Google talks ‘singularity’ while scaling up agentic AI for enterprises

Google is recasting its enterprise AI roadmap around autonomous systems and AGI, with DeepMind CEO Demis Hassabis telling I/O attendees the industry now sits at the “foothills of the singularity.”

“When we look back at this time, I think we all realise that we were standing in the foothills of the singularity,” Hassabis said in his speech at Google I/O. “It will be a profound moment for humanity.”

The remarks capped a keynote spanning AI agents, cybersecurity systems, scientific research tools, coding platforms, and simulations — suggesting Google increasingly views AI not as standalone enterprise features, but as a broader operational platform capable of executing complex tasks across environments.

“AGI is now on the horizon, and it will be the most profound and impactful technology ever invented,” Hassabis said. “If built right, it could propel human progress and flourishing beyond our imaginations.”

While terms such as AGI and singularity have historically remained largely confined to AI research circles, Google’s keynote suggests those concepts are increasingly shaping how major vendors position long-term enterprise AI strategy.

Google pushes an agentic enterprise vision

Much of the enterprise AI market over the past two years has centered on copilots designed to assist employees with coding, productivity, and search tasks.

At I/O, however, Google repeatedly emphasized autonomous agents and long-running AI systems capable of orchestrating workflows, generating code, executing background tasks, and interacting across applications and environments.

Neil Shah, vice president for research and partner at Counterpoint Research, said Google is increasingly positioning its AI offerings as a unified enterprise platform rather than isolated AI tools.

“Google’s positioning to CIOs highlights its full AI stack capability powering a comprehensive enterprise-ready AI platform to build autonomous agent factories,” Shah said.

“Demis emphasizes that CIOs should look at this as a platform rather than singular AI tools to build their AI strategy,” he added.

That broader positioning was also reflected in Hassabis’ comments around AI safety and governance.

“It’s important that we are clear-eyed about the potential challenges and use all the tools at our disposal to ensure the safety of our agentic systems — and ultimately AGI itself,” Hassabis said.

Enterprise buyers face architecture and lock-in questions

For enterprise technology leaders, Google’s keynote may signal a broader shift from AI-assisted workflows toward increasingly autonomous enterprise architectures built around long-running agents and reasoning infrastructure.

Yugal Joshi, partner at Everest Group, said CIOs should interpret Google’s announcements less as a literal AGI prediction and more as a push toward “autonomous enterprise” architecture.

“Google wants them to use its platform not just for productivity and agents for specific tasks but to build end-to-end, well-orchestrated and governed agentic enterprise,” Joshi said.

The broader AGI framing could also reshape procurement and infrastructure decisions as enterprises commit more deeply to AI-native platforms.

“When a vendor positions their tech as a path to AGI, it is designing its AI strategy, which is future-proof but can introduce vendor lock-in, and that changes the math completely,” Shah said.

Cybersecurity and scientific AI emerge as proof points

Capabilities of AI models in cybersecurity are rapidly improving, with many now able to perform end-to-end, multi-stage penetration tests. These performance levels are doubling every five months or less.

Google sees a big opportunity here. “One area of risk that’s gained a lot of attention recently is cybersecurity,” Hassabis said in his speech, adding that Google is applying its “frontier capabilities and deep expertise to help secure the world’s code bases.”

Google introduced CodeMender, which Hassabis said can “automatically find and fix critical software vulnerabilities,” alongside a new CodeMender API for a select group of testers.

At the same time, Google heavily emphasized scientific discovery and simulation systems, areas Hassabis described as core long-term applications for AI.

“The whole reason I’ve worked on AI my entire career was because I saw it as the ultimate tool to advance science and our understanding of the world,” Hassabis said.

Hassabis also introduced Gemini for Science, a collection of AI tools designed to help researchers analyze papers, generate hypotheses, produce code, and accelerate scientific workflows. Google also highlighted simulation systems such as AlphaEarth Foundations and WeatherNext, which the company said improved hurricane prediction capabilities during the 2025 hurricane season.

The company further pointed to Isomorphic Labs, the Google-backed AI drug discovery company now working on treatments for immune disorders and cancer.

“Our mission is to reimagine the drug discovery process with the goal of one day solving all diseases,” Hassabis said.

While timelines surrounding AGI remain heavily debated, Google’s keynote suggests the company increasingly sees autonomous reasoning systems, rather than productivity copilots alone, as the long-term direction of enterprise AI.