McKinsey: AI is changing work. Now it has to change the organization
Source: McKinsey & Company, AI is changing work. Now it has to change the organization, October 2026
Key takeaways
AI is changing the constraints around which organisations were designed. McKinsey argues that agentic AI changes three fundamental constraints that have shaped the modern organisation: human capacity, expertise and coordination.
Individual productivity is moving faster than enterprise value. 80% of respondents say AI has improved their individual productivity, but only 37% report positive EBIT impact and just 6% qualify as AI high performers.
Workflow redesign separates the leaders. Nearly three quarters of AI high performers report substantially redesigning workflows, compared with only one quarter of other organisations.
Only 11% of organisations have reached McKinsey’s reinvention horizon, where they rethink how work gets done and how value is created rather than simply deploying additional AI tools.
The operating model needs to move from an org chart towards an orchestration chart. Organisations can start with the outcome and then assemble the humans and agents required to deliver it.
AI changes jobs without necessarily eliminating them. Agents can take on research, drafting, coding, analysis, monitoring, summarisation, routing and coordination while the broader role remains.
Released capacity needs an explicit destination. AI productivity gains can support greater output, different work, lower cost or more sustainable workloads, but the organisation has to choose where the dividend goes.
Career architecture also needs redesign. If AI absorbs more coordination and entry level work, traditional management and apprenticeship pathways may no longer build expertise or progression in the same way.
Skills adjacency matters for workforce transitions. McKinsey argues that the most realistic moves into new work tend to draw on skills people already possess.
Personal readiness is significantly ahead of organisational readiness. 70% of respondents say they are personally ready for AI, while only 27% of leaders say their organisations are ready for the shifts required for an agentic future.
Trust is part of organisational readiness. McKinsey finds that trust that the organisation will support employees as AI changes work is associated with enterprise value across every stage of AI maturity.
AI reinvention requires an execution engine. McKinsey proposes an agentic mission factory that begins with a consequential business outcome and redesigns the human and agent workflow, technology, controls and organisational changes around it.
Executive summary
McKinsey’s October 2026 article argues that the first phase of enterprise AI focused primarily on putting powerful new tools into employees’ hands. Capturing the larger opportunity from agentic AI now requires something more fundamental: redesigning the organisation around what humans and agents can do together.
The starting point is that agentic AI changes three constraints that have shaped organisational design for more than a century: human capacity, expertise and coordination. Tasks, jobs, teams, management layers and hierarchies were built around those constraints. AI does not remove them completely, but it changes them enough that leaders can begin asking what the company can now do that was previously impossible.
Most organisations have not yet reached that point. McKinsey’s 2026 State of AI research shows a persistent gap between individual productivity and enterprise economics. 80% of respondents say AI has improved their individual productivity, but only 37% report positive EBIT impact and just 6% qualify as AI high performers. Nearly three quarters of those high performers report substantially redesigning workflows, compared with only one quarter of other organisations.
McKinsey’s AI readiness research reinforces the same point. Only 11% of organisations have reached the reinvention horizon, where companies move beyond individual tools and workflow automation to rethink roles, workflows and operating models around what AI now makes possible.
The article develops this argument through four connected shifts. The operating model moves from an org chart towards an orchestration chart, where humans and agents are assembled around outcomes. The talent system needs to become more dynamic as work inside roles changes. Organisational change has to move beyond adoption towards new behaviours, incentives and decision rights. Finally, companies need an execution engine capable of redesigning consequential work, building agents and controls, and changing the organisation at the same time.
The core message is that enterprise AI value depends less on how many tools or agents an organisation deploys and more on whether it is willing to redesign how work, roles, decisions and accountability operate around them.
Key stats and quotable claims
80% of respondents say AI has improved their individual productivity.
37% report positive EBIT impact from AI.
Only 6% qualify as AI high performers.
Nearly three quarters of AI high performers report substantially redesigning workflows, compared with roughly one quarter of other organisations.
Only 11% of organisations have reached McKinsey’s reinvention horizon.
McKinsey’s operating model research surveyed more than 700 executives and senior leaders.
In one product development organisation, traditional pods of eight to ten people shifted to hybrid teams of four to six people supported by agents.
Across McKinsey client work, coordinator roles have grown at roughly 1.5 to 2 times the rate of line roles over the previous five years.
70% of respondents say they are personally ready for AI.
Only 27% of leaders say their organisations are ready to make the shifts required for an agentic future.
One technology company using an AI powered internal talent marketplace generated more than half of its hires internally in a recent quarter.
McKinsey finds that trust in the organisation to support employees as AI changes work is associated with enterprise value across every stage of AI maturity.
Overall summary
McKinsey’s central argument is that AI is beginning to expose the limitations of organisational structures designed for a pre AI world. For more than a century, companies were built around constraints in human capacity, expertise and coordination. Specialist knowledge sat inside functions, information moved through management layers and jobs provided relatively stable packages of work around which companies could build learning, accountability and progression.
Agentic AI changes those assumptions. Agents can perform parts of work previously done by people, move information across workflows, prepare analysis, monitor activity and coordinate execution. Expertise can increasingly become available through systems rather than only through individual employees, while capacity can expand without adding people at the same rate.
The problem is that many organisations are placing these new capabilities inside an old organisational model. Employees receive AI tools, but functions, workflows, management layers, career structures and decision rights remain largely unchanged. McKinsey argues that this helps explain why individual productivity can rise without producing equivalent enterprise value.
The alternative is to begin with the outcome rather than the existing organisation. Leaders can ask what needs to be achieved, which humans and agents should work together to achieve it, which handoffs remain necessary and where judgement and accountability should sit. That change then flows into the talent system, career architecture, management model and mechanisms used to implement organisational change.
The result is a broader definition of AI transformation. It is not simply about automating more activities. It is about redesigning the organisation around a new mix of human and machine capacity.
Deep dive
AI changes the constraints that shaped the modern organisation
Agentic AI changes three fundamental constraints that have shaped the modern organization: human capacity, expertise, and coordination. For more than a century, the familiar building blocks of organizational design, from tasks and jobs to teams, management layers, and hierarchies, have reflected those constraints. AI doesn’t remove them entirely, but it alters them enough that leaders can ask a more consequential question: What can this company do that wasn’t possible before?
Most organizations are not yet asking, let alone answering, that question. AI adoption is scaling rapidly, and individuals are already seeing meaningful productivity gains, but enterprise value has not kept pace. In McKinsey’s 2026 state of AI survey, 80 percent of respondents say AI has improved their individual productivity, while only 37 percent of respondents report positive EBIT impact, and just 6 percent qualify as AI high performers.
Tellingly, nearly three quarters of those high performers report substantially redesigning workflows, compared with only one quarter of other organizations. This is the gap between using AI and redesigning the organization around it.
Recent McKinsey AI readiness research reinforces that gap. Just 11 percent of organizations have reached the “reinvention” horizon, where they rethink how work gets done and how value is created. Reimagining means going beyond individual AI tools and workflow automation to redesign roles, workflows, and operating models around what AI now makes possible.
As AI capabilities advance, the existing organization can become the constraint. Yet many transformation efforts address only parts of that system, workforce, organizational design, or change management, rather than considering how they need to change together.
McKinsey therefore focuses on four interconnected shifts: organizing work around outcomes, reshaping the talent system to reflect the changing mix of human and AI work, building the mechanisms that allow change to compound, and putting clear authority and accountability behind the transformation.
Operating model: From org chart to orchestration chart
For much of the modern corporation’s history, companies have organized people into functions and hierarchies, with work passing from one part of the organization to another. AI enables a different logic: Start with the outcome, then assemble the humans and agents needed to deliver it.
There are signs that this shift has already begun. New McKinsey research surveying more than 700 executives and senior leaders finds that companies furthest along in AI reinvention are more likely to organize cross functional teams around end to end products, customer journeys, and business processes, putting decisions and accountability closer to where value is created.
There is no single winning operating model structure. Rather, the common thread is organizing work end to end, reducing unnecessary handoffs, and moving decisions and resources faster when priorities change.
As agents take on meaningful parts of work, organizational design increasingly becomes a question of orchestration: which humans and agents should come together around an outcome, what each should do, and which handoffs and interfaces are still necessary. An “orchestration chart” can make those choices explicit.
In some cases, agents may do more than simply join an existing team. They may replace the need for coordination between people or functions altogether, changing the network through which work moves. These shifts also make risk mitigation an integral part of orchestration, particularly when organizations are moving quickly, as decisions about agent autonomy, access, oversight, and escalation become embedded in the design of work itself.
What hybrid teams can look like
No company offers a finished blueprint, but early examples provide signals of what these changes might look like. In one organization’s product development function, traditional pods of eight to ten people shifted to hybrid teams of four to six people supported by agents.
An orchestration chart of the new team shows which work has moved to agents, which human roles now span more of the product development process, which handoffs have disappeared, and where human judgment and accountability matter most.
The significant change is therefore not simply smaller teams. Work is redistributed across the team. Some handoffs disappear, some human roles become broader and agent execution sits alongside areas where people remain responsible for judgement and accountability.
Coordination and the role of meetings
That same shift may change coordination mechanisms such as meetings, which are essentially an organization’s way of transferring information, creating alignment, and making decisions. In a human agent organization, agents can increasingly track which decisions have been made, what work is still outstanding, and what needs attention next.
Meetings don’t disappear, but their purpose shifts away from the routine transfer of information toward building judgment, commitment, and trust.
The practical test of a redesigned operating model is whether the organization can make decisions, move information, and coordinate work with fewer layers and handoffs and less manual translation between functions.
Across McKinsey client work, “coordinator roles”, internal roles focused largely on moving decisions across the organization, such as project management, have grown at roughly 1.5 to 2.0 times the rate of line roles over the past five years, signalling how complex organizations have become.
AI creates a significant opportunity to reduce some of that complexity rather than add to it. But the flatter pyramid can enable value only if the lost coordination is replaced by better intelligence, clearer decision rights, and stronger human judgment.
Markers of operating model progress
McKinsey identifies several ways to tell whether the operating model is genuinely changing. Organisations should seek structural change rather than simply making the old model faster. Teams should increasingly own outcomes across functional lines, while agents take on more execution and coordination. Team size, composition and management layers should change materially where AI takes on meaningful parts of the work.
The new operating model also needs to be visible. Orchestration charts can show how humans and agents combine around outcomes, who does what and where human judgement and accountability remain essential. The measure of progress therefore becomes a change in the design of work itself rather than simply higher AI adoption.
Talent: Building a system around work that keeps changing
Organizing work around outcomes rather than inherited roles and functions has consequences for the entire talent system. Most companies still hire, develop, deploy, and advance people through relatively stable jobs and career paths. A human agent organization needs a talent system that can move people as the skills required by the work change.
To make that shift, organizations can focus on rebuilding the pathways through which people learn, advance, and develop expertise.
AI changes the relationship people have with their work. For decades, the job has been the basic unit through which companies have organized work and developed talent. It defined what work someone did, how they created value, how they learned, how they took pride in their output, and how they advanced.
A junior analyst learned by doing analysis. A designer improved by making more important design choices. A manager progressed by taking responsibility for larger teams and more complex work. The job has traditionally linked the work people do with how they create value, build expertise, find satisfaction, and advance.
AI alters those connections by completing parts of the work without necessarily replacing the job itself. Agents can take on tasks that once gave a role much of its substance: research, drafting, coding, analysis, monitoring, summarization, routing, and coordination. That creates a productivity dividend that allows time and capacity to be used differently.
Where does the productivity dividend go?
McKinsey’s Exhibit 1 illustrates the choices created when AI frees capacity inside a customer service role. Rather than assuming there is one inevitable workforce outcome, the organisation can use that capacity in several ways. It can perform the same work with fewer people, increase output with the existing workforce, move people into higher value customer work, or recombine activities into different roles.
While individual productivity gains are welcome, they don’t automatically translate into enterprise value. Capturing more value requires an explicit choice about where those gains will go. In a recent conversation with McKinsey, Stanford economist Erik Brynjolfsson argues that focusing primarily on workforce reduction misses the larger opportunity to reconfigure and rebundle work around what people can now do with AI.
The productivity dividend is therefore an organisational design question. The important issue is not only how much time AI saves, but what the enterprise deliberately does with the capacity that becomes available.
How AI changes the experience of work
Agentic capabilities also change how people experience the work. If the system writes the first draft, creates the image, summarizes the research, prepares the analysis, or recommends the decision, the human may feel less like the producer and more like the reviewer of someone else’s output.
In creative fields, that concern is already visible among artists, writers, designers, and actors confronting AI generated work. Similar questions will appear inside companies as AI moves deeper into professional, managerial, and technical roles.
As roles combine human judgment with a changing mix of AI enabled activities, companies will also need to manage cognitive load. Too little human involvement can lead people to become overly reliant on AI, while too much can leave them overloaded by an expanding mix of responsibilities.
The design challenge is therefore not simply to maximise agent activity. Organisations also need to determine the level and type of human involvement that produces the strongest outcome.
Career progression also changes
AI changes career progression because traditional advancement has often been tied to managing larger teams, coordinating more work and sitting higher in the hierarchy. If AI absorbs more status tracking, synthesis, routing, and coordination, some of the work that justified management layers will shrink, and middle management may become less of a default career destination.
Companies will need career paths that reward craft mastery, judgment, human agent orchestration, customer or client trust, and the ability to make high quality decisions with AI support. A senior individual contributor who can lead agents around a complex business outcome may create as much value as a traditional manager with a large team.
That creates an alternative view of progression. Seniority can increasingly come from expertise, judgement and the ability to orchestrate complex human and agent work rather than simply from the number of people someone manages.
The apprenticeship problem
While deep specialization will remain important, there is a trade off. If agents take on the entry level work through which people have historically built expertise, today’s productivity gain can become tomorrow’s judgment gap.
Recent research showing employment pressure among early career workers in highly AI exposed occupations makes the question of how apprenticeship is handled increasingly concrete. Companies will need to redesign how people build expertise as the underlying work changes, with managers spending less time coordinating activity and more time creating opportunities for practice, feedback, and increasingly consequential judgment.
If foundational work disappears, the organisation therefore needs to replace the learning that used to happen through performing it. Otherwise greater efficiency in the short term could weaken expertise in the longer term.
Skills adjacency and future work
Forecasting which jobs will grow or shrink is therefore not enough. New McKinsey Global Institute research finds that people’s ability to move into new work depends heavily on skill adjacency: the most viable moves tend to be into jobs that draw on skills people already have.
Some occupations offer many adjacent pathways, while others have fewer options or face credential barriers. For employers, the most salient questions revolve around which skills future workflows will require, who has a realistic path into that work, and where the organization will need to retrain, hire, or create new pathways.
This reframes internal mobility around realistic transitions rather than generic reskilling. The organisation needs to understand not only which skills will be needed, but whether employees have a credible path from the work they perform today into the work the future organisation requires.
Connecting skills data to real opportunities
Some organizations are beginning to connect those questions directly to internal mobility. One technology company uses an AI powered talent marketplace to map employees’ existing skills to personalized learning, stretch assignments, and open roles.
In a recent quarter, more than half of its hires came from inside the company. That approach connects skills data to actual opportunities rather than treating reskilling as training disconnected from where people can move next.
The distinction matters because training is not the same as transition. Employees need a credible destination for the capabilities they are being asked to build.
Protecting what should remain human
Companies are also setting principles for how employees will experience the transition rather than going case by case. One global life sciences company has stated that AI may support work but should not replace human creativity, expertise, autonomy, agency, or oversight.
Employees remain accountable for AI assisted work. Such principles don’t resolve every workforce decision, but they do establish boundaries before individual decisions about roles and automation arise.
McKinsey therefore suggests that organisations should make the productivity dividend explicit, map realistic paths into future work, create new routes through which employees can build expertise and be clear about which elements of work should remain human.
Turning AI activity into sustained organisational change
Redesigning workflows and talent systems leads to enterprise value only if people themselves begin working differently. Yet one of the paradoxes of the current AI moment is that widespread adoption can coexist with surprisingly little organizational change.
Employees may use AI every day while the same workflows, decision rights, incentives, and management practices remain intact. The challenge is to make new ways of working take hold across the organization.
McKinsey’s recent AI readiness research points to this divide: 70 percent of respondents say they are personally ready for AI, while only 27 percent of leaders say their organizations are ready to make the shifts required for an agentic future.
The gap is important because it suggests that employee willingness is not necessarily the main constraint. The organisation itself can become the slower moving part of the transition.
Trust at the centre of the transition
That preparedness gap puts trust at the center of the transition. People can embrace AI themselves while questioning whether their organization will manage the consequences well.
McKinsey data suggest that this is more than an employee sentiment issue. Trust that the organization will support people as AI changes their work is associated with greater enterprise value across every stage of AI maturity.
As roles, career paths, and expectations change, trust will increasingly be shaped by what employees see leaders do, including how openly workforce decisions are explained, whether people have realistic opportunities to adapt, and whether commitments about human oversight and agency survive difficult choices.
Trust therefore becomes part of organisational readiness rather than a separate culture initiative.
New behaviour, not simply new skills
Sustained change also requires people to work differently, not just learn new skills. Doing all the work oneself is not necessarily a sign of competence if AI can improve the outcome. Teams can test ideas faster and more cheaply with AI while being clear about where mistakes would carry serious consequences.
Contribution becomes less about completing a list of activities and more about owning the outcome. McKinsey’s readiness research finds that such organization wide behavior shifts are associated with enterprise value for companies in the reinvention horizon.
That means performance expectations also need to change. Employees cannot be trained to use AI while still being assessed as though the previous way of performing the work remains the ideal.
Embedding learning into work
Some organizations are beginning to build those behaviours into how work is learned and reinforced. One large financial institution has evolved its approach from broad AI training toward role based learning, peer coaching, dedicated development for managers, and repeated opportunities to build and demonstrate solutions in real work.
Internal champions, hackathons, weekly demos, and peer learning circles help successful practices spread across teams, and the number of AI solutions reaching production has continued to grow. The model embeds learning and experimentation into how teams work rather than treating AI adoption as a separate training effort.
McKinsey identifies four markers of sustained change: looking beyond adoption to whether workflows themselves are changing, reinforcing new behaviours through performance management and manager expectations, building trust throughout the transition and shortening the time it takes to convert new AI capabilities into new ways of working.
Enabling AI reinvention: Building the transformation engine
Reinvention on this scale can’t be carried out through disconnected AI experiments. Companies need a way to repeatedly redesign consequential work, build the technology and controls around it, and change the organization alongside it.
That starts with clear ownership. Because AI reinvention cuts across business, technology, talent, risk, and other organizational boundaries, no single existing function has a natural mandate to deliver it end to end.
Companies need a senior leader with the authority to drive the agenda across those boundaries and make the trade offs that reinvention requires. The title matters less than the levers attached to the role: the ability to move funding and scarce talent, shape strategic investment decisions, and resolve competing priorities.
Authority therefore needs to extend beyond coordinating stakeholders. The person accountable for reinvention needs the ability to make decisions across the boundaries the transformation crosses.
The agentic mission factory
Some companies are beginning to build dedicated execution capabilities for this type of work. One emerging model is what McKinsey calls an agentic mission factory.
Its job is to take a consequential business outcome, the mission, and redesign the work through a hybrid human agent workflow, build the required agents and controls, launch the new way of working, and create reusable assets for the next mission.
The accountable leader provides the direction and authority while the mission factory turns that mandate into working solutions. The factory also creates the AI operating system, which compounds institutional intelligence by creating learning loops as individual agents execute their tasks.
In this model, the factory starts with the mission, not the tools. A mission might be to approve and fund a good loan faster, resolve a customer issue in one interaction, reduce claims leakage, accelerate product development, or improve supply chain resilience.
The business owner brings the outcome, economics, constraints, and accountability. The factory brings the cross functional capability to redesign and build the work.
Building the workflow and controls together
That capability needs more than AI engineers. It brings together business, technology, product, design, data, risk, compliance, cybersecurity, legal, change, and HR from the start.
Controls are built into the workflow rather than added at the end, and product ownership is assigned at launch, not after the pilot. Evaluation, monitoring, permissions, autonomy, and life cycle management are part of the operating model.
The model builds on the digital factory, which used dedicated cross functional teams to change the business while the rest of the organization continued to run it. The agentic version extends that approach to redesigning workflows around humans and agents.
That requires new capabilities for building and governing agents, as well as the shared AI infrastructure, evaluation, permissions, and life cycle management needed to operate them at scale.
Organisational choices have to be made during redesign
The mission factory is also where the organizational changes described above come together. Redesigning a mission means deciding how humans and agents will work together, how people’s roles and development need to change, and how new ways of working will take hold across the organization.
Those choices need to be made as the work is redesigned, not after the technology is built. That is what turns a series of AI deployments into organizational reinvention.
McKinsey identifies four practical markers of progress. The accountable leader should have enough authority to move funding and scarce talent and resolve cross enterprise trade offs. Business owners, builders and control functions should work together from the start. Evaluation, monitoring, permissions and lifecycle ownership should be designed before agents go live. Finally, every completed mission should leave reusable agents, controls, components and decisions that shorten the path for the next one.
The leadership test
No single company can claim to be a finished human agent organization, but the direction is becoming clearer. AI expands the choices leaders can make about how teams are shaped, how hierarchies work, how jobs develop people, and how quickly the organization can change.
McKinsey suggests six questions for leaders:
Where would a different mix of humans, agents, and data create more value, and how should roles, accountability, and controls shift as a result?
As AI takes on more tasks, which parts of each role remain distinctively human, meaningful, and worth designing the job around?
Where is AI removing the work through which people have built expertise, and what new apprenticeship ladder will replace it?
How will people be developed, coached, and promoted to cultivate judgment in human agent workflows?
What new structure or capability will redesign workflows, build agents and controls, and change the organization as one coordinated effort?
What behaviors are we modeling as leaders, and would our teams conclude from watching us work that AI is genuinely changing how we lead?
These questions move the conversation away from the number of agents deployed and towards whether the organisation itself is changing.
Conclusion
If leaders can answer these questions, AI is no longer just an experiment. It is changing the organization’s ability to break the constraints of the pre AI age.
The companies that capture the most value from AI will not necessarily be those that deploy the most agents or automate the greatest number of tasks. They will be the ones that use AI to rethink how value is created, then make the organizational choices required to deliver it, from how work and talent are organized to how transformation is led and built.
The larger opportunity is not simply to make today’s organization work better, but to build an organization capable of doing what was previously out of reach.
Key takeaways
AI is beginning to alter the fundamental assumptions around which organisations were designed. Human capacity, expertise and coordination are becoming less fixed constraints, which opens the possibility of redesigning teams, workflows, management layers and decision structures around a different mix of humans and agents.
The important distinction is between using AI and redesigning the organisation around AI. Individual productivity gains are already common, but enterprise value remains much more limited. McKinsey’s high performers are distinguished by the extent to which they redesign workflows rather than simply accelerating existing work.
The operating model therefore needs to become more outcome oriented. Human and agent orchestration, fewer unnecessary handoffs, clearer decision rights and explicit accountability become more important than preserving the existing org chart.
The talent system must change with the work. Companies need to decide where released capacity goes, how employees move into future roles, how expertise is developed when foundational work is automated and how careers progress when management layers become less central.
Finally, reinvention needs a repeatable execution model. The agentic mission factory provides one possible structure by bringing business, technology, product, risk, talent and control functions together around a consequential outcome and redesigning the work, agents, controls and organisational changes as one effort.
The broader message is that AI transformation becomes organisational transformation once the company changes how work is structured, how talent develops, how decisions are made and how humans and agents combine around outcomes.
