AI is everywhere. The agentic organization isn’t—yet

Yes, AI is astonishing: fast, powerful, and learning every day. But even as leaders strike up new pilots across their organizations, most still struggle to translate experimentation into enterprise value—and now, agentic AI is raising the stakes. In this episode of The McKinsey Podcast, McKinsey Senior Partner Alexis Krivkovich speaks with Global Editorial Director Lucia Rahilly about what it will take to build an “agentic organization”—from reimagining workflows to reshaping leadership roles, skills, and culture for a future where humans increasingly operate above the loop.

The McKinsey Podcast is cohosted by Lucia Rahilly and Roberta Fusaro.

The following transcript has been edited for clarity and length.

The great AI paradox

Lucia Rahilly: Alexis, we hear a lot about companies investing in agentic AI without realizing corresponding returns. In fact, our own research shows the complexity of delivering on the promise of agentic AI at any kind of scale. Where are we now?

Alexis Krivkovich: The great paradox is just that. Companies expect massive transformations from AI and have invested with that mindset. Yet more than 80 percent of companies say they’re not yet seeing impact on the bottom line from those investments.

The real question is “How do we meet this moment and create the agentic organization of the future?” In every conversation I have, leaders feel like they’re on the precipice of that question. What they’re trying to think through is really a set of questions: “How are roles going to change? What are the skills we’ll need in the future? How do I bring along our workforce with excitement, not fear? How do I drive that change so it hits every corner of the organization?”

Lucia Rahilly: The research outlines five pillars that make up the new agentic organization. Let’s start with the first, an organization’s business model. What are some examples that help illustrate the value at stake for leaders?

Alexis Krivkovich: Suppose the world could move toward near-zero marginal cost of delivery. How would that change what you’re able to bring to customers, how you’re able to tailor and hypersegment to the unit of one, and how you think about that in the context of mobilizing growth? What does it look like if you’re a tech player that does content distribution and you can tailor experience down to the individual consumer across millions of people? What does that open up in terms of how you create and enable commerce?

If you take that a step further, the question becomes “On the receiving end, what if the small business or individual has agents interacting on their behalf?” For example, imagine a customer has an agent that can move money frictionlessly across bank accounts to seek the best rate. That fundamentally changes the moat that has existed in financial services since the beginning of time. We don’t do this now because the process of making these changes is incredibly hard. If that suddenly goes away, the entirety of how you think about your business model, both the opportunity and the risk associated with competitive threat, changes.

Rethinking teams for the agentic era

Lucia Rahilly: You’re one of the global leaders of our People & Organizational Performance Practice. Talk to us about how you see team structures functioning in this agentic era.

Alexis Krivkovich: The real promise with agentic, relative to generative AI or previous evolutions of AI, is that you can have the equivalent of superhuman capabilities added to your teams. But the day-to-day workflows and the rituals around ways of working will need to fundamentally change.

That’s what we mean when we say the operating model needs to shift. You need to think about how the hours of the day happen differently, the process of overseeing an agent population, how you engage in problem-solving as a team—and put the right governance and risk controls on top of that.

For most companies, this is all new ground. We now have the opportunity and the challenge to grapple with it nearly everywhere simultaneously because the use cases are so broad.

Lucia Rahilly: As you note, it’s super early, and everyone’s model is being upended. But let’s take your work with pioneers in this area. Where are folks starting in terms of rethinking their org charts, for example?

Alexis Krivkovich: Org charts are tricky. We’re still too early to say, and the answer may look quite different across domains.

Take pharma: A lot of life sciences and pharma companies are envisioning huge squads of agents in the R&D space. That doesn’t mean the researchers and scientists they have today will be replaced in the org chart. It means they’ll be able to supercharge the speed at which they can innovate.

But in some other areas—HR, finance, legal—companies envision corporate functions as cost centers that deliver critical execution capabilities but don’t drive the front end of value generation. There’s a real hope that the agentic capability can transform how that work gets done and who does it, to enable more capacity in other aspects of the business.

Most companies have added at least one layer to their structure, from the CEO all the way to the front line, over the past decade. In some organizations, it’s more like two or three layers. Not only is that expensive, it also slows decision-making because more people weigh in before any decision can be made. There’s hope that AI will enable leaders to have a more superhuman capacity to manage across bigger scopes. That would allow companies to flatten their structure and get faster in the process.

Lucia Rahilly: What should leaders be doing to develop the skills they need both to manage broader spans of talent and to oversee agentic tech?

Alexis Krivkovich: Just about everybody in the workforce is going to need a new job description in the next two to three years. Most roles won’t go away, but they’ll be reshaped. Seventy-five percent of roles need fundamental reshaping right now. That includes people leading teams and those who report to them.

So when we ask about capability-building needs, nearly half of leaders say they think they see skill gaps in their organization. And most would say they’d really benefit from more training, more capability building, and more support—all the way up to senior leadership.

What ‘reimagining workflows’ really means

Lucia Rahilly: We talk a lot about reimagining workflows from end to end. What does this look like in practice, and what are the potential implications?

Alexis Krivkovich: The best use cases, where AI is enabling scalable impact, are where a workflow can be reimagined in its entirety. That’s because those workflows typically cut across multiple teams and areas of a company. In a traditional context, there’s a lot of connecting the dots across touchpoints and people. This is where agentic can be incredibly valuable because it can be part of that connection stream in a fast, multifaceted way.

So for companies, instead of point solutions, where you take one task and use AI to accomplish it better and faster, we’re now talking about a whole workflow: How do I take something like insurance underwriting and rethink that end to end? How do I take the HR “hire to onboard” process and reimagine that end to end? That’s where we’re seeing the magic start to happen.

But that’s also deep, granular work. You have to rethink a whole process and ask, “Where does it benefit from having an agent? Where do I need a human in the loop, or above the loop, to supervise? Where do I need a team of agents? How do I make them reusable, so once they’re trained, I can deploy them in multiple places?”

That’s the systems thinking that is starting to unlock opportunities at scale, and it has exciting potential. But it requires a set of individuals who can do that type of problem-solving and leaders who can pick areas ripe for being lighthouse use cases.

Lucia Rahilly: What would the composition of a team thinking about reimagining certain processes look like?

Alexis Krivkovich: I think this change, bringing together multiple levels of the workforce simultaneously, is one of the really exciting opportunities with AI. The potential is actually up and down the chain of command, the hierarchy of the organization. You need people at all levels thinking about how that process could look different.

A huge part of the unlock with agentic is executive function—reviewing the results of work done. You still need human judgment, but you have the potential to take tasks historically done by managers, or even senior leaders, and do much of the quality control, review, and pattern recognition assessment through the agentic process. This means you need folks at all levels, from leaders down to employees deep in the day-to-day flow of work, to be involved in thinking about how the process would look different.

Lucia Rahilly: Does that also have implications for the organization of the tech function?

Alexis Krivkovich: There are different models playing out. Many companies are creating SWAT-type teams that they can draw on, from within tech, to enable change. Others are embedding expertise inside each area of the business. It’s a classic example of trying to both change yourself and transform the organization at the same time.

What the agentic paradigm shift means for leadership

Lucia Rahilly: In the research, you call this the biggest paradigm shift since the industrial and digital revolutions. How should executives be thinking about leading organizational change at that scale?

Alexis Krivkovich: One big question for leaders is “How do I need to show up differently to lead the organization forward?” And as one said, “If this will radically transform everything about our business, we as a leadership team need to start by radically transforming ourselves.” In other words, is 50 percent of my time spent differently because I can access AI to do my job? Or am I dabbling—tightening an email here, querying there, but not really rethinking the hours of my day? If I am radically changing how I spend my time, how do I talk about that within the organization?

A huge trust gap exists today. Some of the early experimentation with AI has proven to have real faults—the hallucination stories we read in the headlines are real. And the slop that moves around a company through poor use of AI can create more work, not less. So how do you, as a leader, get people excited about what’s possible while overlaying good judgment and risk management? How do you acknowledge that we’re in a learning mode rather than a mature state of deployment?

Lucia Rahilly: There must be a certain amount of resistance that’s fear-driven as well. Folks think they may lose their jobs and be replaced by agentic AI.

Alexis Krivkovich: The technology is not going to stop. There’s that line, you won’t be replaced by AI, but you may be replaced by someone who embraces AI before you do. I think that’s a real question for the employee base. And as a leader, how do I get you excited about new skills, new tooling that could make you more successful? And if that’s not in this role, because this role will transform so dramatically, perhaps it’s in an adjacent role.

When judgment becomes the job

Lucia Rahilly: On that topic, do you see any new talent profiles that you think will be essential? How might employees begin to develop those skills?

Alexis Krivkovich: Some of the skills being affected you’d expect, because AI is really good at research, math, data science. But other aspects of those skill profiles are becoming even more prominent: strategic thinking, systems orientation, people management soft skills—they’re all on the rise. In an AI-augmented world, the first set of skills is still needed but you can do much more with them with the assistance of agents. What we’ll need even more of is oversight, judgment, problem-solving capabilities, and the operational capabilities human leaders bring as the overlay.

Lucia Rahilly: We’ve previously referred to humans in the loop. In this latest piece, I noticed we seem to be evolving toward humans “above the loop.” Is that a relevant difference?

Alexis Krivkovich: Yes. It’s a great distinction. To be clear, we will see both in the future. Having a human in the loop suggests agents are doing pieces of the process, then passing it to a human who does other pieces. Having a human above the loop suggests that if we get to a place where teams of AI agents are able to do most, if not the entire core process, the human’s role becomes judgment on top.

Let me give you an example. We’ve reimagined the American Arbitration Association’s process when a case is sent in for review. The traditional process involves gathering hundreds, if not thousands, of different data points, including photographic exhibits of contracts and email exchanges; reviewing the case file; and deciding on the right answer based on the terms of the agreement. This can take a really long time.

So we asked: “Could we train agents, using closed case files, to put together the timeline, review the fact base, look at both sides of the argument, and come to a summary decision?” We found that these agents could not only do much of the core work but, in some cases, do it better. You still need a human to look at that whole process and ask, “Do I agree with the decision the agents have come to?” It’s still incredibly critical to have that judgment layer, that human layer. But a lot of the work underneath can be done end to end by agentic AI.

Lucia Rahilly: Judgment is developed over time. How do junior employees develop the skills to oversee AI without the hazing grunt work their leaders did in the pre-agentic era?

Alexis Krivkovich: This is the billion-dollar question. If you eliminate every new software engineer, you have a very expensive model of only senior folks. If you roll that movie forward ten years, you’re also missing the next generation of talent you need.

The investment in learning and development is a bit of a sidecar. It’s viewed as important. It’s certainly valued by employees. But it runs periodically. In the future, I think it should be at the center of the journey people go on as employees, because the next generation will have access to all these tools from day one. So while they may not have the pattern recognition of 20 years of reviewing case law, they also won’t have the hurdle of being 20 years into a career and trying to figure out a massively disruptive and powerful technology for the first time. So leaders should ask, “How do I make that access from day one actually work to their advantage?”

The other aspect is capability building to help people see that change management is no longer an episodic thing. It’s a perpetual state. And we’re going to have to get really good at being comfortable in constant change without introducing chaos and risk into the organization.

Talent in a time of change

Lucia Rahilly: Any thoughts on how these changes will affect the shape of an organization’s talent hierarchy?

Alexis Krivkovich: I think it’s too early to have an answer on the final state of the shape of organizations. AI will absolutely enable that.

For example, there’s a lot of excitement about organizing around pods of work as opposed to traditional job hierarchies—pods that form and reform, are reusable, and move across the organization more nimbly. That’s really hard for companies to do in practice. You need a job hierarchy so people can be evaluated and know whom to go to for managerial support.

So I think we’re still pretty far away from seeing most companies reorganize themselves around that principle. But I do think this idea of fluid talent flows is going to increase. Organizations that have an HR function that enables this flow and can support nimbler movement of talent will have a competitive advantage.

Lucia Rahilly: What are the implications for organizational culture in the context of a transformation of this magnitude?

Alexis Krivkovich: In this moment, organizations that foster curiosity and continuous learning will be in a great space because so much is still unknown. We need people who want to experiment and learn.

The people who can do joint problem-solving, who can bring together people from across the organization, coach them on how to work into new models to explore the change with optimism and not fear—those people will be a premium as well.

We also need people who can strike the critical balance between taking some risks and not betting the company. This is going to be one of those moments where we have to be highly iterative. We need two-way doors, not one-way doors. We need situations where we can experiment and explore. And if it doesn’t work, we pull back and go into the next pathway.

Lucia Rahilly: Super interesting. Alexis, thanks so much for joining us today.

Alexis Krivkovich: It was such a pleasure.