AI Did Not Automate Your Jobs — It Revealed They Were Never the Real Work

You have probably seen the headlines: Block laid off 4,000 people and Jack Dorsey pointed at AI. The story got framed as yet another cautionary tale about automation eating jobs. But if you stop at that headline, you miss what is actually happening — and what it means for how you run your organisation right now.

The Question Everyone Is Asking Is the Wrong One

When AI enters the conversation at the executive level, the reflex is always the same. Someone decomposes roles into task lists. Researchers publish percentages. A headline appears. And leadership maps the org chart against it, counting exposed positions like inventory.

This framing has a fatal flaw. It treats the organisation as a fixed structure — a set of roles performing a set of tasks — and treats AI as a force that automates some of those tasks while leaving others untouched. That model is wrong on every count.

The real impact is not two or three times larger than the headlines suggest because AI automates more tasks. It is larger because AI eliminates the reason many of those tasks exist in the first place.

The Coordination Tax: What Most of Your People Are Actually Doing

Pull up your calendar from last week — the real version, not the idealised one. Count the hours. Be honest about which activities created something that did not exist before, and which activities existed only to synchronise humans who could not share a brain.

Research puts the split at roughly 60/40. Microsoft's 2025 Work Trend Index found employees spend 57% of their time communicating and 43% creating. Other studies land at 60% on work about work. The average knowledge worker sits through more than 11 hours of meetings per week — a number that has tripled since 2020.

None of that overhead is waste in the lazy sense. It is a structural tax — the cost of running an execution layer made of people. Consider what coordination work actually is:

• PRDs and specs: Translating a decision into a form someone who was not in the room can act on.
• Status meetings and stand-ups: Synchronising state across people who cannot see each other's progress directly.
• Design handoffs: Converting one person's vision into instructions another person can execute.
• Onboarding programmes: Transferring institutional context that lives in people's heads into a new human brain.

These activities are not symptoms of bad management. They exist because human bandwidth is finite, context does not transfer instantly, and communication loses fidelity in translation. Remove the constraint, and you remove the activities that exist to manage it.

What Happens When the Coordination Layer Evaporates

Agentic AI systems — software that reads context, writes code, verifies outputs, and iterates — do not just automate tasks within your current structure. They dissolve the structural reasons your organisation is shaped the way it is.

An agent starting a task reads the codebase, the progress file, and the git history. Full context in seconds. The onboarding tax does not apply. There is no institutional memory stored in human heads that needs transferring. The status is the commit history — inspectable at any time, by anyone, without a meeting.

This changes the math on every role that exists primarily to manage human-to-human handoffs:

• The coordination roles disappear — not because their work was automated, but because the handoffs they managed no longer happen.
• The translation roles disappear — not because AI writes better specs, but because the spec and the implementation become one artefact.
• The oversight roles shrink — not because quality matters less, but because verification is built into the agent loop rather than delegated to a separate function.

The Org Is Moving to Code — and That Is Not What You Think It Means

When AI-native companies say the organisation is moving to code, they do not mean everyone becomes a software engineer. That is the 2015 version of this conversation. What has changed is simpler: code is now readable and writable in natural language. The interface between human intent and machine execution is a conversation, not a programming language.

This means every knowledge worker gets closer to the product itself — not to a description of the product.

• A product leader shapes the actual artefact, not a document that an engineer will later interpret.
• A marketing lead builds the landing page, tests the conversion, and sees the result — in the same session.
• A designer iterates on the final implementation directly, not a mock-up that approximates it.

This is not a productivity gain in the conventional sense. It is a fundamentally different relationship with work. The translation layers between people get deleted, not automated. That distinction matters for how you plan your transformation.

What This Means for CEOs, CIOs, and CTOs Today

There is a spectrum here. Many organisations are at the stage where AI helps write smarter PRDs, generates draft specs, or summarises meeting notes — and that has real value. But do not mistake it for the destination.

The organisations already operating at the frontier have eliminated entire coordination layers. No sprint planning. No design-to-engineering handoff. No separate QA function. The question for leadership is not "which roles are at risk" but "which coordination structures in my organisation exist only because humans need them."

Some practical starting points for the executive team:

1. Audit the coordination tax. Ask your senior leaders to map what percentage of their team's time goes into synchronisation versus creation. The numbers will surprise you.
2. Identify your translation layers. Where does work stop and get re-explained to the next person? Every translation layer is a candidate for elimination, not just automation.
3. Pilot direct artefact ownership. Pick one product area and let a small team iterate directly on the final deliverable using AI, skipping the handoff chain. Measure cycle time.
4. Redefine roles around outcomes, not tasks. When the coordination overhead drops, the humans who remain get closer to value creation. Redesign their roles to reflect that shift.

The Bitter Pill Is Actually Good News

The uncomfortable truth that the Block story illustrates is not that AI replaces people. It is that AI makes visible how much of what we called "work" was never the actual work. The code is the value. The shipped product is the value. Everything between "we know what to build" and "it exists and it works" is process — necessary overhead in a world where execution runs on human bandwidth.

As that constraint changes, organisations that were built around managing it will need to change shape. The leaders who act on this now — by redesigning structures rather than just adopting tools — will find this bitter pill considerably easier to swallow than the alternative: being surprised when the coordination layer evaporates on its own schedule.

If you want to think through what this restructuring looks like for your organisation specifically, that is a conversation worth having sooner rather than later.