When the Orchestration Layer Gets Smart Enough to Matter

A deeper look at the Agentic Operating System and why managing this transformation requires adaptability, cross-departmental thinking, and a new focus on the human-AI relationship.

In my last piece, I talked about the Great Unwinding, or how the enterprise software stack will be hollowed out and replaced with an orchestration layer I called the Agentic Operating System. I want to dig deeper into what’s starting to happen here, because after decades in enterprise software, I’m seeing the early stages of something that’s going to require fundamentally different thinking about how we organize work.

This isn’t just another technology shift, rather it’s the arrival of a new approach to how organizations operate when technology isn’t a destination but an organic operator itself. If this all plays out, this orchestration layer of business may become genuinely intelligent and autonomous. To get there the process of extracting business logic from legacy systems will force us to rethink not just our architecture, but our entire organizational structure, operating model, and definitions of success.

The Unwinding Has Begun

As AI agents start to move beyond simple task automation, they’ll begin to coordinate work that’s been trapped inside legacy applications which were built on database architectures from the 1990s. Remember all that business logic your company painstakingly encoded into ERP systems, CRM platforms, and homegrown applications over the past 30 years? The process of extracting and reimagining it through intelligent agents has just started and it’s going to require unprecedented collaboration across departments that have traditionally operated in silos.

The research shows we’re approaching an inflection point. BCG’s latest data reveals that 70% of AI’s potential value concentrates in core business workflows like R&D, sales, marketing, and manufacturing. Agentic AI already accounted for 17% of total AI value in 2025 and is projected to nearly double to 29% by 2028. This is the beginning of that concentration of intelligence I described reaching critical mass, but we’re still in the early innings.

Why This Requires New Organizational Thinking

McKinsey describes what they call the “agentic organization”, where operating models anchor around reimagined AI-first workflows, with humans and IT systems selectively reintroduced in AI-native design. Read that carefully. This isn’t about IT implementing new tools. It’s about reimagining operating models, which means Finance, Operations, Sales, IT, Support, and Legal need to be in the same room, redesigning processes together around what agents can do autonomously, then deciding where humans add the most value. That requires a level of cross-departmental collaboration many enterprises aren’t set up for today.

The technical trajectory is clear. The length of tasks AI agents can complete has been doubling every seven months, reaching roughly two hours of autonomous work as of late 2025. McKinsey projects agents could handle four days of work without supervision by 2027. The jury is still out whether that happens so quickly, but directionally, this is where we are headed. The coming challenge will be how best to merge what feels like an exponential growth of agentic capabilities with how the human aspects of organizations adapt and adopt. Managing this gap will require intentional, adaptive leadership. And, it will challenge decades long management practices.

The Cross-Departmental Challenge

This transformation is different from past technology shifts in that extracting business logic from legacy systems isn’t just a technical exercise, it’s also an organizational one. When you pull a procure-to-pay workflow out of your ERP and CRM systems, and reimagine it in the orchestration layer, you’re not just touching the underlying IT systems, you’re redesigning how Procurement, Finance, Legal, Support, and Operations work together. You’re challenging assumptions about decision rights, approval workflows, and exception handling that have been encoded in those systems for decades.

Deloitte’s research confirms this is just beginning. Further, Gartner predicts 15% of day-to-day work decisions will be made autonomously through agentic AI by 2028, up from essentially zero in 2024, while 33% of enterprise software applications will include agentic AI capabilities by that timeframe. That’s a three-year window to fundamentally rethink how work gets done across your organization. Again, that may be optimistic, but it is directionally accurate.

The Strategic Shift Required

This connects directly to my earlier point about compressing the distance between intent and execution. The Agentic Operating System doesn’t just coordinate faster, it requires you to think differently about where intelligence lives in your organization. Business logic is migrating from your legacy systems into a dynamic orchestration layer, but that migration path needs to be managed thoughtfully, with input from every function that relies on those systems.

Your legacy systems aren’t going anywhere for the foreseeable future, but their role is fundamentally changing. They’re becoming the persistent layer, the modern equivalent of the mainframe database, while agents become the intelligent layer. The question isn’t whether this transformation will happen, rather it’s whether you’ll manage it proactively with cross-departmental collaboration, or reactively as departments optimize for agents independently and create new silos in the process.

Managing the Intensity of Change

In my Future Intense framework, I talked about the concentration of intelligence reaching critical mass. We’re entering that phase now, but the intensity isn’t just technological it’s also organizational. The unwinding of the enterprise software stack requires unwinding decades of process assumptions, departmental boundaries, and ways of working.

The companies that’ll get this right won’t simply be asking “How do we add AI to our processes?” They’ll be activating cross-functional teams to ask “What processes should exist in an agent-native enterprise?” They’ll invest as much in organizational adaptability as they will in technology. They’ll recognize that extracting business logic from legacy systems is as much a change management challenge as it is an engineering one.

The unwinding has begun. But managing this transformation successfully requires leaders who can think across traditional boundaries, build adaptive organizations, and orchestrate change at the speed of exponential technology growth. The orchestration layer is finally getting smart enough to matter—the question is whether our organizations can become adaptive enough to harness it.