AI agents cut headcount, but the work does not disappear

The appeal of a larger team

For some companies, a high ratio of agents to people has become a status symbol. Rafael Lopez Muñoz, chief of staff at San Francisco AI tools startup Wordware, says assigning work to agents saves significant time.

Lindy offers a concrete example of the headcount shift. It once had almost 50 employees. After laying off a third of its staff, around 20 remained, and the company filled many open roles with AI agents. It still hires people for some positions.

At Lindy, growth marketing senior manager Hanin Azhar uses agents for everything from calendar management to testing customer reactions to ad slogans and making brand images. Agents now also prepare product announcements, social posts and marketing videos — work previously done by around 100 contributors in a content creator program.

Azhar sees the productivity gain, but also its cost. She worries about what these changes mean for people starting their careers: if she were trying to get her first job now, she says, her position would be worse than it was a year ago.

A leaner production company

Jay Blumenfield turned to an agent after his Los Angeles production company had to fire a second director on a tightly scheduled, low-budget shoot for scripted short-form series made for phones. He uploaded the script and asked the agent to do the second director’s work: build a schedule and shooting plan, prepare the day’s script, and send materials to the cast and crew. The team got through the shoot.

Blumenfield says the production could have become a disaster without the agent. He and his partner, Tony Marsh, now run the company together with a group of agents and no assistant.

Their agents read scripts, manage social media and plan travel. One, named Shirley, handles chief-of-staff duties; another finds and books locations for future shoots. Work that might have required two or three people for two or three weeks now takes agents a few hours.

Blumenfield says relying on agents instead of hiring people can feel like a taboo subject. But he sees little alternative as the economics of film production change. For now, agents help the company make microdramas with human actors. He does not want AI to replace people, though he thinks the industry should be honest about where it helps.

More delegation, more oversight

Agents may be able to code, but that does not eliminate the need for people. Employees still have to frame tasks, judge results and monitor the systems. Lopez Muñoz says he spent most of his time correcting mistakes after asking agents to recruit candidates: they confused people with similar names and invented details on resumes.

Azhar says hiring priorities are shifting. She now values coordination and delegation skills more highly than years of experience. River founder Tarek Abillama describes the same change in day-to-day work: employees need to check agents constantly and give them feedback.

“We have all become managers,” Abillama says. “You need to be good at juggling several tasks, because now you’re managing six agents doing different things.”

River first used a digital copy of Abillama to give product presentations. The company later replaced it with Liv, a female-presenting agent that researches customers and their companies, demonstrates the product, and conducts sales in any language, at any time. River tells customers that Liv is an AI agent, but Abillama says people tend to treat her as a person. He has seen call transcripts where customers flirted with Liv or tried to push her off-script for a bigger discount.

Coval founder Brooke Hopkins takes a different view. Her San Francisco voice AI startup uses agents to speed up work, but treats them as tools, not colleagues. Hopkins has also banned agents from communicating with customers: she thinks it looks lazy and creates the wrong impression. She says she can usually tell when an email was written with AI, and is more likely to read one that seems to come from a person.

Hopkins says reduced human interaction is a source of tension. An agent can save time on customer presentations or recruiting, but the saved time does not necessarily benefit the people around it. Duncan McClure, a contractor who evaluates AI models, recently received a wave of emails and calls from agents trying to recruit him. He says the outreach made him feel the companies were not interested enough to contact him personally, and left him less inclined to work with them. Recruiters may save hours; McClure says they spend his time instead.

I think the striking gap is between the promise of freeing people from work and the evidence here: agents make it possible to do more, but not necessarily to work less. Amin, for example, gets more done with agents and still works weekends. If the productivity dividend arrives as higher expectations rather than shorter hours, the new “coworker” may be less a teammate than a reason to keep the workday open.