AI coding pushes startups toward leaner teams

Small startups are increasingly using AI coding tools to build products with much smaller teams, creating a sharper divide between experienced developers who can direct these systems and entry-level programmers trying to enter the industry.

At Giftory, the online gift-giving platform he runs, CEO Eric Lauer is looking for “mid-career people who are lazy in a smart way”—experienced developers he describes as architects who understand software workflows and can use AI to multiply their output.

“To be an architect, you need that previous world experience, and you need to know all the workflows.”

Candidates with “no knowledge whatsoever of the workflows” are “a lot less appealing” to his company, Lauer told AFP. Tools such as Anthropic’s Claude Code and OpenAI’s Codex are changing programming from line-by-line implementation into a management task: developers prompt systems to write, test and fix software, then oversee the resulting work.

Recommended reading

A Humanoid Robot Lost Its Head—and Kept Fighting

AI coding adoption at startups

The shift is already widespread. In a survey of developers at small startups by The Pragmatic Engineer, 75% reported using Claude Code. Jared Friedman, managing partner at Y Combinator, said a quarter of startups in the Winter 2025 batch were built on code that was 95% AI-generated.

Giftory’s roughly 30 employees each receive a premium AI subscription costing about $200 a month. Lauer called that “peanuts” beside an average annual salary of $100,000, and said the economics are strong enough to make offshoring “uncompetitive.”

Haitham Mengad, co-founder of Stems Labs, said his company also chose to increase output from its existing staff rather than expand quickly.

“We already had a pretty lean team and very talented engineers, so the approach I took was, let’s do more with the people that we have.”

At software company Espresa, AI is saving the customer-success team “millions of dollars a year,” according to Lindsay Euller, its vice president of customer success. She said future headcount requests could face a new condition: managers may first be asked how they are optimizing AI before a new hire is approved.

Entry-level software jobs are under pressure

The productivity gains are arriving alongside weaker prospects for younger programmers. A Stanford Digital Economy Lab study of payroll data from millions of U.S. workers found that employment among 22- to 25-year-olds in occupations most exposed to AI, including software development, had fallen nearly 20% from a late-2022 peak.

Harvard researchers examining résumé and job-posting data covering about 62 million U.S. workers across 285,000 firms found that junior employment at companies adopting generative AI fell roughly 9% relative to nonadopters within six quarters. Senior employment continued to rise.

Ian Amit, CEO of cybersecurity startup Gomboc AI, said companies are interviewing candidates but hesitating to make offers.

“I think there’s a lot of hesitation in hiring right now. I’m hearing of a lot of companies that are interviewing multiple candidates across the board but are not pulling the trigger on actual hiring decisions.”

Amazon Web Services CEO Matt Garman has called replacing junior developers with AI “one of the dumbest things I’ve ever heard,” warning that the industry could lose its next generation of leaders. Yet computer-science enrollment is already declining, down 6% across the University of California system and falling at two-thirds of computing programs nationwide, according to the Computing Research Association.

For now, startups still see lean teams as the rational choice. Giftory’s Lauer said the company remains in a “hypergrowth phase” and continues weighing whether to invest in more resources or more people. In the tech sector, AI is increasingly winning that trade-off.