AI Is Making the Work Faster. And the Decisions Harder.

A recent Fortune article offered one of the better descriptions I have seen of how AI is changing work.

Princeton computer scientist Arvind Narayanan depicts traditional work as a hamburger. The thick patty in the middle represents execution: writing the code, producing the analysis, building the presentation, drafting the contract. The thinner buns represent deciding what to do and delivering the result.

AI changes the proportions. The execution layer shrinks. The decision and delivery layers grow. We end up with a skinny hamburger and an enormous bun. (Fortune)

I think that is exactly what is happening.

I also think the metaphor reveals a much harder problem than the future of any individual job.

When execution becomes cheap and fast, organizations have to make many more decisions about what deserves to be built, who should trust it, when it is ready, and who owns the consequences. Most companies have spent decades optimizing the execution layer. They have processes for allocating engineering hours, reviewing budgets, approving headcount, and managing projects.

They have much weaker systems for scaling judgment.

That weakness is already showing.

In the Fortune article, Yahoo Finance product chief George Leimer describes a world in which his team can build and ship with astonishing speed. His team reportedly released more than 100 features for its AlphaSpace investing product within roughly two months. When almost anything can be built quickly, the constraint shifts. Teams have to be far more certain that they are building something worth releasing. (EUROPE SAYS)

This is where many companies will get AI wrong.

They will count output.

More campaigns produced. More code committed. More customer concepts tested. More reports written. More features shipped.

The dashboards will look terrific.

Some of the work will have little value. Some will create more work for everyone downstream. Some will reach customers before anyone has seriously questioned whether it is accurate, useful, secure, or aligned with the company’s strategy.

AI allows a team to move from a vague idea to a plausible deliverable before the organization has formed a clear opinion about the idea itself.

That feels like productivity. Often, it is just creating more chaos and slop.

Kate Smaje of McKinsey has described this as a “false productivity” trap: organizations can build almost anything instantly, including things they do not need. The Fortune article also points to “workslop,” AI-generated material that looks finished but transfers the burden of checking, correcting, and interpreting it to someone else.

Anyone who has received a polished, six-page AI-generated memo that should have been a thoughtful three-sentence recommendation knows what this feels like.

The work has been completed technically. The thinking has not.

This is why the growing “decide” layer matters so much.

Deciding includes framing the problem correctly, identifying the tradeoffs, noticing what the model missed, and determining which evidence deserves weight. It requires someone to understand the business well enough to know whether an answer is merely coherent or actually useful.

AI can produce twenty reasonable paths in seconds. It cannot relieve leaders of the responsibility to choose one.

In fact, it makes that responsibility heavier.

When creating an option was expensive, organizations naturally limited how many options they considered. Scarcity imposed discipline. A proposal had to earn engineering resources, budget, or executive attention before it could become real.

AI removes much of that friction.

Now every team can produce endless ideas, prototypes, analyses, and recommendations. The cost of creating an option approaches zero while the cost of choosing among them keeps rising.

Decision fatigue will become a serious operational risk.

Leaders will face more proposals with less time to understand the assumptions behind them. Employees will encounter more AI-generated recommendations that appear equally credible. Boards will receive more detailed analysis without necessarily receiving better insight.

The danger is that organizations respond by creating more committees, more approval stages, and more meetings about the work. That would make the upper bun bigger without making it smarter.

The answer is stronger judgment closer to where the work happens.

People throughout the organization need clear authority to make decisions within defined boundaries. They need to understand when they may rely on AI, when they must verify it, and when the possible consequences require escalation. Senior leaders need to become far more explicit about which decisions matter, which risks are acceptable, and what outcomes the organization values.

Otherwise, AI speed will collide with organizational ambiguity.

The delivery layer presents a different set of problems.

Generating a product, report, campaign, or policy is only part of the job. Someone still has to review it, integrate it into existing systems, persuade people to use it, explain how it affects their work, and remain accountable when something goes wrong.

This is frequently the least glamorous part of transformation, and it is where much of the real work lives.

I saw this repeatedly during major technology shifts across mobile, cloud, consumer software, and AI. The technology often moved faster than the organization’s ability to absorb it. A product could be technically ready while the sales team lacked a clear story, customer support lacked the right training, legal had unresolved concerns, and the metrics rewarded yesterday’s behavior.

AI intensifies that gap because the output arrives so quickly.

A model can generate code in minutes. The code still has to survive security review, testing, deployment, maintenance, and actual customer use.

An AI system can produce a strategy deck before lunch. The company still has to make the choices in it.

A tool can automate a workflow. Someone still has to redesign roles, address exceptions, change incentives, and decide what happens to the people whose work has changed.

This is why I agree with the Fortune article’s suggestion that the future of work may need to look less like a distorted hamburger and more like a pizza, with decision, execution, and delivery distributed more evenly across the work itself.

But that redesign will demand more than teaching everyone to use AI.

It will change who gains influence inside organizations.

For decades, many companies promoted people because they were exceptional at execution. They possessed scarce knowledge. They knew the systems, mastered the process, and could personally solve the hardest technical or operational problem.

Those strengths still matter. Their monopoly on organizational value is weakening.

As AI handles more execution, influence will move toward people who can frame ambiguous problems, exercise judgment without complete information, connect decisions across silos, challenge confident answers, and take responsibility for outcomes they could not fully predict.

These capabilities have often been treated as secondary to expertise. In an AI-heavy organization, they become central to performance.

The workforce implications are substantial.

Entry-level employees have traditionally developed judgment by doing the work. Junior lawyers reviewed documents. Analysts built models. Marketers drafted early versions. Engineers debugged code. Much of that execution was inefficient, but it allowed people to see how decisions were made and how quality was assessed.

As AI absorbs more of the entry-level work, companies may unintentionally remove the training ground where future leaders developed judgment.

We cannot tell young employees to supervise AI systems without giving them the experience required to recognize when those systems are wrong.

Organizations will need to redesign apprenticeship deliberately. Junior employees should spend less time producing first drafts and more time comparing approaches, interrogating assumptions, observing consequential decisions, and understanding what separates a plausible answer from a sound one.

Managers will also need to change.

A manager who built authority by knowing more than everyone on the team may struggle when everyone has access to powerful intelligence. The manager’s value will increasingly come from setting direction, raising the quality of decisions, resolving tensions, and creating the conditions for others to exercise judgment.

This transition may be uncomfortable. It exposes how much leadership authority has been tied to information control.

The companies that benefit most from AI will be the ones that recognize what the skinny hamburger is telling us.

Faster execution creates a demand for better judgment.

More output creates a demand for sharper choices.

Greater autonomy creates a demand for clearer accountability.

AI can dramatically reduce the effort required to produce work. It also removes many of the natural constraints that once forced organizations to decide carefully.

The work in the middle is getting thinner.

What surrounds it now matters more than ever.