Don't automate the old organization.
Design the next one.

Walk into almost any boardroom today, and you will hear the phrase "AI transformation." Look closer, and most of what is being called transformation is something more modest: AI-enabled continuity. The org chart is the same, so are the approval chains or, for that matter, the handoffs between functions.

The work simply moves through faster.

Most organizations are using AI to go faster inside structures that were built for a different era, and then wondering why the returns feel thin. That is the wrong ambition. And the evidence is mounting.

McKinsey Global Institute estimates that AI could automate up to 70% of the tasks knowledge workers perform today, yet the same research is clear that automation alone does not improve outcomes without a matching redesign of roles and workflows. Deloitte finds that only 22% of organizations have restructured roles and teams alongside their AI deployments. The rest are layering new tools on old foundations. Gartner expects that by 2027, 50% of companies that invested heavily in AI productivity tools will have seen negligible impact on business outcomes, because they never changed how work is organized around those tools.

The structures were built for scarcity

It is worth remembering why our organizations look the way they do. Functional silos, management layers and handoff-heavy processes were not mistakes. They were rational answers to a world in which information was scarce, and coordination was expensive. Managers existed to gather, filter and route information. Hierarchies existed because only a few people had access to enough information to make decisions. Handoffs existed because expertise was concentrated and data lived in separate systems.

AI removes much of that scarcity. When a team can see real-time customer, operational and financial data, and an AI agent can synthesize it in seconds, the case for routing every decision upward weakens considerably. While the structures remain, their justification does not.

So, the deeper question is not "how do we automate this process?" It is more fundamental: "does this process still need to exist in its current form?" BCG Henderson Institute research suggests the answer matters enormously: companies that use AI to redesign work processes, rather than simply digitize them, achieve 2.5 times more value than those that automate unchanged existing processes.

None of this makes the old structures easy to dismantle. They carry real weight. Hierarchies provide accountability. Functional controls manage risk. Layers coordinate across complex, regulated operations. If you remove them carelessly, you do not get agility; you get confusion, duplicated effort and decisions that lack ownership. Any leader who has lived through a botched reorganization knows the cost.

Autonomy is an architectural choice

The alternative is not the absence of structure. It is a different structure altogether: smaller teams organized around customers and outcomes, with genuine ownership of results, direct access to the data they need, and decision rights pushed to where the work actually happens.

Self-managing teams are often dismissed as a soft, cultural idea. They are the opposite. They are a hard architectural choice, and they demand more discipline, not less. Ownership must be explicit, accountability must be measurable, data must be trustworthy and available. Decision rights must be written down, genuinely delegated and insulated from the instinct to escalate.

When those conditions hold, the payoff is significant. Research published in Harvard Business Review indicates that self-managing teams with clear ownership and access to data outperform traditionally managed teams on customer outcomes by up to 25%. The caveat is the whole point: only when decision rights are genuinely delegated, and not just nominally so. Autonomy on paper, with approval on demand, delivers the worst of both worlds.

This is why the real barrier is rarely the technology. MIT Sloan Management Review reports that 85% of executives see their organization's model and culture, not the technology itself, as the biggest obstacle to AI value. Leaders have to redesign governance as deliberately as they redesign workflows: defining which decisions stay central, which move to teams, and how risk is managed when the hierarchy is no longer doing that work by default.

The real risk

The common fear is moving too fast. I believe the greater danger runs the other way. The risk is not moving too fast. The risk is spending the next three years automating the wrong organization, investing heavily to make yesterday's structure marginally more efficient while competitors build something fundamentally better.

The organizations redesigning work now will compound their advantage. Their teams will learn faster, decide faster and serve customers better, and that gap will become structural. It will not close with another software licence. Every leadership team will deploy AI. Far fewer will use it as the reason to rethink how they are organized. The question for each of us is simple: are you willing to design the next organization, or are you content to automate the last one?