Last August, NetDragon Websoft — a Hong Kong-based online gaming firm with $2.1B in annual revenue — appointed a CEO to helm its flagship subsidiary.
The new chief, Tang Yu, was responsible for all of the typical duties of a company figurehead: reviewing high-level analytics, making leadership decisions, assessing risks, and fostering an efficient workplace.
She worked 24/7, didn’t sleep, and was compensated $0 per year.
But there was a catch: Yu wasn’t a human. She was a virtual robot powered by artificial intelligence.
So far, having an AI CEO hasn’t had any catastrophic consequences for NetDragon Websoft. In fact, since Yu’s appointment, the company has outperformed Hong Kong’s stock market.
Zachary Crockett / The Hustle
As AI tools have become more robust, automation experts have philosophized about replacing vast swaths of workers.
McKinsey Global Institute recently predicted that 45m workers, or ~28% of the entire American workforce, would lose their jobs to automation by 2030.
Most automation efforts have been centered around eradicating so-called lower-level and blue-collar jobs like warehouse workers, truckers, clerical assistants, and food prep workers. More recently, AI has threatened white-collar roles like accountants and journalists.
But while executives at the top of the corporate food chain celebrate the cost-cutting virtues of AI displacement, they rarely seem to turn the spotlight on themselves.
The incentives for workplace automation are largely financial. So why not start by replacing the highest-paid employee of them all — the CEO?
At Fortune 500 firms, the average CEO pay is now ~$16m per year.
Over the past 45 years:
As a result, today’s average CEO is paid the equivalent of 399 median workers.
At larger companies, this ratio is often many multiples higher: For instance, in 2021, Amazon CEO Andy Jassey received a package worth $213m — equal to the collective wages of 6,474 Amazon employees. That’s enough workers to fully staff four fulfillment centers.
Zachary Crockett / The Hustle
Even die-hard free market capitalists have had trouble justifying these pay packages.
Most economic scholars who have studied CEO pay have concluded that executives have substantial “rents” — that is, they earn far more than what they give back by measure of productivity.
Research has shown that there may actually be an inverse relationship between CEO pay and long-term company performance.
One study examined executives at 400 firms between 2006 and 2015 and found that:
Zachary Crockett / The Hustle
One reason for this is that executive pay structures incentivize CEOs to chase short-term profits rather than meaningful long-term growth.
Most of a CEO’s compensation is dependent on boosting metrics like earnings-per-share, which can be fudged through maneuvers like stock buybacks. The result of this is that CEOs are often handsomely rewarded even when they lead their companies to abysmal financial outcomes.
Among the many recent examples:
While many CEOs have gotten more expensive and less effective over time, technology has simultaneously become cheaper and more reliable.
Replacing CEOs with AI would not only save firms millions of dollars in payroll costs, but would minimize, or altogether eradicate, the personal motives that often lead to less-than-ideal corporate outcomes.
In late 2022, the Organisation for Economic Co-operation and Development (OECD) — an international consortium that aims to “stimulate economic progress” — put together a report analyzing the likelihood of automation affecting different occupations.
CEOs (categorized as “top executives”) were nearly dead last, right next to religious workers.
Zachary Crockett / The Hustle
“My gut feeling is that CEO will be the very last job to be automated,” Marguerita Lane, a labor economist with OECD, told The Hustle.
Lane says that most of the components of a CEO’s role that can’t be replicated by AI are rooted in the human touch: being a figurehead for accountability, selling a vision, communicating with the public, negotiating.
A large part of a CEO’s job is to essentially serve as a company mascot — and mascots are fairly AI-proof.
But that’s not to say that many of the other elements of a CEO’s job can’t be automated.
McKinsey has estimated that ~25% of a CEO’s time is spent on tasks that machines and/or AI could potentially replicate — reviewing financial performance, sending emails, forecasting trends.
Some CEOs have openly admitted to automating a much larger portion of their job by outsourcing the bulk of their responsibilities to other workers.
Several years ago, an American entrepreneur named Christine Carrillo, who bills herself as “The 20 Hour CEO,” posted a thread on Twitter detailing how her executive assistant — hired in the Philippines, where the average salary is $9.5k/yr— performed 60% of her duties, including:
As Will Dunn at The New Statesman later wrote: “If most of a CEO’s role can be outsourced, this suggests it could also be automated.”
Zachary Crockett / The Hustle
Another area of potential automation is the executive decision-making process.
Each year, executives make ~3B decisions — and there is a direct link between the effectiveness of these decisions and a firm’s financial performance. So, there’s an incentive for optimizing the rate of success as much as possible.
CEOs are often heralded as decision-making geniuses. But by their own admission, they aren’t much better at making good decisions than the rest of us:
Executives are already increasingly relying on aid from algorithms and machine learning to improve these ratios.
A new field of machine learning called decision intelligence automates and augments the executive decision-making process by “linking data with decisions and outcomes.” Firms like IBM, Google, and Alibaba have all jumped into the space in recent years.
Some scholars are skeptical that such tools could model the contextual complexities of executive problem-solving.
Oded Netzer, a professor at Columbia Business School who specializes in text-mining techniques, estimates that current tools could probably automate “a good 30%-40%” of executive tasks. But he argues that human decision-making requires contextual awareness that AI can’t replicate.
“For AI to work, it needs to train on data,” he says. “The less repetitive a job is, the harder it is to collect data on. Each executive decision requires different inputs and considerations that make it hard to apply a predictive framework.”
What does AI itself think about all of this?
We asked ChatGPT, the powerful AI chatbot that is currently rekindling fears of automation, to ruminate on the likelihood of a great CEO replacement.
It admitted it wasn’t up for the job yet — at least not “in the near future.”
ChatGPT thinks that human CEOs will be around for a while (OpenAI)
Technological feasibility aside, there are other blockades to automating CEOs.
Part of what makes some jobs less likely to be automated is bargaining power — and CEOs are extremely good at convincing shareholders that they are indispensable.
“If there were a proposal to replace them with AI, CEOs would be very well-positioned to protect their interests, much in the same way they are in salary negotiations,” says OECD’s Lane.
Even if AI were capable of entirely replacing the CEO, Lane imagines that social norms would serve as a barrier of protection.
“If two roles inside a company suddenly became automatable — one an executive position and the other an entry-level position — the employer would still let the entry-level worker go first,” she says. “Because the value that we have ascribed to CEOs exceeds that of almost any other worker.”
Nonetheless, the broader workforce isn’t entirely opposed to getting rid of their bosses:
And at least one prominent voice thinks the clock is ticking for CEOs.
Several years ago, Jack Ma, the Chinese billionaire who co-founded the Alibaba Group, hypothesized that the emotionless logic and efficiency of AI would eventually find its way into the corner office.
In 30 years, he suggested, “a robot will very likely be on the cover of Time magazine as the best CEO.”
If NetDragon’s bot-in-chief can keep outperforming the stock market, that prediction might not be so far-fetched.