Before You Automate, Simplify.

McKinsey published the State of Organizations 2026 report this week, drawing on input from more than ten thousand executives across fifteen countries and sixteen industries. It is a dense document. But buried in a chapter called "From Structure to Flow" is one of the clearest and most important arguments I have seen made at this scale about how organizations should be thinking right now.

The finding: two-thirds of senior leaders say their organizations are overly complex and inefficient. Nearly forty percent say redesigning process flows is their single biggest productivity unlock over the next one to two years. And the prescription -- stated plainly -- is: simplify first. Then automate.

That sequencing matters more than almost anything else your organization will decide about its AI strategy this year.

Someone had to say it. McKinsey just said it with ten thousand data points behind it.

The Productivity Play That Has Stopped Working

For the past decade, the default organizational response to productivity pressure has followed a familiar pattern: restructure, delayer, reduce headcount, cut costs, and hope that the remaining people find efficiencies. McKinsey's report is direct about what has happened to that approach: it is hitting diminishing returns. The organizations that have run this play repeatedly are not finding new gains. They have already cut what was cuttable.

The bigger opportunity -- the one that most organizations have not yet seriously pursued -- is not about fewer people. It is about how work moves. Or more precisely, about why it moves so slowly and with so much friction in organizations that are otherwise well-staffed and well-resourced.

The culprits McKinsey names are familiar to anyone who has spent time inside a complex enterprise: unnecessary meetings, unclear decision rights, redundant handoffs, duplicated work across functions, and approval processes that exist because nobody ever stopped to ask whether they still need to. These are not technology problems. They are design problems. And they cannot be solved by adding more technology on top of them.

This is exactly the argument I have been making inside transformation programs for fifteen years. The organizations that get the most out of new technology are not the ones that deployed it fastest. They are the ones that redesigned the work first.

What "Automate a Broken Process" Actually Produces

Let me make this concrete, because it is easy to nod at in the abstract and miss in practice.

When an organization automates a process that is already working well, automation produces a real gain: faster output, lower error rate, freed capacity for higher-value work. That is the promise, and in a well-designed process, it often delivers.

When an organization automates a process that is poorly designed, the result is different. The automation runs faster. The broken output arrives more quickly. The decisions that were unclear before the automation are still unclear -- except now they are unclear at higher volume, with less time to catch the errors, and with a human workforce that has been partially removed from the loop at exactly the point where their judgment was needed most.

This is not hypothetical. The HBR research I wrote about last week noted that AI creates conditions for "workslop" -- lower-quality output delivered faster as employees are pushed to produce more with less time for quality checks. Workslop is not caused by AI. It is caused by deploying AI into work systems that were not designed to sustain quality at accelerated pace.

McKinsey's prescription and the HBR research are pointing at the same problem from different angles. Complex, inefficient systems plus AI does not produce simple, efficient outputs. It produces complex, inefficient outputs faster.

What Simplification Actually Means

I want to be specific here, because "simplify your processes" can sound like advice that is easy to agree with and impossible to act on.

McKinsey names the specific interventions: reduce unnecessary meetings, clarify decision rights, cut redundant handoffs, streamline approvals. These are organizational design tasks, not technology tasks. They require someone to look at how work actually moves through the organization -- not how the org chart says it should move, but how it actually moves -- and to make explicit choices about where the friction is and why it exists.

In practice, this work almost always reveals one of three things.

The first is decisions that are not being made. Processes stall at points where nobody has the authority or the clarity to move them forward without convening a meeting, escalating to a senior leader, or waiting for alignment that never quite arrives. The friction is not inefficiency -- it is an unmade governance decision wearing the costume of a workflow problem.

The second is accountability that is distributed so broadly it effectively belongs to nobody. When everyone is responsible for an outcome, the outcome tends to happen slowly, poorly, or not at all -- not because the people involved are underperforming, but because diffuse accountability produces diffuse effort. This is not a motivation problem. It is a design problem.

The third is process steps that exist because they once made sense and nobody has revisited them since. Legacy approvals, inherited review cycles, standing meetings that outlived the decisions they were created to support -- these are not malicious. They are organizational sediment. And they accumulate faster than anyone cleans them out.

None of these three problems are solved by AI. They are exposed by it. And McKinsey is saying, with ten thousand executives' experience behind it, that clearing them is worth more than the next automation project on the roadmap.

What This Means for Your AI Roadmap

Here is the question I would encourage every leader and change professional reading this to bring into their next AI strategy conversation:

What process is this automating, and is that process currently working well?

Not "does this process exist." Not "is this process documented." Working well -- meaning decisions are clear, accountability is named, handoffs are clean, and the output is reliable at current pace. If the answer to that question is "not really," the honest recommendation is to start with the process, not the tool.

This is not an argument against AI investment. It is an argument for sequencing AI investment in a way that actually returns value. McKinsey's report is clear that the productivity frontier in 2026 is not more automation. It is better flow. AI can dramatically accelerate flow in an organization that has designed its work well. In an organization that has not, it accelerates something else.

The organizations that will win on AI ROI over the next three years are not the ones moving fastest. They are the ones that paused long enough to simplify what they were about to scale.

Try This

Before your next AI initiative gets scoped and resourced, run this two-question diagnostic on the process it is targeting:

First: can you describe in one sentence who owns the decision at each major step of this process -- not who is involved, but who owns the call? If the answer requires a paragraph or a committee, the decision rights are not clear enough to automate around.

Second: if this process ran ten times faster tomorrow, would the outputs be better or just more? If more is the honest answer, you are about to automate volume rather than value.

If either question surfaces uncertainty, that is your signal. Simplify first. The automation will still be there when the work is ready for it. And it will work considerably better.

This is the second article in the Unbroken Work AI series. The research keeps pointing in the same direction: the organizations that will get AI right are not the fastest adopters. They are the most intentional ones.

If you found this article first, the series begins with "AI Does Not Reduce Work. It Intensifies It." -- worth reading in order.