There is a visceral fear about AI taking our jobs. The latest Silicon Valley retort is that we will work for AI rather than being replaced. This creates the meat puppet economy, in which humans perform the physical or legal tasks machines cannot yet do. These include appearing in court, owning a patent, or signing a wet-ink contract. With no limit on AI agents, humans might eke out a living as the manual interface for a world run by algorithms.

West Coast thinking is evolving. A year ago, the goal was the one-person billion-dollar company. Today, the vision is a company led by AI and staffed by humans. Sam Altman has speculated that OpenAI is the natural first company to put an AI in charge.

Technically, AI is an ideal CEO. It possesses total data recall, zero emotional fatigue, and infinite scalability. It bypasses layers of bureaucracy, reviewing every line of code or every customer ticket without a filter. Human leaders are waking up to this, even without assuming it may be their roles at risk.

A January 2026 survey from Section showed that 19% of C-suite executives already save over 12 hours a week using AI. Only 2% of non-managers reported similar gains. The executives are automating their own analytical functions. If the CEO’s primary value is processing information to set strategy, the human is becoming the bottleneck.

The critical question is whether AI is setting the right goals. Efficiency is a tool, but the benchmark is the master. Companies must pursue the right outcomes whether they are run by humans or machines.

Some industries lend themselves to algorithmic control. In financial markets, where there is a lot of data and the rules are clear, price is the arbiter of success. Market participants are on constant look out for binary outcomes, which can be digitised and traded. Prediction markets are the latest example of the financialisaton of everything. But pure efficiency begins to breakdown in hybrid industries.

Healthcare is the cautionary tale for the AI-led future. Machines excel at diagnosis because it is pure pattern recognition. Yet the industry is often governed by efficiency algorithms that prioritise process over outcomes.

In the U.S., many physicians are managed toward a 15-minute consultation window. The implicit goal is to minimise liability and maximise billable referrals.

If an AI CEO were to manage this data, it would ruthlessly perfect the wrong goal. It would dismiss doctors who take 20 minutes to listen to a patient, ignoring the fact that those five extra minutes might uncover a life-saving alternative diagnosis. Without a fundamental shift in benchmarks, such as moving from throughput to five-year survival rates, an AI CEO is just a faster way to achieve the wrong result.

The physical economy remains the final frontier for AI autonomy. The Bureau of Economic Analysis estimates that roughly 12% of U.S. GDP comes from the digital economy. While this sector grows three times faster than the overall economy, it cannot yet fix a leak or wire a building.

In Silicon Valley, where productivity is defined by lines of code, the intelligence solves everything worldview makes sense. Machines do not need to pause for empathy or ask after a colleague’s well-being. But the rest of the world operates on relational value.

Productivity software that measures screen time misses the unproductive moments where trust is built. An AI CEO might view a long, difficult conversation with a frustrated client as a system latency to be eliminated, rather than the pivot point that saves a multi-million dollar contract.

Even before AI, consultants made fortunes creating metrics for managers to track. These metrics often destroyed trust by focusing on the means rather than the ends. The inappropriate application of AI would accelerate this process.

The more binary a business outcome, the easier it is to automate. But most high-value human work is not straightforward. The salesperson taking time to understand a client’s real needs is inefficient in the short term, but essential for long-term growth.

The true risk of an AI CEO is total devotion to the metrics we provide. If we continue to use 20^th century industrial benchmarks in 21^st century service industries, AI will destroy value. Speed, volume and short-term margin are not the best benchmarks for every industry. Before ushering in the era of the AI boss, we must first master the art of defining what a good outcome looks like.

I do not know how many of us will become entrepreneurs with armies of agents running businesses. Maybe most people will continue to work for companies and it is the boss who will change. What is clear however, is that in a world dependent on data, the right benchmarks are essential for delivering the optimal outcomes.