Why C-suite leaders need to rethink organization design for the agentic era

The market is filled with bold predictions about the impact artificial intelligence will have on organization design. Headlines warn of workforce reductions, pressure to reinvent operations, and autonomous agents replacing entire layers of management.¹

In a rush to become “AI-native,” many organizations are making significant workforce and operating-model decisions before the value of AI has been proven in their organization. In 2026, leaders in some sectors have announced full replacement of functions with AI agents, removal of entire layers of the workforce, and broad requirements that every worker must use AI without always providing the necessary guidance on how or when to use it.² At the same time, other organizations are investing heavily in scarce human capabilities such as leadership, judgment, creativity, and relationship-building, paying a premium for talent they view as increasingly differentiated in an AI-enabled world.

The mixed signals are telling. These contrasting bets reinforce why leaders must be deliberate about how human and agentic work fit together. AI does not eliminate the need for clear roles, sound decision rights, healthy spans of control, and strong accountability. Leaders must first determine how work should be distributed between humans and agents, where accountability should sit, and how the organization can continue to support innovation, adaptability, and growth.

The good news is that many of the decades-old principles that guide effective organization design still apply. With deliberate redesign, your organization will be better positioned to capture AI’s benefits without defaulting to blunt headcount reductions that can erode customer experience and reduce employee morale.

Organization design is fundamental to realizing AI value

Many organizations are deploying “AI everywhere” for individual productivity, leveraging out-of-the-box LLMs across all functions, levels, and activities. But the next evolution of AI will be workflow-specific, with data-backed agentic tools that require a deep understanding of the work itself.

Before changing reporting lines, consolidating teams, or redefining roles, leaders need to understand how work is performed at a granular level. This means breaking work down into its core activities, decisions, accountabilities, and outcomes to identify where AI can create value and where human involvement remains essential.

Agents can increasingly support both routine and complex tasks, process information, and execute defined workflows. But many activities still rely on human decision-making informed by judgment, experience, contextual awareness, and the ability to navigate ambiguity. Both human accountability and human authority to override AI outputs will make or break the success of organizations in the years to come.

To clarify and document decision rights, the traditional RACI-style approach to defining who is responsible, accountable, consulted, and informed is no longer sufficient. Consider adding an “O” for Override authority: the designated human decision-maker who can reject or modify AI-generated outputs, apply context and judgment when recommendations are inappropriate, and authorize exceptions when needed. In many cases, this may also be the decision-maker marked as accountable for the outcome. In others, the Override authority may be from a control function, such as risk or legal, due to exceptions to defined AI models that follow regulatory rules.  

Override authority belongs to the human in the loop

An organizational chart has traditionally represented reporting lines, hierarchy of accountability, decision-making authority, and escalations. While AI agents can perform tasks, provide recommendations, and execute defined processes, they cannot be held accountable for business outcomes in the same way people can. That accountability must remain with human leaders.

For example, if a team member makes a mistake, the human leader takes the steps to work with them to ensure root causes are addressed. It becomes a learning moment. Similarly, a team leader is responsible for driving improved team performance. The same logic should apply when an AI agent is involved. Leaders should remain accountable for performance, outcomes, and continuous improvement, even if the work is performed by an agent.

If organizations choose to represent agents on the org chart, they should do so only when they can clearly establish who the agent supports, who governs it, and who remains accountable for its outcomes.

At a minimum, the agent should:

1. Operate within a single defined team or function so it’s clear who it should report to even if the agent’s work contributes to a broader process (e.g., Invoice Review Agent reporting into accounts payable).
2. Have a single human leader accountable for the agent’s performance and ongoing improvement, working in partnership with IT and/or data teams for technical changes as needed.
3. Support a defined set of outcomes with clear human accountability for results and Override authority.

For many organizations, this means some agents may be on the organization chart. For example, an Invoice Review Agent embedded within Accounts Payable supports a defined team, process, and set of outcomes. In contrast, a personal AI agent used by an employee to analyze reports or validate hypotheses would not be represented because it is a productivity tool that supports an individual, rather than an organizational role with defined accountability and governance.

A useful parallel is the contingent workforce. Organizations that effectively integrate contractors, consultants, and other external resources may place some contractors on the org chart if they support a specific team and set of outcomes. Other external resources, such as consulting project teams, do not appear, because they serve multiple teams and are accountable to multiple stakeholders. In either case, external resources are not treated as true owners of outcomes. They are contributors who help execute work, while permanent employees retain accountability, provide oversight, and exercise judgment. AI agents should be approached in a similar way.

Core organization design metrics still matter

Spans of control, layers, and organizational shape have long served as indicators of organizational health. Despite the excitement surrounding AI-enabled organizations, the fundamentals of effective organization design remain.

In many cases, leaders can apply most of the same top-down organization design principles they use today. The difference lies in how those principles are interpreted in an environment where humans and agents increasingly work together.

Consider span of control. Traditional models often viewed wider spans as a sign of efficiency.³ In the agentic age, AI assumes more routine execution. Managers spend less time overseeing tasks and more time coaching employees through judgment-intensive work. At the same time, a manager may oversee an increasing number of outcomes delivered through AI agents. As we highlighted earlier, managing AI performance should become a core part of the manager’s job.

In this context, span of control may be better measured by the number of distinct outcomes a leader oversees, known as “span of outcomes.” This allows for consideration of the complexity of a leader who manages humans, team-specific agents, and the outputs of agents that serve multiple teams. Similar to span of control, “span of outcomes” should decrease where the outcomes a leader is accountable for are more complex and business critical, and should increase when outcomes are less complex or business critical.

Organizational shape may also evolve. As agents augment or perform work once handled by transactional or administrative roles, layers decrease and the shape begins to resemble, in many cases, an inverse pyramid. Organizations will see the number of employees focused on repetitive execution decrease in comparison to those focused on oversight, exception management, decision-making, and specialized expertise. This could also create stronger career paths for highly skilled individual contributors at all levels. However, organizations must intentionally prioritize the development of junior employees so that a healthy succession pipeline remains in place. Furthermore, we expect this shift to allow for the shape to adjust to the work at the micro or functional level, not just fit the macro or enterprise view.  

The appropriate balance between humans and AI will vary by function, risk profile, and strategic decisions around where human work provides outsized value to customers. Finance, customer service, and internal support functions may require very different models.

The role of managers in the age of AI

As organizations seek efficiency gains, some are exploring flatter models that rely on senior executives, frontline employees, and AI agents. These same organizations misguidedly categorize the work of people leaders as “middle managers,” with the connotation that their work is repetitive, replicable, and judgment-free. Some assume technology can remove management layers while preserving performance. In practice, this approach risks creating a critical gap between those setting direction and those executing the work.

• Executives bring strategic perspectives, but they are often removed from day-to-day operational realities.
• Frontline employees work closest to customers, processes, and operational challenges, but may not yet have the experience required to make complex decisions or navigate difficult trade-offs.
• Managers bridge this divide by translating strategy into action, providing oversight, and ensuring accountability for results.

As AI becomes more embedded in workflows, management’s bridging role becomes more important.  

Managers sit at the intersection of proximity to the work and judgment informed by experience. They are often well-positioned to assess whether AI-generated recommendations make sense in context, identify exceptions that require human intervention, and balance competing priorities when decisions are not straightforward. Rather than simply reviewing outputs, they can help govern how AI is used so it enhances decision-making without introducing new risk.

In a future where people and agents work side by side, managers play a critical role as both coaches and executors. Their value to an organization increases relative to their cost, as their time is spent increasingly on advancing strategy and market differentiation, rather than administrative tasks.

How Deloitte can help

Deloitte helps organizations establish clear principles for human-agent teaming, including decision rights, escalation paths, accountability, and risk thresholds. We also help business leaders reassess spans of control and organizational shape by function, team, and work type, to maintain efficiency without added risk.

Realizing AI’s potential requires focusing on what has always defined a healthy organization: work is thoughtfully designed, agents support the right activities, managers apply judgment, and accountability remains clear. Those who do this well will be the winners of the next few years of rapid change.