Google vs OpenAI: The Workplace Agent Coworker War Has Started

The video opens by covering Google's annual Cloud Next event in Las Vegas, where the headline announcement was 'Workspace Intelligence.' Powered by Gemini, this feature understands semantic relationships across Google's suite of tools—Docs, Gmail, Slides—as well as third-party platforms like Asana, Jira, and Salesforce. Three standout capabilities were highlighted: an interactive Sheets canvas for building dashboards and Kanban boards, infographic creation inside Google Docs, and custom-built automation 'skills' that allow teams to codify recurring workflows and publish them across Workspace with one click. The host frames Google's structural advantage as its deep integration with tools people already use daily, arguing that rivals like ChatGPT and Claude can only approximate this through connectors.

OpenAI responded in the same week with its own workplace agents, powered by Codex, capable of taking real actions—writing code, updating CRMs, and sending emails—rather than just providing information. The host used OpenAI's own sales team as a case study, where an agent that previously required 5–6 hours of manual work per week now runs automatically. A product feedback agent example was also shown, which reads web feedback, groups it into issues, posts Slack summaries, and updates Linear tickets. The host notes that Google is playing a structural game while OpenAI is competing on speed and reach, and that Anthropic's Claude is also shipping workplace features rapidly. Neither company has clearly won the enterprise AI race yet.

The video then covers Shopify CEO Tobi Lütke's project called Flipbook—an infinite visual browser that generates pages entirely as AI-produced images on demand, with no HTML or pre-built links. Though slow and early-stage, the host raises conceptual questions about what happens to UI systems, tutorial tools, and adaptive learning platforms if interfaces become dynamically generated per user interaction. A new email app called Extra, built by ex-Pinterest leaders, is also briefly mentioned as a potential inbox management tool for Gmail users.

On data and trends, the host discusses internal Google AI adoption figures leaked by former Google engineer Steve Yegge, who claimed only 20% of staff are power users, 60% use basic chat tools, and 20% don't use AI at all—and that DeepMind reportedly uses Claude rather than internal Gemini products. Demis Hassabis disputed these figures, though Yegge says internal Googlers corroborated them. Separately, Sundar Pichai revealed that 75% of Google's new code is AI-generated, painting a somewhat contradictory picture of AI adoption internally.

OpenAI's Codex growth data was highlighted as impressive: 4 million weekly active users, up from 3 million two weeks prior and 2 million a month ago, supported by partnerships with Accenture, Capgemini, and Bain. Finally, a Bain & Co. report on AI search behavior found that 44% of US online buyers now start their purchase journey in an LLM or split between AI and traditional search. Critically, Bain's analysis of 500 million AI citations found that 89% of responses to unbranded prompts drew from third-party sources like G2, Reddit, TechCrunch, and analyst reports—not brand-owned websites—a key strategic implication for product and marketing teams relying on SEO.