enterprise transformation advisory through State of Mind Strategies.