They digitized channels. Automated tasks. Modernized systems. Connected platforms. Built apps. Improved reporting. Introduced workflow. Reduced manual effort. In many cases, they also started redesigning around the customer rather than around internal silos. That has been the dominant transformation journey for a long time.
But I think something else is now starting to happen.
We are moving beyond a purely digital operating model and into something more agentic.
That is a bigger shift than it may first sound.
The digital era was largely about using systems to support the work. What is changing now is that systems are starting to do parts of the work themselves.
Not all of it. Not everywhere. Not without controls. But enough that leaders need to think differently.
Systems used to support the work
Historically, enterprise systems mostly sat alongside people.
They helped capture information. Route tasks. Produce reports. Trigger workflows. Store documents. Enforce rules. Improve visibility. Speed things up.
But in most cases, the person still stayed at the centre of execution.
The person interpreted the issue. The person made the judgment. The person decided the next step. The system helped, but it did not really participate in the work in any meaningful sense.
That is beginning to change.
AI is making it possible for systems to do far more than store, route, or calculate. Systems can now classify, summarize, explain, draft, reconcile, interpret, recommend, and in tightly governed cases, trigger action. Done properly, they can participate inside the flow of work itself rather than sitting outside it as a passive support layer. I touched on this in a recent article on agentic AI: the model plans and explains, but the surrounding system provides the context, tools, guardrails, and observable execution.
That is where the operating model question starts to become real.
The question is no longer just:
Where can we use AI?
It is increasingly:
How should the enterprise operate when governed AI agents begin to participate in execution?
That is a very different question.
This is not just another tool discussion
I think one of the easiest mistakes leaders will make is to treat AI as just another productivity layer.
Another assistant. Another interface. Another feature. Another pilot. Another shiny thing to bolt onto an existing process.
That approach will still produce some value. There is no doubt about that.
But it will also miss the bigger shift.
The real issue is not simply that people can work faster with AI. The real issue is that work itself can now be redesigned around a different mix of human effort, machine effort, workflow logic, contextual intelligence, approvals, and orchestration.
And once that becomes possible, the conversation changes quickly.
You have to think about where decision rights sit.
You have to think about what must remain under human approval.
You have to think about how context is assembled, how actions are bounded, how thresholds are enforced, and how accountability is preserved.
At that point, you are no longer talking about a tool. You are talking about the shape of the operating model.
The progression makes sense if you step back
If you zoom out, the direction of travel is not actually that surprising.
Many businesses started in a product-centric world. Structures, systems, and processes reflected product lines, internal ownership, and functional silos.
Then many moved toward a customer-centric model. That forced organizations to design around journeys, needs, and outcomes rather than departmental boundaries.
Then came the digital operating model, where platforms, integrations, automation, and data became central to how the business functioned.
What I believe we are now seeing is the next layer on top of that.
Not a replacement for digital, but a progression from it.
Digital created the rails. Agentic changes what can move on those rails.
Why this gets serious in complex enterprises very quickly
This matters most in large, heavily interconnected organizations.
That is where work is rarely a single step. It moves across multiple systems, multiple roles, multiple approvals, multiple exceptions, and multiple layers of policy and control.
In those environments, a chatbot on the side does not really solve the hard problem.
The hard problem is whether AI can participate inside real workflows, in real architecture, under real controls, and still be trusted.
That is why I do not think the next phase is mainly about better prompts or nicer interfaces.
It is about how AI gets embedded into execution in a way that is bounded, governed, and architecturally sound.
AI fits best inside a well-governed execution platform, with controls, policy-aware decisioning, traceability, approval points, and connected data rather than floating above the organization as a novelty.
The implications are much broader than technology
Once systems start doing parts of the work, several other things start to shift with them.
Workflow design becomes much more important.Architecture becomes more strategic.Governance has to move earlier.Human roles start changing.Institutional knowledge becomes more valuable, not less.And leadership can no longer treat this as a narrow IT matter.
Because the real questions are not only technical.
They are operational.
Who approves what?
What is the boundary between human judgment and machine action?
Where do exceptions go?
How are policies enforced?
How is cost controlled?
What gets logged?
How is trust maintained?
How do you stop AI from simply amplifying fragmentation, ambiguity, and poor design?
Those are enterprise design questions.
This is not a story about replacing people
Whenever this topic comes up, people tend to jump straight to replacement.
Will AI remove jobs? Replace teams? Eliminate decision-making?
In most serious enterprise environments, I think that is the wrong place to start.
The more useful question is this:
How do we redesign work so that AI can absorb appropriate effort, while people move toward higher-value judgment, supervision, intervention, and leadership?
That is a much more practical conversation.
In a well-designed environment, AI should not weaken control. It should strengthen it.
It should not reduce transparency. It should improve it.
It should not blur accountability. It should make it clearer.
And it should not operate in a vacuum. It should operate inside a visible, governed system of work.
The organizations that will get the most from this
The organizations that benefit most from AI will not necessarily be the ones talking about it the loudest.
They will be the ones that understand what kind of shift this really is.
They will recognize that AI is not just a clever layer on top of the enterprise. It has implications for workflow, architecture, governance, control, and operating model design.
They will redesign work deliberately.
They will strengthen architecture before chasing automation.
They will think in terms of orchestration, not isolated tools.
They will build governance into the design from the beginning.
And they will treat the move from digital to agentic as a serious transition in how the business operates.
That is where I think the next real wave of transformation sits.
Not simply more digital.
Something more embedded than that. More participative. More orchestrated. More agentic.
And the leaders who see that early will shape what comes next.