From org charts to work charts: organising for AI
McKinsey argues the next advantage comes from redesigning the organisation around AI agents. The unit of work shifts from the task to the outcome.
TL;DR
- McKinsey contends that competitive advantage will come from redesigning the enterprise around AI agents.
- The traditional org chart, built on hierarchical delegation, gives way to a work chart built on tasks and outcomes.
- People move "above the loop," supervising agents and managing trade-offs.
In The agentic organization, McKinsey makes a claim that is easy to nod along to and hard to act on: the advantage in the AI era will not come from using individual tools, but from redesigning the enterprise so that humans and AI agents create value together.
The five-pillar redesign
The argument rests on a structural shift. Traditional organisation charts encode hierarchical delegation. McKinsey suggests these will pivot toward agentic networks, or work charts, organised around the exchange of tasks and outcomes rather than reporting lines. In that model, employees move from performing tasks to orchestrating them: setting goals, supervising agents, and managing exceptions and trade-offs. The firm frames the redesign across five pillars.
| Pillar | The shift |
|---|---|
| Business model | AI-native channels, hyper-personalisation, proprietary data |
| Operating model | AI-first workflows and small agentic teams |
| Governance | Real-time, embedded controls, with humans accountable |
| Workforce and culture | From doing tasks to orchestrating outcomes |
| Technology and data | Platforms that let agents operate at scale |
One line deserves emphasis. McKinsey argues that governance can no longer be a periodic, paper-heavy exercise. As agents act continuously, oversight has to become real-time, data-driven, and embedded, with humans holding final accountability.
Sequence matters
We agree with the direction and would add a caution about sequence. Most organisations cannot leap to a work chart, and they should not try. The agentic operating model presumes clean data, well-defined workflows, and real-time governance, none of which exist by default. The realistic path is to redesign one workflow end to end, prove the human-plus-agent pattern there, and let the operating-model change follow the evidence.
The hardest change is managerial. A manager whose status came from owning a large team faces a different role when the team includes agents and the job is orchestration. That transition is where most of these redesigns will succeed or stall.
Practical steps
- Map work by the outcomes it produces, and identify where agents can own steps under human supervision.
- Redesign governance for continuous operation before scaling agents, well ahead of any incident that would force the issue.
- Reskill managers for orchestration. The scarce capability is judgement over a whole system, and that is what to hire and develop for.
Related reading
- Closing the gen AI paradox: a leadership agenda
- Rethink management and talent for agentic AI (McKinsey)
Source: The agentic organization: A new operating model for AI, McKinsey.