The state you bring to the decision Why AI readiness starts with the leader in the room, not the technology on the table

Sixty-five percent of organizations are now using generative AI in at least one business function, nearly double the share reported just one year earlier (McKinsey, 2024). That number sounds like momentum. It is also a warning. Gartner estimates that through 2025, more than 85 percent of AI projects will deliver below-expected returns, with poor problem framing and inadequate change management cited as the leading causes of failure (Gartner, 2024). Speed of adoption and quality of implementation are not the same measure, and right now the gap between them is widening fast. J. Caresse and Company convened a global roundtable of practitioners spanning robotics and automation, enterprise software go-to-market, change management, and software engineering leadership. Participants represented operations across North America, South America, Europe, Africa, and South Asia. What they surfaced was not a technology problem. It was a human one. The forces shaping AI outcomes in their organizations, leader burnout, competitive pressure, workforce fear, governance gaps, and the longer curve of physical AI, were already in motion before any system was selected or any workflow was changed. What follows is an honest account of what they found.

The decision-maker is the first variable in AI readiness

The roundtable opened with a claim that stopped the conversation: burnout among senior leaders is not an occasional condition. It is a near-universal one, and it is materially distorting AI implementation decisions. One automation consultant with more than two decades of experience across the Americas estimated that 99 percent of the organizational leaders he encounters are operating in a depleted state before any technology conversation begins. That depletion, the group agreed, is not incidental to poor AI decisions. It is a primary driver of them. Deloitte's 2023 Global Human Capital Trends report found that 77 percent of executives report experiencing burnout at their current jobs, with decision quality and strategic clarity cited as the first casualties of prolonged stress (Deloitte, 2023). Harvard Business Review research reinforces this: cognitive overload reduces a leader's ability to assess second-order consequences by as much as 40 percent, making complex technology decisions significantly more prone to error under conditions of chronic stress (HBR, 2022). A change management practitioner in the group described the downstream consequences directly. When leaders are depleted, stakeholder relationships deteriorate, trust erodes, and organizations enter what she called a state of relationship treachery: large decisions made in rooms where credibility has already been spent. The practical implication the group landed on was unambiguous. The first intervention in any AI initiative is not a technology selection. It is a pause. Before scoping automation, before assessing vendors, before building a business case, the organization must establish a clearly stated problem. As one participant summarized: most requests framed as 'we want AI' do not require technology first. They require clarity. That clarity is precisely what a burned-out leader cannot reliably produce. Organizational AI readiness, in this framing, is not a systems audit. It begins with an honest assessment of the cognitive state of the people making the calls.

Competitive pressure is automating the wrong things

Alongside burnout, participants identified a second force driving premature AI adoption: the pressure to scale revenue, pipeline, and productivity in competitive markets. A go-to-market leader covering enterprise software across Africa, Europe, and Asia Pacific described the dynamic precisely. In software sales environments, the belief that AI is the key to the scale we are chasing is nearly universal, and it produces a willingness to implement before validation is complete. The pattern that emerged across multiple participants: testing in production and automating processes that have not been confirmed to work at the manual level first. This is the operational equivalent of adding a faster engine to a car with a broken transmission. BCG research published in 2024 found that organizations that moved to scale AI before completing a controlled pilot phase were 2.5 times more likely to report implementation failure than those that validated outcomes in a limited environment first (BCG, 2024). A robotics and automation specialist in the group made the point bluntly: automating a broken process does not fix it. It produces better bad results at higher speed and cost. PwC's 2024 AI Business Predictions report found that 54 percent of executives who described their AI rollouts as disappointing attributed the shortfall to insufficient process validation before deployment (PwC, 2024). This is where process thinking before AI becomes a concrete operational discipline, not a slogan. The group converged on a distinction a software engineering leader described as separating what has already been proved from what still requires controlled experimentation. Proven processes can move fast. Unproven ones require a baseline measurement, a narrow pilot, and confirmed results before any broader rollout. The practical skill, as one participant framed it, is knowing which category a given process belongs in before the pressure to move makes that judgment harder to form. That categorization is itself a form of systems thinking: seeing the work as a set of interdependent processes rather than a list of tasks to be accelerated.

AI behavioral change starts on the floor, not in the announcement

Fear of job displacement is the dominant emotional response to automation among production-floor workers and established functional teams. This is not a perception problem to be managed with better messaging. It is a structural risk that shapes implementation outcomes. A 2023 EY Workforce Survey found that 65 percent of frontline employees identify job loss as their primary concern about AI adoption, and that organizations which failed to address this fear directly experienced implementation resistance rates nearly three times higher than those that engaged workers proactively (EY, 2023). The automation consultant in the group described his standard approach for production-floor engagements. Before walking the floor and identifying placement sites for automation hardware, he brings the maintenance and operations teams into a meeting room and demonstrates the technology on its own terms. That single step, he noted, changes the dynamic entirely. Workers who understand what a system does and who are asked where it would help their own work stop perceiving the consultant as an external threat and begin contributing domain knowledge that improves the implementation itself. MIT Sloan Management Review research confirms the pattern: successful AI transformations share a common early step, bringing workers into the conversation before deployment, in a neutral setting, and allowing them to identify how the technology intersects with their own work (MIT Sloan, 2023). Participants were explicit that this approach is not a courtesy. It is risk management. An implementation that proceeds without worker inclusion must overcome active resistance after deployment, at far greater cost than early engagement would have required. KPMG's 2024 CEO Outlook found that organizations ranking in the top quartile for AI adoption success were significantly more likely to have invested in change management and workforce education programs before rollout, rather than after (KPMG, 2024). Changing how employees think about work, specifically shifting them from task thinking toward understanding how their role fits within a larger process, is not a soft initiative. It is the work that determines whether AI lands on a process that can hold it.

Governance is the operating system for AI decisions

Governance was the term that generated the most debate in the roundtable, with participants acknowledging its necessity while resisting its bureaucratic connotations. A go-to-market transformation leader described governance, at least in enterprise software environments, as an obstacle to pipeline. That framing reflects the speed pressures described above. But the group ultimately agreed on a more functional definition: governance is the set of recurring structures that give stakeholders a place where decisions are made, debates happen, and no significant change proceeds without appropriate involvement. A change management practitioner in the group described this as building an organizational operating system around the change. The goal is not to add process steps. It is to reduce the isolation that amplifies burnout and the ad hoc decision-making that produces implementation regret. Deloitte's Human Capital Trends research identifies the absence of structured decision forums as one of the primary contributors to transformation fatigue, noting that leaders who lack a defined governance process report higher stress levels and lower confidence in their AI-related decisions than those operating within a formal framework (Deloitte, 2024). McKinsey found that organizations with clearly defined AI governance structures were 1.7 times more likely to report that their AI investments had met or exceeded financial expectations (McKinsey, 2023). In practice, participants described governance expanding incrementally as demonstrated reliability grows. One described the approach his organization uses: the AI cannot edit the CRM independently. That constraint is a hard limit maintained regardless of how capable the model becomes in adjacent tasks. As reliability is demonstrated in lower-risk domains, the field in which the AI operates is gradually widened. This incremental expansion model mirrors Gartner's guidance: establish non-negotiable constraint zones for high-consequence actions and expand AI autonomy systematically as trust is earned through measured performance (Gartner, 2023). The participants acknowledged that industry-level standards remain unsettled. The leaders of the major AI development organizations are still establishing what responsible operational limits look like. That uncertainty makes internal governance structures more important, not less.

Physical AI and the longer curve operations leaders are not yet planning for

The roundtable's most forward-looking discussion emerged when a robotics engineer introduced the distinction between software AI adoption, which is already underway at scale, and physical or embodied AI, where machine intelligence must coordinate dozens of sensors within a hardware body in real time. The hardware to build such systems exists in principle: servos and actuators capable of mimicking human range of motion are commercially available, and humanoid platforms with 24 to 30 degrees of freedom are in active development. The unsolved problem is not mechanical. It is computational. Coordinating 50 to 60 sensors simultaneously within a physical body, in real time, at the precision required for production-floor tasks, exceeds the current capacity of available AI engines. Gartner's 2024 Emerging Technology Hype Cycle identifies humanoid robots and physical AI as technologies still two to five years from mainstream enterprise adoption, with integration complexity and compute requirements cited as the primary barriers (Gartner, 2024). McKinsey's 2024 Technology Trends report notes that while the global market for industrial robots is expected to reach 70 billion dollars by 2030, the share of that market represented by AI-native physical systems remains below 5 percent today, with the majority of deployments still relying on pre-programmed motion sequences rather than adaptive intelligence (McKinsey, 2024). MIT Sloan research identifies quantum computing as the most plausible pathway to the real-time parallel processing that physical AI coordination requires, though commercially viable systems at the required scale remain a research-stage proposition (MIT Sloan, 2024). For operations leaders, this trajectory has direct planning implications. The adoption curve for physical AI is longer, more capital-intensive, and more technically complex than the software AI curve that most enterprise transformation frameworks currently address. A robotics specialist in the group described the current moment as a data collection phase: workers in production environments are being equipped with sensors and motion-capture tools so that their skilled movements can be translated into training data for future robotic systems. That data, once sufficient, will form the basis of the machine learning models physical AI requires. The roundtable framed this not as a distant abstraction but as an active management challenge. Operations leaders who are building AI strategy today will need to sequence investment across both the software and hardware horizons simultaneously, and the leaders best positioned to do that are those who have moved their teams from task thinking to systems thinking before either horizon arrives.

What comes next

The forces examined here, leader depletion, competitive pressure, workforce fear, governance gaps, and the physical AI frontier, are not temporary conditions to be addressed sequentially. They are structural features of the environment in which AI implementation decisions will be made for the foreseeable future. The organizations that navigate this environment most effectively will not be those that move fastest or those that move most carefully. They will be those that can accurately read which mode a given decision requires, and that have built the internal conditions to make that read correctly under pressure. That means a clearly stated problem before any technology is selected. A workforce brought into the conversation before deployment. A governance structure that creates recurring forums for decision and debate. And a categorization discipline that separates what is ready to scale from what still requires controlled experimentation. J. Caresse and Company will continue to examine these themes in subsequent roundtable sessions, with particular attention to how organizations across different geographies and industries are resolving the tension between measurement discipline and market speed. The near-term work is not primarily technological. It is behavioral and structural: moving people from task thinking to systems thinking, building processes that can hold AI when it arrives, and protecting enough cognitive capacity in the leaders making the calls to ensure that when the decision comes, the state they bring to it is one that serves the organization well.