In this second article in the series on AI, organizational design and leadership capability, Ann Harten and Craig Pattison FCIPD continue to share thoughts on another impact to organizational design.
AI is not simply changing the work family-owned and privately held businesses do. It is changing the shape through which work is organized, judgement is developed and leadership capability is passed on.
In our first article, The Pyramid Under Pressure, we argued that the traditional organizational pyramid was more than a reporting structure. It was also a development model. People entered near the base, learned the business through experience, built judgement over time, and progressed as their capability matured.
That model is now under pressure. But what follows may not be a flatter pyramid. Increasingly, the emerging structure resembles a diamond.
The base narrows as automation and AI agents absorb routine, rules-based and high-volume work. The top remains constrained because most businesses are not adding senior leaders in proportion to the complexity they face. The middle becomes more consequential because it must make sense of AI-generated outputs, coordinate across functions, manage exceptions and translate strategic intent into operational reality.
But, the more important shift is this: the future middle manager may not only manage people. They may manage the interface between people, AI agents, operating systems and owner intent. Managing output and governing systems are not the same thing. If AI is generating recommendations, routing exceptions or initiating workflows, owners and boards need to know where human accountability sits — who reviews the recommendation, who owns the exception, and who is responsible when the system is technically correct but commercially wrong.
For family-owned and privately held businesses, this matters because judgement is often held close. It sits in owners, founders, long-serving managers and trusted operators who understand the history behind customer relationships, supplier dependencies, margin decisions and reputation risk. Much of that knowledge is not written down. It has been learned through proximity, experience and judgement passed from person to person.
The question is whether AI strengthens that concentration of knowledge, weakens it, or quietly disrupts it in ways owners and boards have not yet examined.
The risk is not simply that AI creates a diamond-shaped organization. The risk is that it creates one faster than the business can develop the managers, judgement and accountability needed to make it work.
The diamond may be emerging by accident
In many businesses, diamonds are formed as the result of a series of decisions that make sense individually but are not being governed collectively.
A business case is approved to automate a reporting process. Another introduces AI-supported customer service. A third deploys intelligent workflow tools in finance, operations or supply chain. Each initiative may improve productivity, reduce cost or increase speed.
Together, those decisions can reshape the organization in ways no single business case has properly assessed.
The base narrows, but no one asks what learning previously happened through those roles. The middle absorbs more complexity, but no one redesigns the capability required. Owners and boards receive richer, faster information, but governance rhythms and decision processes often remain largely unchanged.
The result is an organization that is structurally becoming a diamond while continuing to manage itself as though it were still a pyramid.
In family and private businesses, the warning signs may be quieter than in listed companies. The business may still perform because experienced people compensate. The AI business case may still look positive. But beneath that, the organization may be becoming more dependent on fewer people who understand how the business really works. AI may capture data, summarize activity and identify patterns. But it does not automatically preserve institutional memory, commercial instinct or the values that shape long-term owner stewardship.
A business can become more automated and less resilient at the same time.
The middle layer is the pressure point
A common assumption is that AI will reduce management layers. In some cases, it will. But that misses the more important point. AI does not remove the need for management. It raises the standard of management required.
In the emerging diamond, the middle layer has three critical jobs.
First, it must translate owner, board or executive intent into operational direction for teams that may include both people and AI-enabled systems. That requires understanding what the technology can do, where it may fail and how its outputs align with the real priorities of the business.
Second, it must interpret AI-generated outputs and manage exceptions. AI can surface correlations, anomalies and recommendations. But it cannot reliably determine which are commercially significant, which are misleading and which require human challenge. As routine work becomes more automated, the easy cases are handled by systems. The difficult ones land with managers.
Third, it must coordinate across functions and oversee work carried out by both people and AI-enabled agents. AI makes interdependencies more visible, but it does not resolve them. A decision in supply chain may affect customer service, margin, working capital, staffing and reputation. Someone still needs to hold the whole picture.
These are not narrow technical questions. They are management questions.
The capability gap should not be underestimated
This is not unique to family-owned or privately held businesses. Across many types of organization, middle management has been under-invested in for years. Selection has often been based on technical competence, tenure or individual performance rather than the capabilities now required: judgement, interpretation, cross-functional coordination, ambiguity tolerance and the ability to oversee work carried out by both humans and AI-enabled systems.
In the traditional pyramid, that weakness could sometimes be absorbed by the depth of the organization. Strong employees below the management layer could compensate for poor supervision. Informal networks moved knowledge around. Capable people often carried weaker structures because there were enough people close enough to the work to see what was really happening.
As the base narrows, that margin for error reduces. There may be fewer people beneath the middle layer learning by doing, spotting anomalies or compensating for gaps in management judgement. AI does not only compress layers. It can also compress learning. If junior roles disappear or become too narrow too quickly, people may no longer gain the exposure through which judgement, culture and commercial instinct were historically developed.
The middle layer becomes more consequential at precisely the point when many businesses have not properly selected, equipped or developed it for the role they now need to play. For owners and boards, this is a direct governance concern. A diamond-shaped organization cannot rely on pyramid-era assumptions about management capability.
Could AI remove the middle altogether?
There is a fair challenge to the diamond argument. In some businesses, AI may appear to make the middle less necessary. Owners and executives may receive AI-curated intelligence directly. Workflow tools may allocate tasks automatically. Dashboards may surface exceptions without requiring the same chain of human interpretation.
That may reduce some management roles — and may remove layers that added limited value. But it does not remove the need for judgement. It simply moves the risk.
If the middle layer is removed without replacing its interpretive, coordination and challenge functions, the business becomes more dependent on systems and a smaller number of senior decision-makers. Owners may see more information, faster, but with fewer people able to test whether that information reflects operational reality.
The real question is not whether every business settles permanently into a diamond. It is whether the organization has deliberately decided where judgement, accountability and institutional memory will sit as AI changes the flow of work.
If the middle disappears without those functions being redesigned, the business may not have become more intelligent. It may have hollowed out one of the places where understanding used to be built.
What owners and boards should do differently
The practical challenge is not to resist AI. It is to govern the organizational shape that AI is helping to create. Before approving AI-led redesign, owners and boards should test whether management can answer five questions.
First: where is the base narrowing? Which roles, tasks or processes are being automated or redesigned? What work is disappearing, and what learning did that work previously provide?
Second: what will replace the lost learning route? If people no longer develop judgement through entry-level work, how will they gain exposure to customers, suppliers, trade-offs, operational constraints and commercial consequences? How will culture and judgement be transferred if the apprenticeship effect of moving through the business is weakened?
Third: what is the new role of the middle layer? Which middle-management roles are now carrying new accountabilities because of AI-enabled work? Have those roles, authorities, decision rights and control points been formally redesigned, or have the responsibilities simply been absorbed into existing jobs?
Fourth: is the middle layer capable enough for the new structure? Have managers been selected and developed for judgement, ambiguity, cross-functional coordination and AI literacy? Or are businesses assuming that people promoted under the old model will automatically be effective in the new one?
Fifth: does AI reduce key-person dependency or increase it? Does the new operating model spread judgement more effectively, or does it concentrate understanding in fewer experienced people? If one or two key individuals left, would the business still understand why important decisions are made the way they are?
If owners and boards cannot answer these questions clearly, the issue is not whether they have an AI strategy. It is whether they understand the operating model they are building.
The diamond needs deliberate design
The emerging diamond may be a better structure for some businesses. It may remove low-value work, improve speed and allow talented people to operate with better information and stronger tools. But it will not work simply because it is leaner.
A narrower base, a constrained top and a more consequential middle require deliberate design. The middle layer needs clearer accountability, stronger capability and better support. Junior and early-career employees need alternative ways to develop judgement. Owners and boards need better visibility of how work, decisions and institutional knowledge are moving through the business.
For family and private businesses, the test is not only whether AI improves productivity. It is whether the business remains intelligible to the next generation of leaders.
Systems may process work, but they do not inherit stewardship. They do not understand family intent, customer history, supplier trust, margin instinct or the quiet judgement built through years of operating experience.
The diamond may be emerging. But unless owners and boards deliberately design the capability inside it, they may not be building the organization of the future. They may simply be removing the layers that once helped the business understand itself.
About the Authors
Ann M. Harten is the founder of C-Suite Advisor, an AI strategy advisory practice working with corporate enterprises on governance, operating model design and leadership capability in the age of AI. She brings 25 years of executive experience spanning CIO, CHRO and VP of Global AI Transformation roles, and completed MIT’s Executive Education program in Deploying AI for Strategic Impact. Ann is an NACD certified board director and serves on three corporate boards.
Craig Pattison FCIPD is an executive coach, fractional Chief People Officer and strategic advisor working with CEOs and senior leaders at the intersection of people, technology and organizational change. A former Chief People Officer and Chief Technology Officer, he has led enterprise people, technology and transformation agendas in PLC, private-equity-backed and regulated environments. He is the creator of the Root-to-Result® coaching model and writes on leadership, capability, operating models and execution.
