workflow redesign – After two years of companies chasing the question of “how to use AI,” momentum is shifting toward a tougher premise: many existing processes were never built to work with AI systems. The argument is that AI doesn’t simply accelerate workflows; it highlights st
For the past two years, companies have been asking the wrong kind of question: how do we use AI in our processes?
At first, it sounded practical. When large language models arrived, the instinct was to take what already existed—workflows, functions, decision chains—and accelerate them. Add copilots. Add assistants. Add automation layers. Improve productivity.
But that approach has run into the limits of reality.. The technology itself hasn’t been the main failure point, the argument goes.. The mismatch sits higher up: large language models were never designed to “run” an entire company. and simply embedding them into existing processes doesn’t correct the structural gap.
Now a different question is starting to surface, quietly but persistently: what if the problem isn’t how to use AI in corporate processes, but that those processes were never designed for AI in the first place?
The return of an old idea (this time for real)
In the 1990s. business process reengineering promised a radical shift: redesign companies around information systems rather than layering technology on top of existing workflows.. The promise was strong, but execution was uneven.. Many efforts turned into expensive reorganizations with limited long-term impact. in part because the underlying systems were rigid. fragmented. and unable to adapt in real time.
This time, the pitch is different because systems are changing shape.. In that earlier era, systems were passive: they stored information, enforced rules, and supported decisions made by humans.. Today, systems are described as becoming active—able to generate, evaluate, coordinate, and increasingly act.. That shift is said to change the equation. pushing companies away from simply digitizing workflows and toward redefining what a process even is.
McKinsey’s latest research on AI adoption reinforces the same direction: usage is widespread. but real impact tracks closely with workflow redesign. not just tool deployment.. The organizations gaining measurable benefits are portrayed as those rethinking how work is done—rather than only adding assistance.
Why most processes are incompatible with AI
The uncomfortable claim is that most enterprise processes aren’t just inefficient; they’re structurally incompatible with the kind of systems AI is becoming.
They are described as:
Fragmented: spread across tools. teams. and data silos
Sequential: built around handoffs and delays
Context-poor: dependent on individuals to reconstruct state
Decision-latent: optimized for review. not action
Human-centric by design: assuming that cognition. memory. and coordination are scarce
Those characteristics, the argument continues, made sense when humans were the limiting factor. In a world where systems can maintain context, apply constraints, and operate continuously, the same design patterns don’t fit.
Deloitte’s analysis of agentic AI is cited as capturing the tension directly: organizations often try to automate processes designed for humans instead of rethinking the work itself.. The result is described as predictable—complexity rises, but outcomes don’t improve in proportion.. That’s framed not as a tooling flaw but as a design problem.
AI doesn’t optimize processes: it exposes them
One of the repeated patterns across enterprise AI efforts is straightforward: the more AI is applied to an existing process, the more visible that process’s limitations become.
What used to be hidden by human effort is said to come into view:
missing data
inconsistent rules
unclear ownership
duplicated work
delayed feedback loops
In that framing, AI behaves less like an optimization layer and more like a diagnostic tool. It shows the gap between how a company believes it operates and how it actually operates.
That helps explain why pilots often stall.. Not because the model fails. the argument says. but because the process it is inserted into cannot absorb what the model produces.. MIT Sloan is invoked to make the point that the challenge isn’t only adopting AI—it’s redesigning organizations so they can use it effectively.
The implication is the limiting factor isn’t technology anymore. It’s the company.
From processes to systems
If the last phase of enterprise AI was about adding intelligence to tasks, the next one is described as redesigning systems so that intelligence is embedded from the start.
Instead of asking:
“How do we automate this step?”
Companies are pushed toward questions like:
“Why does this step exist at all?”
“What would this process look like if it were designed around continuous context?”
“Where should decisions actually happen?”
“What constraints should be enforced automatically?”
These are presented as structural questions, not incremental improvements.. They point toward a different kind of organization—one where processes stop being static sequences of actions and become dynamic systems that maintain state. integrate data. operate under constraints. and adapt continuously based on outcomes.. The same system-oriented traits described earlier are referenced again as the defining characteristics of the systems being discussed.
The companies that move first will look very different
The shift becomes visible in how early movers operate.. The claim is that companies that redesign processes around these principles won’t just be faster or more efficient; they’ll operate differently:
decisions will happen closer to data
coordination will require fewer handoffs
feedback loops will shorten dramatically
execution will become more continuous
roles will evolve around systems. not tasks
Microsoft’s Work Trend Index is brought in as a hint of the transition, describing organizations moving toward more dynamic, outcome-driven structures where humans and AI collaborate around goals rather than functions.
From the outside, these organizations may not appear radically different at first. Internally, though, their operating logic is described as shifting—and that shift is said to compound.
This is not optional
It’s described as tempting to treat the coming changes as opportunity. It may be an opportunity, but the argument treats it also as a constraint.
Once some companies begin to operate this way, others aren’t only competing against better tools.. They’re competing against a different kind of system—one that learns faster. adapts continuously. coordinates more efficiently. and executes with fewer delays.. The point is that you can’t match that by adding another copilot or deploying another model.. Redesign is framed as necessary.
The next phase of enterprise AI is organizational
The roadmap is laid out in phases: experimentation first. realization second. transformation third.. The twist is that transformation is described as driven not by models but by structure.. The argument says it’s not a move from “worse AI” to “better AI.” It’s a move from companies built for humans to companies that must operate with machines as part of their core logic—requiring rebuilding how companies actually work.
The real question
So the question shifts away from “how do we use AI?” toward whether companies are willing to redesign so AI can actually work. If the answer is no, the outcome is stated as already clear: AI will not fail—your processes will.
The core thread runs straight through the argument: pilots stall because existing processes can’t absorb what AI produces. and the same mismatch is then framed as a mismatch between systems that are becoming active and workflows that remain fragmented. sequential. and human-centric—so workflow redesign becomes the lever that connects adoption to measurable gains.
enterprise AI workflow redesign agentic AI business process reengineering pilots productivity