Return-to-office arguments have sucked up oxygen while a bigger shift has slipped in through the side door. Managers are burning out, with global manager engagement falling from 27% to 22% in one year, and overall employee engagement at 20%, a drag Gallup estimates at 10 trillion dollars in lost productivity, or 9% of global GDP (Gallup, 2025). At the same time, 82% of leaders say they plan to use digital labor, meaning AI agents, to expand their workforce capacity in the next 12 to 18 months (Microsoft WorkLab, 2025). The managerial layer that once coordinated human effort now faces coordination of humans and agents at scale.
From hierarchy to a work chart
For a century, the org chart has been the firm’s operating system. It is good at reliable execution of known work in stable settings, yet it struggles with nonroutine, cross-functional problems and tends to harden status differences that stunt development (Lee and Edmondson, 2017). Self-managing organizations, defined as those that systematically decentralize authority and eliminate the formal manager–subordinate relationship, arose as a response to these limits (Lee and Edmondson, 2017). The idea was to push decisions to those closest to the work and let structure follow outcomes.
Agents bring this logic into the mainstream. Microsoft proposes replacing the fixed org chart with a dynamic work chart, a map of teams, people, and agents aligned to outcomes rather than job titles (Microsoft WorkLab, 2025). In companies it labels Frontier Firms, those that have deployed AI across the organization and actively use agents, 71% of workers say their company is thriving compared with 39% globally. That is a self-reported perception rather than a causal estimate, but it points to a pattern worth watching (Microsoft WorkLab, 2025). McKinsey describes a similar shift, estimating that 89% of organizations still run on an industrial-age paradigm, 9% use digital-age product or platform models, and only 1% operate as decentralized networks. Its view, based on client work, is that the agentic operating model replaces hierarchical delegation with flat networks of outcome-aligned teams coordinated by work charts (McKinsey, 2025).
What managers actually do in a human–agent organization
If agents handle more of the tasks, managers do not disappear. Their job changes. Microsoft reports that 46% of leaders already use agents to fully automate workflows or processes, and introduces new managerial roles that span human supervision and agent orchestration, sometimes called an agent boss (Microsoft WorkLab, 2025). McKinsey, drawing on client experience rather than peer-reviewed evidence, argues that a small human team of two to five people can supervise an agent factory of 50 to 100 specialized agents that run end-to-end processes like onboarding customers or closing the books (McKinsey, 2025). The implication is concrete. Management shifts from monitoring attendance and activity to specifying outcomes, governing data and prompts, setting guardrails, and auditing results.
That asks for a different kind of clarity. Traditional supervision rides on proximity. Agent orchestration rides on well-defined interfaces. The work chart forces leaders to name owners for outcomes, to publish the rules agents operate under, and to define escalation paths for exceptions. It also changes where human judgment matters most, moving it to boundary conditions, ethics, and cross-team trade-offs.
A caveat belongs here. Both Microsoft and McKinsey have commercial stakes in AI, and their estimates are based on a large proprietary survey and client experience respectively, not randomized trials. Treat their numbers as directional signals, not proofs of causation (Microsoft WorkLab, 2025; McKinsey, 2025).
Safety first, then speed
Organizations rarely fail to automate. They fail to learn safely at speed. Amy Edmondson’s field study of 51 teams showed that psychological safety, the shared belief that a team is safe for interpersonal risk taking, predicts learning behaviors. Those learning behaviors in turn link to performance, while team efficacy alone does not when safety is accounted for (Edmondson, 1999). The agent era raises the stakes. When a workflow is partially automated, speaking up about an odd model behavior or a silent failure is not optional. It is how harm is prevented and improvement compounds.
The self-management literature adds the other half of the design brief. Hierarchy is efficient for reliable execution of known tasks. Decentralization helps with novel, cross-boundary problems, but it needs explicit mechanisms to avoid drift and ambiguity (Lee and Edmondson, 2017). Human–agent organizations must blend both. Define crisp standards for routine agent-run work, including audit trails and rollback procedures. Then create local authority for teams to adjust prompts, data sources, and process logic within clear guardrails, and to escalate when the work crosses into new terrain.
A practical start
Do not redraw the whole company. Build one visible work chart and scale from there.
- Pick one end-to-end process with measurable outcomes, significant volume, and real customer impact. Examples include onboarding, claims adjudication, or financial close. Name a single accountable owner for the outcome.
- Stand up a small supervisor cell. Two to five humans test, deploy, and monitor a set of specialized agents that cover the process steps. Treat McKinsey’s ratio as a hypothesis to validate in your context, not a target (McKinsey, 2025).
- Publish the work chart. List the human roles, the agents, their inputs and outputs, the service levels, and the exception paths. Make it easy for adjacent teams to see interfaces and for auditors to trace decisions (Microsoft WorkLab, 2025).
- Engineer for learning. Instrument the process so that failures surface quickly. Run blameless reviews. Train managers in psychological safety, and measure it in the team, because learning behaviors are the path to performance (Edmondson, 1999).
- Mind the human energy. Track engagement, particularly among managers. The worldwide dip in manager engagement is a warning that role clarity and support lag behind expectations. Best-practice organizations show that design choices, not the economic cycle, explain much of the gap, with 79% manager engagement in top performers versus 22% globally (Gallup, 2025).
The near-term prize is not a flattened org chart for aesthetics. It is a shift in managerial attention from presence to outcomes, and from static structure to transparent interfaces. The survey data suggest momentum. Leaders plan to expand capacity with agents, and a meaningful share already run fully automated workflows (Microsoft WorkLab, 2025). The research on self-management and team learning offers the rules of the road. Push authority to where the work and the data live. Specify outcomes, not activities. Make it safe to say when the system is wrong. Firms that do this will find the work chart is not theory. It is the artifact of a healthier organization that learns faster than it automates.
Sources
- Microsoft WorkLab (2025). Work Trend Index 2025 — The Year the Frontier Firm Is Born. Reports that 82% of leaders plan to use AI agents within 12–18 months, 46% already automate workflows with agents, and introduces the outcome-based “work chart” concept along with survey evidence that workers in Frontier Firms are more likely to say their companies are thriving. https://www.microsoft.com/en-us/worklab/work-trend-index/2025-the-year-the-frontier-firm-is-born
- McKinsey (2025). The agentic organization: contours of the next paradigm for the AI era. Consultancy view that most firms remain in industrial-age models, and that small human teams can supervise large factories of AI agents within outcome-aligned networks coordinated by work charts. https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/the-agentic-organization-contours-of-the-next-paradigm-for-the-ai-era
- Lee and Edmondson (2017). Self-managing organizations: Exploring the limits of less-hierarchical organizing. Defines self-managing organizations and summarizes why hierarchy suits routine work while decentralization helps complex cross-boundary problems, informing the design of human–agent structures. https://www.sciencedirect.com/science/article/abs/pii/S0191308517300059
- Edmondson (1999). Psychological Safety and Learning Behavior in Work Teams. Shows that psychological safety predicts team learning behaviors which link to performance, a critical condition when deploying partially automated workflows. https://journals.sagepub.com/doi/10.2307/2666999
- Gallup (2025). State of the Global Workplace 2025. Finds global engagement at 20%, manager engagement dropping from 27% to 22% in a year, and estimates a 10 trillion dollar productivity cost, underscoring the managerial strain that human–agent models must address. https://www.gallup.com/workplace/659610/state-global-workplace.aspx