Dr. Klint Kendrick’s Insights on Why Enterprise AI Success Is Harder Than It Looks

What’s Inside: Enterprise AI projects have an 80+% failure rate. But when you look at why do AI initiatives end up failing, the reason isn’t usually technology limitations, but the fact that most organizations are not structurally prepared to scale it beyond pilots. In this conversation, Sarrah Pitaliya, VP of Digital Marketing at Radixweb, explores this challenge with Dr. Klint Kendrick of NYU, an expert in workforce transformation, M&A, and AI-enabled operating models. The discussion goes beyond surface-level AI adoption and examines the deeper organizational constraints that determine success or failure of AI projects.

Enterprise AI is entering a critical phase where adoption is no longer the challenge; execution is. Nearly 85% of enterprise AI initiatives fail and a majority of organizations report limited or no measurable financial impact from AI deployments at scale. This is despite rapid advancements in AI capabilities.

A lot of organizations continue to struggle when moving from fancy AI prototypes to fully-functioning AI software solutions in production. This reveals a deeper issue that goes beyond technology itself. The real constraint lies in how work is structured, how decisions are made, and how effectively organizations are prepared to operate alongside AI at scale.

In this conversation, Sarrah Pitaliya, VP of Digital Marketing at Radixweb, speaks with Dr. Klint Kendrick, Adjunct Faculty at NYU School of Professional Studies, where he focuses on HR’s role in M&A and organizational transformation. With extensive experience spanning workforce strategy, large-scale integrations, and AI-enabled operating models, Dr. Kendrick brings a rare systems-level perspective on enterprise change.

The discussion explores why AI initiatives consistently lose momentum after pilot stages and offers actionable inputs on what enterprise leaders should do to take their AI demos to successful scalable solutions. It examines practical breakdown points including workflow design, governance structures, data integrity, and managerial capacity, offering a grounded view of where enterprise AI efforts succeed or fail.

About Dr. Klint Kendrick

Dr. Klint Kendrick is an Adjunct Faculty at NYU School of Professional Studies and a global expert in organizational leadership, workforce transformation, and AI-enabled operating models. He has led large-scale workforce and M&A initiatives across more than 150 deals valued at over $15B, including major transformations at Walmart. His work focuses on integrating AI into enterprise operating systems, scaling workforce decisions, and redesigning organizations to balance human judgment with automation.

In this conversation, we discussed key factors shaping enterprise AI success at scale including:

• Why workflow design and ownership gaps break AI at scale
• How roles, managers, and decision paths must evolve for AI success
• Why data readiness and governance determine long-term performance

In Conversation with Industry Insiders

Dive in as we explore the organizational realities that determine if enterprise AI succeeds in practice and the strategic mindset shift required for impactful AI outcomes.

1. You've spent decades working on organizational leadership and workforce capability. How has your perspective evolved with the rise of AI in enterprise environments?

Thirty years in organizational leadership has taught me that major technology waves follow a predictable pattern. The technology improves quickly, leaders get excited and promote it, but the organizational system changes slowly, and outcomes land well below the promise.

AI is different in speed and scale, but that basic dynamic is playing out again. Organizations are putting AI tools in front of hundreds or thousands of employees in weeks. The tools may work, but the surrounding work system often is not ready for them.

That is why I do not think AI success is a technology issue. It is an operating model issue. Leaders have to make the right choices about workflow design, data quality, decision rights, manager capability, governance, and value capture. If those pieces do not move together, AI creates activity faster than it creates value. You end up with organizations swamped by AI slop.

2. Many organizations approach AI as a technology deployment. Why does this assumption often lead to failure in real-world operations?

When leaders treat AI as a technology deployment, the work gets handed to IT and success gets measured by adoption rates. The organization declares victory because people are using the system. But usage is not enough. Usage does not tell you whether the work is better, decisions are better, or whether the organization is capturing financial value.

The biggest failure pattern right now is the pilot-to-nowhere. The pilot works because the organization creates unusually favorable conditions: a narrow use case, an engaged team, high levels of direct support. Exceptions get managed manually. The tool looks great.

Then the company tries to scale into normal operations and results drop. The tech did not get worse but the conditions changed. Workflows are not redesigned, managers are not empowered, escalation paths are unclear. Ultimately, employees either work around the tool, trust it too much, or revert to the old process. None of those produce ROI.

3. AI is fundamentally changing how work gets done. What are the most critical shifts organizations underestimate when adopting AI?

Three come up consistently.

First, the exception problem. AI handles common use cases well. But it handles edge cases poorly, which is a problem because a lot of risk lives in the edge cases. Harvard Business School researchers found that when people used AI for tasks outside the tool's capability range, they were less likely to produce correct answers than people who did not use AI at all. The boundary is jagged, and most organizations do not work to understand this.

Second, manager bandwidth. AI adoption creates new work for managers as they end up reviewing outputs, coaching different behaviors, catching errors earlier. Most organizations add the tool without adding management capacity and then wonder why adoption stalls.

Third, the absorption problem. Freed time does not automatically become business value. Unless leaders make an explicit decision about what to do with that capacity, it gets reabsorbed into meetings, rework, and low-value activity. The productivity gain is real but the income statement never sees it.

4. In your experience, what typically breaks first when AI tools move from pilot to live operational environments?

Workflow clarity breaks first. In a pilot, the use case is tightly defined and ambiguity gets absorbed through direct support. When operations go live, that support disappears. Employees encounter cases the training did not cover and outputs that look credible but are incomplete or flat out wrong.

5. How should organizations rethink roles, responsibilities, and decision-making structures when integrating AI into workflows?

When designing and building AI software solutions, organizations should start with decisions, not tasks. AI changes where judgment sits in a workflow. Some decisions that required senior expertise can now happen closer to the work. Others need more human oversight because the risk of an AI error is higher than a comparable human one.

Before any AI-assisted workflow goes live, three things need to be clear. Who owns the output so if the AI is wrong, someone is still accountable. What will be measured, since leaders need to know whether the system is improving or drifting. And what stops, because if the old process keeps running next to the new one, the efficiency gain disappears. People end up doing both, which takes more time, not less. Organizations miss that last one most often.

6. You've emphasized that AI adoption is an operating challenge, not just a software challenge. What does a well-designed "AI-ready" operating model actually look like?

It is the operating environment that allows AI to create value once it leaves the pilot and enters real workflow. I look for five things, roughly in this order.

• Business ownership: The accountable owner of AI outcomes should be the business leader whose metrics are affected, not the AI team or IT.
• Baseline metrics: Leaders need to know how the work performs today before they redesign it around AI. Without a baseline, ROI becomes a debate about impressions.
• Workflow clarity: Once current performance is understood, the organization can redesign with discipline.
• Exception handling: What happens when AI cannot complete the work or produces a questionable answer.
• Feedback loops: Someone needs to own ongoing review and iteration.

The simplest test: can the organization explain who owns the outcome, how the work performs today, how it will change, what happens when AI is wrong, and how the system improves over time?

7. What role do data quality, system integration, and process clarity play in determining whether AI initiatives succeed or fail?

They matter, but not in the same way.

Data quality is a prerequisite. If the data the model needs is not clean and accessible, that is a data infrastructure problem that has to be resolved before AI enters the picture. System integration is a cost and complexity problem that organizations consistently underestimate.

Process clarity is what determines whether AI creates sustained value once it is running. McKinsey's State of AI research has identified workflow redesign as one of the strongest predictors of financial impact from AI. The organizations getting value are not just adding AI to work. They are changing the work around AI.

8. How should leaders design oversight, escalation paths, and feedback loops to ensure AI systems perform reliably in production environments?

Assume AI errors will be invisible at first. Many AI errors do not look like system failures. They look like confident, plausible answers that are just wrong enough to create downstream problems. Escalation paths need to be designed for uncertainty, not just known exceptions. Employees need to know when to stop and who resolves it.

Feedback loops need a named owner. If feedback is everyone's responsibility, it is nobody's responsibility.

9. Many organizations struggle with internal resistance or capability gaps. What practical steps can leaders take to prepare their workforce for AI-driven ways of working?

Stop treating every adoption issue as a training issue. Capability gaps and resistance are different problems.

For capability building, role-specific training beats broad AI literacy. Show people what good output looks like, what bad output looks like, when to challenge the tool, and where to escalate. And train managers first.

10. If you were advising an enterprise planning to scale AI across functions, what would be your top priorities to ensure long-term success?

Make one deployment work well before scaling. The organizations that scale AI successfully usually prove the operating model in one meaningful use case first.

Second, separate productivity from value capture. AI may make people faster. That is not the same as financial impact.

Third, govern the AI portfolio at the executive level. Someone needs to make tradeoffs, allocate resources, and stop work that is not delivering.

Radixweb Perspective on AI in Enterprise Transformation

Dr. Kendrick’s insights align closely with what we see in real enterprise AI implementations projects. At Radixweb, we work hands-on with organizations scaling AI systems across functions, and a consistent pattern emerges: over AI initiatives struggle at scale not because of model quality, but due to weak data foundations, fragmented workflows, and unclear ownership structures. Even well-built models fail when they are introduced into environments where systems are not integrated or decision pathways are not clearly defined. AI does not fail in isolation, it fails inside operating models that were never redesigned for it.From a delivery standpoint, the enterprises that succeed are the ones that treat AI as an engineering and architecture challenge, not just a deployment exercise. We consistently see that when organizations invest in end-to-end system and AI integration, workflow redesign, and continuous feedback loops across data and decision layers, AI performance stability improves by 40–50% in live environments.The key takeaway is simple: AI value is not unlocked at the model layer, but at the system level where data, processes, and decisions converge. Organizations that focus only on adoption miss this entirely, while those that rebuild how work flows through the enterprise are the ones turning AI into measurable business impact.If you’re planning your AI roadmap, schedule a consultation with our AI practitioners, who can help you align AI ambition with engineering execution.