Google held its annual I/O developer conference today, and this year’s keynote had one clear message for enterprise buyers: the agentic era is not coming, it is here. The announcements were less about research milestones and more about production-ready tools that businesses can deploy right now.

Here is what matters for organisations evaluating AI strategies.

Gemini 3.5 Flash Is the Affordable Frontier Model You Have Been Waiting For

Google’s flagship announcement was Gemini 3.5 Flash, which is available today via Google AI Studio, the Gemini Enterprise Agent Platform, and Antigravity. The headline numbers are striking: 4x faster output than comparable frontier models, and roughly half the cost of running equivalent workloads on OpenAI or Anthropic’s top-tier models.

More importantly for enterprise teams, Gemini 3.5 Flash is built specifically for agentic tasks. It does not just answer prompts. It can handle long multi-step workflows, orchestrate teams of specialised sub-agents, and operate inside complex pipelines without losing context or coherence. Google’s benchmarks show it outperforming Gemini 3.1 Pro on coding, agentic, and multimodal tasks.

Google put a bold number on this during the keynote: Gemini 3.5 Flash has the potential to reduce enterprise AI inference costs by more than $1 billion annually at scale. That is a claim worth scrutinising, but even directionally it signals what Google is selling: the same capability for dramatically less spend.

Gemini 3.5 Pro is currently in testing and will launch next month, targeting teams that need maximum reasoning capability and can absorb slightly higher costs.

The Managed Agents API Changes How Businesses Deploy AI

One of the quieter but practically significant announcements was the Managed Agents API. Developers can now spin up a fully functional AI agent with a single API call. The agent reasons, uses tools, and executes code inside a secure Google-hosted Linux environment. Enterprise data privacy and governance protections are inherited automatically.

This is the no-ops path to agentic AI. Instead of standing up infrastructure, managing compute, and wiring up security policies manually, teams get a governed, production-ready agent environment out of the box. For IT leaders who have been watching the agent space but hesitating over security and compliance questions, this is a direct answer.

Gemini Spark: A 24/7 AI Agent That Actually Understands Enterprise Context

Gemini Spark is described as a personal AI agent that works around the clock on your behalf. It runs in isolated, ephemeral virtual machines with full encryption and data loss prevention policies enforced by Google Cloud. Critically, it requires explicit approval before taking high-risk actions like sending emails or modifying critical records.

What makes Spark genuinely useful for enterprise is the integration layer. It connects natively to Microsoft SharePoint, OneDrive, and ServiceNow, meaning it works in mixed-stack environments rather than requiring a full Google ecosystem. Practical use cases that Google demonstrated include monitoring system health via ServiceNow and automatically escalating critical issues, pulling Salesforce history and Zendesk tickets to prepare for sales calls, and managing project timelines across Sheets and Docs.

Gemini Spark is available in Gemini Enterprise now, with a Google Workspace preview arriving this summer.

Google Antigravity 2.0: The Agentic Development Platform

Antigravity, Google’s agentic coding and workflow platform, received a significant update. Version 2.0 ships as a desktop application for managing and orchestrating agents at scale, alongside an Antigravity CLI for teams that prefer code-first workflows. It is now integrated into the Agent Platform with Google Cloud’s enterprise security stack.

The adoption list is notable: Accenture, AirAsia, Deloitte, Monks, PwC, and WPP are all running Antigravity at scale. That breadth signals this is production tooling, not a pilot programme.

Google Workspace Gets Serious About AI

For the majority of knowledge workers, the most immediately relevant announcements were in Google Workspace. Voice capabilities are rolling out across Gmail, Docs, and Keep, enabling hands-free drafting and task management. Google Pics, an AI image tool with object-level editing, is being built into Drive, Docs, and Slides. A Google Workspace preview for Gemini Spark features arrives in summer 2026.

CodeMender: AI That Fixes Its Own Security Problems

Google announced CodeMender, an AI security agent that uses Gemini models to autonomously identify vulnerabilities in codebases, recommend fixes, test them, and apply patches across dependent systems. It requires human approval before deploying changes to production. For organisations running large engineering teams, this directly addresses one of the most persistent bottlenecks in software development: security remediation backlogs.

What This Means for Business

Google’s announcements today are a direct challenge to OpenAI and Microsoft’s enterprise foothold. The cost narrative around Gemini 3.5 Flash is calibrated to appeal to CFOs who have been shocked by their early AI infrastructure bills. The Managed Agents API reduces the technical complexity that has kept many enterprises in pilot mode. And Spark’s integration with Microsoft’s tools is a deliberate message: you do not need to rip out your existing stack to run Google AI.

For business leaders, the practical questions are worth asking now:

How much are you currently spending on AI inference? If your team is running heavy workloads on GPT-4o or Claude Opus, the cost comparison with Gemini 3.5 Flash deserves a serious look.

Are you actually running agents in production, or still in pilots? The Managed Agents API and Antigravity 2.0 are designed to close the gap between proof-of-concept and deployment at scale.

Is your team using AI where they actually work? The Workspace announcements are about reducing friction for non-technical users, which is where most enterprises still have the biggest adoption lag.

Google built the search engine that the world runs on. They built the cloud platform that handles billions of enterprise workloads. This year’s I/O suggests they intend to own the agentic infrastructure layer the same way. Whether they succeed against OpenAI’s brand momentum and Microsoft’s enterprise relationships will become clear over the next twelve months. But the tools they shipped today are real, available now, and priced to compete.

For organisations that have been evaluating their AI strategy, this is not a wait-and-see moment. It is a pricing-model moment. The cost of running capable AI agents just dropped materially.

---

Enterprise DNA helps business leaders navigate AI adoption with real data and practical deployment experience. If you are evaluating AI agent strategies for your organisation, start with a discovery call.