AI-generated code – Google CEO Sundar Pichai says 75% of new code is AI-generated at the company. The shift to agentic workflows could reshape productivity, roles, and how companies deploy AI.
Google CEO Sundar Pichai says a majority of the company’s new code is now AI-generated, approved by engineers.
That “AI-first” momentum matters beyond the tech floor: it changes how software is built, how teams measure productivity, and how the workforce adapts as code generation moves from a manual task to an automated workflow.
Pichai’s Wednesday blog post put a clear number on the internal transition.. He said Google has been using AI to generate code internally for some time. and that today about 75% of new code is AI-generated and approved by engineers—up from roughly half the figure reported the previous fall.. In practical terms. it signals that AI is no longer just assisting developers with suggestions; it’s actively participating in producing new software changes.
The shift from code drafting to “agentic” work
The most important part of Pichai’s message is the direction of travel: Google is moving toward agentic workflows.. Instead of AI being limited to producing snippets of code. Google describes a system where engineers orchestrate autonomous digital “task forces. ” with agents helping complete work more independently.
Google’s leadership also framed this as a productivity story with a measurable edge.. Pichai cited a “complex code migration” where the work was completed six times faster than was possible with engineers alone just a year earlier.. The implication is not only that AI can draft code more quickly. but that it may reduce the time spent on iterative groundwork—review cycles. rework. and the manual coordination required to move from an idea to working software.
For the broader business audience, that speed question is the real battleground.. Companies don’t just buy AI for convenience; they buy it to compress timelines—whether that’s shipping features sooner. migrating systems with less downtime. or reducing the engineering effort required to evolve platforms.
Engineers move up the value chain
Google’s own internal narrative goes further than “faster coding.” Richard Seroter. a Google Cloud senior director and chief evangelist. described a shift in what engineers do day to day.. If humans are still approving AI-generated code, then engineers can focus more on system architecture, design, and complex problem-solving.
That’s a significant human impact, because job roles rarely change overnight—but they do change in responsibilities.. In Google’s framing. “software engineer” is becoming closer to “product engineer” or “architect. ” with less emphasis on manual coding and more emphasis on shaping what the system should do and how it should behave.
There is also a cultural point here: as time constraints loosen for engineers and AI helps explore “an endless array of ideas. ” teams can test more paths without paying the same cost in hours.. In business terms, more experimentation can translate into stronger product differentiation—assuming quality controls keep pace.
Productivity gains extend beyond engineering
This isn’t limited to developers. Pichai said Google’s marketing teams used AI models to rapidly generate thousands of variations of creative assets—work that might otherwise take weeks. He also tied the effort to outcomes: 70% faster turnaround and a 20% increase in conversions.
That combination—speed plus measurable commercial lift—signals why AI adoption accelerates when it touches revenue drivers.. Marketing is where many companies feel the pressure to iterate quickly, respond to customer behavior, and optimize at scale.. If AI can increase the volume of testable creative without exploding labor costs, it becomes easier to justify broader deployments.
Still. the underlying business risk stays familiar: AI-generated outputs can create consistency problems. brand drift. or compliance issues if governance lags behind capability.. Google’s emphasis on human approval of code and human steering of work hints at how the company is trying to manage that risk.
From copilots to managed agency
A key phrase in Seroter’s comments is that the “experimental phase” of simple copilots is over.. He pointed to limitations of earlier tools—context-unaware chatbots and simple “please start this for me” requests—and argued the market is moving toward AI that completes relevant work. guided by human operators.
Google also described its next phase as “managed agency. ” where the company provides a governed. enterprise-ready harness to build and scale autonomous agents.. The practical difference is governance: enterprises want visibility into what agents do. how they decide. and how those actions align with internal policies.
What the investment wave could mean for markets
Google’s push is unfolding alongside major infrastructure bets.. At its Cloud Next 2026 event, the company announced new AI chips and a Gemini Enterprise Agent Platform.. Pichai also said Google plans to invest up to $185 billion in infrastructure to power autonomous AI agents.. Separately, Google Cloud has signed a multibillion-dollar deal with Thinking Machines Lab to expand its AI infrastructure.
For investors and industry watchers, those figures are less about curiosity and more about competitive positioning.. Infrastructure spending signals a belief that AI agents will become a standard part of enterprise operations—not a niche experiment.. As more computing capacity is directed to agentic workloads. the winners may be companies that can pair compute with reliable software delivery. security controls. and measurable business outcomes.
Looking ahead. Google’s focus on “agent-first experiences” suggests it wants interfaces and workflows where AI is not just consulted. but effectively deployed as a worker.. For businesses. that raises both opportunity and urgency: teams that redesign processes early may capture productivity advantages. while laggards risk paying a higher cost to retrofit automation later.
U.S. arrests soldier over Polymarket bets tied to Maduro operation
Meta to lay off 10% of staff on May 20—what it signals for AI spending
Bob Iger returns to Thrive Capital as adviser—what it signals for media and AI