Meta exec quit after AI agents replaced entry-level workers

A veteran technology executive who helped build AI products for Salesforce and Meta has left Big Tech to address a problem she helped create: the rapid displacement of entry-level workers by artificial intelligence agents.

Clara Shih spent two decades building enterprise software, most recently leading Meta's business AI group and Salesforce's Agentforce platform. But last fall at Meta, she witnessed something that changed her career trajectory. Product development processes that once required user researchers, designers, product managers, and three types of engineers collapsed into work that one or two people could handle with AI assistance.

"Seeing is believing, and in that moment, I just imagined this amplifying across the economy," Shih said in an interview with Platformer. The pattern repeated across marketing, distribution, and policy review. She began removing entry-level job postings because she no longer needed them.

This spring, Shih left Meta to launch the New Work Foundation, a nonprofit focused on helping workers navigate AI-driven job disruption. The organization offers free tools including a podcast featuring hiring managers, a data platform showing AI exposure by occupation, and a mentoring app for job seekers.

Why it matters

Shih's account offers rare candor from someone who built the automation tools now reshaping labor markets. While many tech leaders maintain that AI will simply change jobs rather than eliminate them, Shih acknowledges her earlier optimism—that automation would free workers for higher-order tasks—has "primarily not been true." Her estimate that one in five corporate roles face significant disruption comes from direct operational experience, not academic speculation.

Three scenarios for AI job impact

Shih outlined a framework from MIT economist David Autor describing how AI reshapes work. In the first scenario, when AI automates a large percentage of job tasks—as in translation or customer service—direct displacement occurs. The second, optimistic case happens when only routine tasks get automated, freeing workers for more complex responsibilities, as occurred with software engineers over the past two decades.

The third scenario creates more jobs but degrades their quality. When AI automates expert tasks rather than routine ones, barriers to entry drop suddenly, flooding the labor market. Shih cited taxi drivers after GPS and rideshare platforms emerged: more jobs existed, but wages fell below living standards in major cities.

Salesforce CEO Marc Benioff said Agentforce helped the company reduce customer support staff from 9,000 to 5,000 people. Meta's Business AI handles customer service across WhatsApp and other platforms, potentially reducing staffing needs for businesses using those tools.

Who faces the greatest risk

Shih identified roles focused on preparing artifacts for internal review—drafting briefs, creating slide decks, generating order forms—as particularly vulnerable. She estimates these represent about one in five corporate positions. When the final decision-maker can use AI directly, intermediate preparation roles lose value.

Contrary to widespread concern about software engineering, Shih predicts the field will expand. She believes the algorithmic thinking trained in computer science programs matches the skills needed to manage agent systems and evaluate their outputs. Her "hot take": software engineers will increasingly take over legal, marketing, and accounting work.

The details were first reported by Casey Newton in Platformer.