Most Companies Are Optimising the Wrong Thing

Right now, every company is adopting AI. Finance uses it to produce reports faster. Customer service uses it to draft responses. HR uses it to screen CVs. These are real gains. They are also the wrong question.

Adoption is not the same as redesign. And the gap between the two is where the next three to five years of competitive advantage will be decided.

The Optimisation Trap

Most of what an organisation does day to day, coordination, status reporting, exception handling, process execution, was never where human intelligence was irreplaceable. It was simply where the work had to happen because nothing else could do it. That constraint is gone.

Yet most companies are still optimising the structure that constraint produced. They're making the old org chart run faster with AI added onto its edges. Faster is not the same as different, and the structure built around slow, manual coordination is now the thing creating friction, not the thing solving for it.

The right question is not what tools to deploy. It's this: if you were building this organisation today, knowing what AI can do, would you build it the same way? For almost every company, the honest answer is no.

Why This Is Urgent, Not Theoretical

Three data points make the timeline concrete.

First, a large share of new companies today are being started by a single founder running the equivalent of a full team through AI agents. Their cost structure has no equivalent in a traditionally staffed competitor.

Second, the companies doing this well aren't just cutting headcount and deploying AI into the same shape. They're redesigning how the work flows first, and the efficiency follows the redesign, not the other way around.

Third, the cautionary examples are just as instructive. Several well-documented AI-driven outages at major technology companies over the past year share a common root cause: a governance failure, not a technology failure. Fast capability, slow governance. That combination is what turns an AI incident into a multi-day trust problem.

The Three Questions, In Order

This is where The Organisational Singularity, the framework I use with leadership teams, earns its name. It organises the redesign problem around three questions, and the order matters more than any individual answer.

Where are we going? What does this organisation look like when it's actually designed for AI, not patched for it? Without a written answer here, every later decision is provisional and every argument in the room is winnable by whoever's loudest.

How does it run when we get there? What's the operating logic once you arrive: how decisions get made, how work moves, how performance gets measured in a structure where coordination and pattern-recognition work are no longer the bottleneck.

How do we get from here to there? What's the sequence, and what does the workforce transition actually require, not in the abstract, but for the specific people in specific roles.

Most organisations try to answer question three before they've answered one and two. That's the single most common reason transformations stall: the migration plan gets built before the destination exists.

What Good Looks Like

A leadership team that has stopped asking "how do we give people AI tools" and started asking "how do we change how we're built." The first question produces a tool rollout. The second produces a different company.