AI is bringing the largest organizational paradigm shift since the industrial and digital revolutions (see sidebar, “The evolution of operating models”). This new paradigm unites humans and AI agents—both virtual and physical—to work side by side at scale at near-zero marginal cost. We call it the agentic organization.
McKinsey’s experience working with early adopters indicates that AI agents can unlock significant value. Organizations are beginning to deploy virtual AI agents along a spectrum of increasing complexity: from simple tools that augment existing activities to end-to-end workflow automation to entire “AI-first” agentic systems. In parallel, physical AI agents are emerging. Companies are making strides in developing “bodies” for AI, such as smart devices, drones, self-driving vehicles, and early attempts at humanoid robots. These machines allow AI to interface with the physical world.
The agentic organization will be built around five pillars of the enterprise: business model; operating model; governance; workforce, people, and culture; and technology and data (Exhibit 1). Imagine, for instance, the bank of tomorrow: When a customer wants to buy a house, a personal AI concierge activates a series of agentic workflows to serve the buyer. A real estate AI agent suggests properties, while a mortgage underwriting agent tailors offers based on the customer’s financial profile. Compliance agents ensure that the deal adheres to bank policies, and a contracting agent finalizes agreements before another agent fulfills the loan. All these workflows are overseen by an agentic team of human supervisors, mortgage experts, and AI-empowered frontline employees. In some cases, the bank could even extend its AI-powered services into furnishing, renovations, energy upgrades, and more. The bank becomes a network of agentic teams—an agentic organization.
In this article, we share early signals from our work with pioneering companies, insights from tech leaders and investors, and the questions executives are asking us. The agentic organization paradigm will undoubtedly evolve, but today’s leaders cannot wait for perfect clarity. In this article, we point leaders to where they can act now to shape the new era—refining their operating models to create more value and rewiring for an AI-first approach—instead of waiting to be shaped by it.
Five pillars of the agentic organization
1. Business model
In the agentic era, companies will gain a competitive advantage by getting closer to customers via AI channels to offer real-time hyperpersonalization, streamlining processes to become AI-first, and building a walled garden of proprietary data as their superpower. AI-native start-ups and agentic companies can potentially disrupt industries, with a fundamentally different level of productivity (revenue per employee), cost decoupled from growth, and greater speed to market and innovation.
AI-native channels enable hyperpersonalization
Consumers are already bypassing shops, apps, and search engines in favor of AI-native interfaces such as ChatGPT. In the future, every consumer could have a low-cost AI personal assistant. One European utility provider has rolled out a multimodal AI assistant to its three million customers. It significantly reduced average handling times, boosted customer satisfaction, improved response speed, and resolved more calls without a human. These assistants won’t just respond; they are personal concierges that will negotiate with other agents 24/7, continuously learning from user behavior and market signals to generate ever-evolving, hyperpersonalized products. This also unlocks new opportunities for the ecosystem economy, in which companies that own customer contact can grow by meeting various customer needs beyond their traditional business model and industry boundaries.
AI-first workflows drive marginal costs toward the cost of compute
Banks already run mortgage and compliance processes with agent squads. Insurers are reinventing claims and underwriting, while reimagining themselves as AI-native. Telcos are using agents in customer service and beyond. One global bank’s “agent factory” manages know-your-customer processes with ten agent squads, which has helped achieve a substantial positive impact on the quality and consistency of output. Another bank has used humans to oversee squads of AI agents in modernizing its legacy core systems, enabling up to 50 percent reductions in time and effort. This is not automation as usual on top of existing processes—it’s a redesign of end-to-end processes with humans “above the loop” for strategic oversight, with potential to bring the marginal cost toward the cost of compute. Going forward, most, if not all, processes can be reimagined as AI-first, with humans and traditional IT systems selectively introduced back in the loop or above the loop.
Proprietary data becomes a key differentiator
If today’s AI is “an intern with the internet in its pocket,” tomorrow’s edge will come from the walled data gardens that the public internet doesn’t offer. Companies can outperform their competition by continuously capturing and refining unique, consented, proprietary data—such as streams of customer behavior, product usage, and sensor data—and converting them into differentiating personalized products and processes. AI can also help by accelerating the build-up of data foundations and data products, as well as data-quality improvements.
2. Operating model
In the agentic era, how organizations are built and operate will evolve as much as the products or services they deliver. Work and workflows will be reimagined as AI-first, and operating models will evolve to flat networks of empowered, outcome-aligned agentic teams.
Work and workflows will be reimagined as AI-first
The operating model of the agentic era will be anchored around reimagined AI-first workflows, with humans and IT systems selectively reintroduced in AI-native design. At a European automaker and a public sector organization, squads of agents are reverse-engineering and modernizing legacy systems while humans steer and validate work. In product development, agents can gather feedback, analyze data, test features, and even run campaigns. Humans will be mostly positioned above the loop to steer and direct outcomes and selectively within the loop where human contact matters.
Outcome-aligned agentic teams will be organizational building blocks
Traditional organizations have been built around functional silos. Digital companies have cross-functional product teams but are still constrained by handovers and human team size limitations, such as the two-pizza team1 and Dunbar’s number.2
In the agentic organization, structure will pivot to small, outcome-focused agentic teams. An agentic team—a smaller group of multidisciplinary humans who own and supervise the underlying AI workflows—can be organized to deliver clear end-to-end business outcomes covering the full functional value chain of marketing, product management, technology, data, and operations. In our experience, a human team of two to five people can already supervise an agent factory of 50 to 100 specialized agents running an end-to-end process such as onboarding a customer, launching a product, or closing the books. Agentic AI can extend the scope and autonomy of a product team more than ever.
Winners orchestrate flat networks of agentic teams
Proliferation of AI agents without the right context, steering, and orientation can be a recipe for chaos. Winning operating models of the future will empower agentic teams, with flat decision and communication structures that operate with high context sharing and alignment across agentic teams to ensure they move in sync. Organization charts (based on traditional hierarchical delegation) will pivot toward agentic networks or work charts (based on exchanging tasks and outcomes).3 Finally, agentic networks are not necessarily limited to the boundaries of a single organization, and different outcomes may be sourced from different parties, opening up new B2B opportunities.
3. Governance
In the agentic organization, governance cannot remain a periodic, paper-heavy exercise. As agents operate continuously, governance must become real time, data driven, and embedded—with humans holding final accountability.
Decision-making accelerates with real-time data
Traditional budgeting, planning, and performance management cycles are too slow for AI-first workflows. Early movers are experimenting with “agentic budgeting,” in which AI agents propose budgets, scenario agents run forecasts, and reporting agents provide real-time insights. Finance leaders shift from collecting spreadsheets to interpreting signals, stress-testing scenarios, and engaging directly with the business.
Agents control agents through embedded guardrails
Just as DevSecOps (development, security, and operations) embedded automated checks into digital delivery, agentic organizations will embed control agents into workflows. Critic agents will challenge outputs, guardrail agents will enforce policy, and compliance agents will monitor regulation. Every action can be logged and explained in real time—from data privacy to financial thresholds to brand voice. An AI governance framework across the life cycle of AI agents—from agent discovery and initiation to decommissioning—can balance speed and scale with the required security and control mechanisms.
Human accountability and oversight remain
Human accountability and oversight will remain essential, but their nature will change. Rather than conduct line-by-line reviews, compliance officers and leaders will define policies, monitor outliers, and adjust the level of human involvement. The challenge is finding the sweet spot: enough oversight to manage risk without pulling agents back to human speed. Companies that get this balance right will capture more of the agentic advantage. Ultimately, the scale of agentic adoption will be capped by how much oversight capacity humans can provide—making governance itself a potential bottleneck to productivity.
4. Workforce, people, and culture
In the agentic organization, humans will move from executing activities to owning and steering end-to-end outcomes. That shift demands new profiles with different skills and a culture that provides cohesion and purpose.
The hybrid agentic workforce needs a new talent system
As agents take on execution, people will increasingly define goals, make trade-offs, and steer outcomes. This will change how companies plan for a hybrid workforce, whom they hire (or borrow), how they deploy human or AI talent, and how they measure success. HR systems not only track human employees but also are a repository of agents and agentic workflows. Performance management anchored in task completion will give way to systems that track how well people orchestrate agents, unlock value, and deliver outcomes. In this new paradigm, the talent system itself must be rethought—from career paths to incentives to leadership models.
New talent profiles with different skills emerge
In our work with pioneering organizations, we see AI agents replacing tasks historically handled by knowledge workers, such as analyzing documents and creating APIs. At the same time, we see rising demand for other skills—for example, deep problem-solving with an end-to-end lens, application of system design, and the ability to apply pattern recognition to edge cases where agents fail.
Three roles are emerging as humans work alongside agents (Exhibit 2):
- M-shaped supervisors: broad generalists fluent in AI, orchestrating agents and the hybrid workforce across domains
- T-shaped experts: deep specialists who reimagine workflows, handle exceptions, and safeguard quality
- AI-augmented frontline workers: employees in sales, service, HR, or operations who spend less time on systems and more time with humans
Leaders themselves will also evolve. CEOs, product officers, and compliance heads will increasingly need the technology fluency once expected only of chief information officers. Filling these roles will require upskilling and reskilling at scale. Early evidence shows that employees without technical backgrounds can learn to manage agentic workflows as quickly as trained engineers. Career paths and performance systems will need to adapt as “boxes and lines” give way to ecosystems of human and digital skills. As these profiles take hold, the constructs of “organization” and “employee” will become more fluid, with ecosystems of human and digital talent blending inside and outside an organization.
Culture acts as glue and ethical compass
Culture will become both the operating glue and the ethical compass of the agentic organization. Pioneering agentic organizations highlight the need for orchestration—to align teams around shared context and outcomes, identify the right mix of human and AI capabilities (as not everything needs agentic AI), and build trust between humans and agents. The culture compass embeds values and long-term purpose into agentic systems, so companies don’t chase short-term efficiency at the expense of cohesion and trust. Early pioneers show that clarity, decisive leadership, and continuous learning are critical—but what will differentiate winners is their ability to preserve cohesion and identity while transforming at pace.
5. Technology and data
In the agentic organization, technology and data will get democratized, supported by an agentic AI mesh. Agent-to-agent protocols will make integration across systems, machines, and humans easier and cheaper. Successful scalers will balance build-versus-buy decisions based on sources of distinctiveness and competitive advantage, avoiding technology or vendor lock-in so they can adapt quickly to a fast-evolving offering landscape.
Distributed ownership of IT and data becomes feasible
In the digital era, technology and data systems evolved from centralized monoliths and databases sitting far from the business toward microservices and data products sitting close to the business. This required significant software and data engineering expertise to design, develop, and maintain the underlying technology and data. In the emerging agentic age, business-side employees will be able to independently create software assets and manage data through agentic AI, which automates the software development life cycle (SDLC) with oversight from deep specialists. Early adopters have seen productivity at least double, with employees from diverse backgrounds—such as a French literature graduate in one of our teams—proving as capable as software engineers in building agentic workflows.
To scale this transformation responsibly, organizations must adopt agentic platforms and architectures, such as an agentic AI mesh. These platforms provide reusable, high-performing “atomic” agents and data products equipped with technical safety guardrails to prevent buildup of technical debt or cybersecurity risks, while unlocking unprecedented levels of democratization and innovation.
Agent-to-agent protocols ease interactions and integrations
Agent-to-agent protocols are redefining interactions between humans, agents, IT systems, and devices. Rather than relying on traditional IT system integrations such as middleware and APIs that require heavy programming and custom system-to-system connections, agent-to-agent protocols enable systems to use agents to communicate with other systems. By moving to agent-to-agent dialogue that sits above the underlying system complexity, organizations can integrate legacy systems, cloud platforms, and even machines such as drones into cohesive workflows more quickly and at lower cost. More important, this allows for faster experimentation—in which new capabilities can be tested, scaled, or deprecated without months of engineering effort.
Dynamic sourcing becomes critical
Many business-critical platforms were historically built in-house or selected in rigorous sourcing processes, with multiyear implementation transformations. These systems were intended to remain largely stable to secure competitiveness for decades. A much more flexible strategy will be needed in the agentic age. Large language models and AI products are evolving so fast that locking in one solution or vendor can lead to technology that is outdated in a matter of weeks. At the same time, organizations will need to wall in proprietary organizational context, institutional knowledge, and nonpublic data for competitiveness. This requires architecture that separates the agentic structure, logic, and data from the underlying vendor landscape.
Agentic organization
Capture enterprise-wide value from AI and agentic technologies
How to start the journey
The most frequent question we heard in our discussions with executives was, “How do I start?” Executives wonder how to create a North Star vision without clarity on what the future holds; how to assess maturity and upgrade needs for data, technical, and governance foundations; how to set priorities for value and feasibility; how to bring the organization along in terms of skills and mindset; and how to scale faster than rivals to create a competitive advantage. The clear and present danger is ending up with “more pilots than Lufthansa,” being disconnected from value drivers, seeing AI everywhere but in your profit-and-loss statements, or ending up with PR fiascos.
Building on our transformation experiences, we believe that companies that want to secure a competitive advantage in developing an agentic organization should think boldly, move fast, and go deep. In the journey to become an agentic AI leader, executives will need a different mindset to get to a coherent set of choices and actions across the 15 themes we have laid out in this article (Exhibit 3).
We encourage leaders to think through three radical shifts to make a step change in how to transform for the agentic era:
- From linear to exponential: While technology develops exponentially, organizations and operating models typically evolve linearly, which can limit how much value an organization can ultimately capture. Don’t let this happen. Leadership teams will need to take bold stances in adapting operating models toward the agentic organization—replacing functional silos with cross-functional autonomous agentic teams, redesigning incentives and support processes to enable the change, and investing in required capabilities.
- From technology-forward to future-back: Delegating the agentic transformation to your technology leader, as you would with a software deployment, will not suffice. Leaders need to envision the organization of the future, its full value potential with AI-first processes and a hybrid human–agent organization—and then work backward to identify the places to begin. You can only learn by doing, not by reading books or talking about it on the golf course. Bringing this to life by boldly reimagining one end-to-end domain will go a long way in building the organization’s learning muscle. And in parallel, leaders should start planning for and building the scaling enablers beyond their first lighthouse.
- From threat to opportunity: Leaders may feel apprehension about agentic AI’s impact on day-to-day operations. It is critical for executives to continuously engage with employees about the new possibilities that this technology can unlock, not just for the organization’s growth and purpose, but also for them as professionals. Overinvesting in upskilling beyond basic literacy—as well as change management, incentives, budget, communications, and performance management to support the transition—will help pave the way.
Organize to Value
Build your operating model with value at the core
Concretely, leadership teams can start by taking these steps: making agentic AI a prominent part of the top team agenda; outlining the CEO’s vision for creating an agentic organization; ramping up an AI center of excellence; upskilling people; and rewiring one or two lighthouse domains1 to launch agentic processes quickly and “learn live.”
Organizational paradigms do coexist: 89 percent of organizations still live in the industrial age, while 9 percent have agile or product and platform operating models from the digital age, and only 1 percent act as a decentralized network. But the time has come for organizations to move as quickly as possible toward the new agentic paradigm to gain a significant competitive advantage or risk being left behind.
Many factors will influence the pace of adoption, including the development of AI models, availability of computing power, progress in robotics, changes in regulations, societal understanding and acceptance, and the human appetite for risk and change. While our insights will evolve along with AI technology in the months and years to come, we are certain that the organizations that adapt and learn faster will be the early winners in this agentic era.
Related Articles
The executive's AI playbook
