1 hour ago
1
Noam Scheiber
April 18, 2026 — 5:01am
Dan Sirk is a so-called fractional executive – meaning he works as the chief marketing officer for not just one company but two. Simultaneously.
It’s a juggling act made far more manageable by artificial intelligence tools such as Claude, Gemini and ChatGPT.
It used to take Sirk three to six months, or longer, to build a custom website with a team of contractors. Now, it takes him about a month, and he can do it by himself. Drafting a messaging strategy used to take a week.
When I spoke to him in March, he had just finished this task in less than eight hours. Thanks in part to these efficiency gains, Sirk is planning to become the chief marketing officer for a third company in the coming months.
And yet, when I asked if I should extrapolate from recent trends and assume he will add still more companies to his roster in the coming years, he looked at me as if I were crazy. He insisted that three was the outer limit of what he could handle, even with the help of AI.
“There are still human relationships,” he protested. Or to put it more bluntly: there are meetings.
The need for those sometimes messy human tasks may limit the number of people AI displaces.
Sirk estimates that he already attends 10 meetings in any given week across the two companies. There is a standing meeting with each team of top executives, not to mention a regular one-on-one meeting with each chief executive. There is a meeting with his own direct report and with the head of sales at one of the companies. And there are meetings about specific projects, like an upcoming presentation for one company’s investors.
Joining a third company is likely to increase the volume of meetings 50 per cent. If he became the chief marketing officer for even one more beyond that, Sirk said, he would be in meetings for almost the entire workweek.
Sirk’s experience, while perhaps extreme, reflects the broader impact of AI in the workplace: it is vastly accelerating many of the tasks conducted by white-collar workers and even replacing some of these tasks altogether. What it can’t automate – at least not yet – are the hard-coded requirements of bureaucracy.
With the help of AI, white-collar workers can generate far more memos or strategy options than in the past and churn out more product prototypes or software features. But some executive still has to decide which option to green-light. Workers can gin up many more sales pitches, but they still have to persuade clients to sign on the dotted line.
As AI makes the production of knowledge work more and more efficient, the job of presenting, debating, lobbying, arm-twisting, reassuring or just plain selling the work appears to be rising in importance. And the need for those sometimes messy human tasks may limit the number of people AI displaces.
“These were always important skills,” said David Deming, an economist who is the dean of Harvard College. “But as the information landscape becomes more saturated, the ability to tell a story out of it – to take a ton of text and turn it into something people want – is more valuable.”
Can you persuade your colleagues?
The idea that automation heightens the importance of personal interactions is not an entirely new one. A 2017 paper by Deming found that as computers became more powerful, a growing portion of jobs required heavy social interaction, while a shrinking portion required a lot of maths know-how but little social interaction — like certain engineering roles.
By automating technical tasks, computers were effectively pushing people into jobs that placed a premium on social skills, Deming said. That didn’t mean emotionally deft people were the most successful by default – the people who fared best tended to combine social skills with substantive knowledge – but it rearranged what employers valued.
In interviews, workers across a variety of white-collar professions said AI had supercharged this pattern. Many declined to be identified for fear of antagonising their employers.
A data scientist at a software company said he and his co-workers used to have to write code for every new feature or improvement they wanted to evaluate. Now they just come up with the idea, and the AI writes the code and runs the analysis.
His company’s interview process, which was once dominated by questions about coding and rewarded socially awkward nerds, now focuses on whether job candidates can identify good ideas and seem capable of persuading colleagues to back them, he said.
Mark Ozaki, a director at KPMG, said the consulting firm had traditionally encouraged younger consultants to specialise either in a subject area such as tax laws and regulations or in a technical area such as coding. But AI was devaluing this expertise and putting a premium on generalists who took the initiative and excelled at cultivating relationships with clients, he said.
Ozaki, who oversees a team developing an AI-based sustainability platform called Sustainlit.com, said his team had sometimes been at the mercy of skilled coders in the past. But it can now mostly use AI to do its coding, he said, and he primarily needed people “who have their phone glued to their head, who are everybody’s best friend, who are go, go, go”.
Other management consultants also underscored the growing value of social skills. Consultants at Accenture often used AI to help make slides for presentations, a manager there said, but the ones who excelled had absorbed the preferences of clients over many of hours of meetings. They know how the target of persuasion likes to consume information. Is he or she a metrics-driven person? Does the client like case studies or personal anecdotes?
A “customer success” worker at Salesforce said she was expected to use chatbots in her job coaching customers to use their sales software effectively and connecting them with technical experts when needed. Worried that she might be effectively training her AI replacement, she had been trying to make herself as “sticky” as possible to those customers, she added.
She makes a point of getting to know them beyond texts and email correspondence, often while schmoozing at site visits and conferences. She goes out of her way to provide emotional support, recently listening to a client who confided that she feared being laid off.
“I’ve had people just be vulnerable with me,” the worker said. “I know you cannot replace that with AI.”
(Salesforce said that AI had freed up employees to focus on priorities, like deepening relationships with customers, and that it had redeployed hundreds of employees to faster-growing areas.)
Goodbye, coders; hello, customer success
Cory Crosland, chief executive of PolicyFly, which sells software that helps insurers issue policies, said AI had reduced both the time it took to set up the software for new customers and the number of employees needed to do it.
In 2024, it took four or five PolicyFly employees an average of six months to get a new customer on board, Crosland said. The number of variables for each type of insurance policy and differences in the way insurers handle these variables meant that PolicyFly had to customise the software for each client.
Using AI to customise the software, a single PolicyFly employee can now get a customer on board in about two weeks, and Crosland expects that time to drop below one week this year.
The shift has allowed the company to charge much less money upfront, which appears to be increasing demand for its services. To keep up, PolicyFly has grown to 28 employees over the past six months, from 20, and only two of the new hires are software engineers. Several are younger employees who help set up customers or work in customer success, helping them get more out of the software.
Still, Crosland said he didn’t think he would be able to automate the process much further, at least not for the foreseeable future. The reason? His customers want to interact with a human.
The customers want PolicyFly to reassure them that the software will work under different situations, and that they have set up their billing properly or are prorating their policies in ways that make sense.
And, of course, there are the meetings to hash it all out – many, many meetings. “With the bigger companies, we have multiple people who are stakeholders weighing in from different departments,” Crosland said. “It’s even harder to get agreement and alignment on stuff.”
This article originally appeared in The New York Times.
Get workplace news, advice and perspectives to help make your job work for you. Sign up for our weekly Thank God it’s Monday newsletter.
