The most revealing number in the technology industry this week came from Google CEO Sundar Pichai. He disclosed at Google Cloud Next 2026 that 75% of all new code at Google is now AI-generated and reviewed by human engineers. The figure was 50% last fall and 25% in October 2024. That trajectory — a tripling in eighteen months — matters more than the number itself.

Google engineers are no longer prompting AI to help them write code. They are directing autonomous AI agents that plan, generate, test, and iterate across entire projects.

Pichai described a complex code migration that agents and engineers completed together six times faster than the same work would have taken a year ago with engineers alone. Google has also tied AI adoption goals to engineer performance reviews. Using these tools has become a job requirement.

The direct question for every software professional reading this: if the world’s largest engineering organization no longer needs engineers to write most of its code, what does it need them for — and how many of them does it need?

Get the eBook of the Substack series that dissects the narratives Tech uses to build its AI Empire. $9.95 flat fee for PDF or ePUB formats; no subscription required. 3.5-hr reading time.

The people building artificial intelligence did not invent their ideas. They inherited them.

The Google disclosure lands alongside a body of research that is now documenting a quieter and more insidious workplace transformation. Researchers and employers have given it a name: AI deskilling.

The pattern is consistent across software engineering, financial analysis, legal research, and medical diagnostics. Workers who rely on AI for core job functions produce more output in less time. Their productivity metrics look strong. Underneath those metrics, their ability to perform the same functions independently is deteriorating.

A January 2025 study found that participants who received AI help during training performed substantially worse on subsequent independent assessments than those who worked through tasks without assistance. The AI completed the task. The worker never learned how to do it.

A Microsoft Research controlled experiment found that developers using GitHub Copilot completed coding tasks 55.8% faster. A separate Microsoft and Carnegie Mellon study of 319 knowledge workers found that AI tool use reduces critical engagement and raises concerns about long-term reliance and diminished independent problem-solving.

Future Forwarded is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Andrea Olson, writing in Inc. Magazine, frames the structural problem clearly: productivity dashboards stay green while the foundation erodes. The dashboards track output. They do not track whether a financial analyst still understands the assumptions behind the model AI just built for her. They do not flag that a junior lawyer can no longer construct a legal argument from scratch because he has been reviewing AI-generated briefs for two years.

Senior workers who built deep expertise before AI tools became ubiquitous retain that knowledge for now. They can use AI as a genuine augmentation layer because they understand their domain well enough to evaluate, correct, and direct AI output.

Junior workers are building their careers on top of AI from day one. Controlled studies suggest many are building on sand. When the tools change, workers without underlying domain expertise will find themselves stranded.

Also listen on Apple Podcasts

The deskilling problem connects directly to a structural shift in how AI companies are now selling their products. A 2026 industry analysis documents the emergence of what analysts are calling vertical AI: companies selling the outcome of labor rather than tools to help humans work.

Insurance, legal, logistics, and healthcare administration are the primary target sectors. The AI coding tools market alone reached $12.8 billion in 2026 revenue, more than double the $5.1 billion generated in 2024.

So, the dollar amount a company previously paid a human worker becomes the revenue target for a vertical AI startup. The labor budget becomes a software subscription.

MIT Technology Review’s roundup of the ten most consequential AI trends of 2026 identifies something the displacement data alone cannot capture: a global backlash is building.

Get the eBook compilation of Substack articles examining how the AI invasion already happened. You just weren’t invited. $9.95 flat fee for PDF or ePUB; no subscription required. 2-hr reading time.

Labor unions, artists, conservative legislators, and progressive activists are converging on AI regulation from different directions and for different reasons. Small regulatory wins are accumulating.

Connecticut passed worker-protection legislation this month. Colorado’s AI law takes effect in June. The federal government has produced a framework with no binding obligations, and states are filling the vacuum.

The backlash is the political story running underneath all the economic data. Workers are sensing displacement before the aggregate statistics confirm it.

A Mercer survey of 12,000 workers and executives globally found that concern about AI-driven job loss rose to 40% in 2026, up from 28% in 2024. Ipsos data on U.S. public opinion finds that almost three in four Americans believe the government should act to prevent AI-induced job losses.

The gap between that sentiment and the current state of federal policy is the space where the next few years of labor politics will be challenged.

Get full access to Future Forwarded at futureforwarded.substack.com/subscribe

Get full access to The AI/Labor Report at ailabor.substack.com/subscribe