Google I/O 2026: What Was Announced, What It Means, What to Do

Sundar Pichai opened the May 19 keynote with a line worth taking literally: "We're now in the part of the AI cycle where people want to see the value in the products they use every day." Ten years after Google declared itself AI-first, the I/O 2026 announcements were not aspirational. The new model shipped to 900 million users before the keynote ended.

This post covers what was announced, what it means for business operations and customer acquisition, and what enterprise leaders should do differently in the next 90 days.

What Google Announced at I/O 2026

The headline model is Gemini 3.5 Flash, which Google launched as the default across the Gemini app and AI Mode in Search for all users, including the free tier, on the day of announcement. It outperforms the previous Gemini 3.1 Pro on coding and agentic benchmarks while running four times faster than comparable frontier models. Gemini 3.5 Pro is in testing and expected next month.

Alongside the model, Google introduced Gemini Omni, a multimodal model that accepts image, audio, video, and text input and produces video output grounded in real-world knowledge. It is available in the Gemini app starting at launch. For developers, Antigravity 2.0 is Google's agent-first development platform, replacing earlier tooling with a desktop application, a command-line interface, and built-in cross-platform sandboxing. It is free.

Search received its most significant structural update in years. The new intelligent search box expands as users type, anticipating intent rather than completing keywords. AI Mode, powered by Gemini 3.5 Flash, now includes Information Agents that monitor topics on a user's behalf continuously, not just at the moment of a query. Universal Cart introduces agentic commerce: a shopping cart that tracks price drops and inventory across retailers and acts on the user's behalf.

In Google Workspace, Gemini Spark is a 24-hour personal agent that manages email, calendar, and tasks across apps. Daily Brief, rolling out to all AI subscribers starting May 19, delivers a morning summary of inbox, calendar, and priorities with suggested next steps, generated overnight. Google Pics is a new AI image generation and editing application. Gmail Live and Docs Live bring voice-to-document capabilities to mobile.

On hardware, intelligent eyewear developed with Samsung, Warby Parker, and Gentle Monster launches as audio-only glasses this fall, with display glasses following. Google Beam, the 3D video collaboration product from HP, is in production with Deloitte and Salesforce as early adopters. Subscription pricing shifted: AI Ultra now starts at $100 per month, down from $250, with a $200 tier matching the previous plan's capabilities.

Search Is Now a Channel You Do Not Control, and Customers Are Already There

The business consequence of AI Mode is not a search ranking problem. It is a customer discovery problem. When Information Agents monitor topics and Universal Cart evaluates products across retailers continuously, the customer's first interaction with your category may happen without a direct query to your website, without a click, and without your sales or marketing team knowing it occurred.

In November 2025 I covered Google's agentic calling and checkout as the first evidence of AI inserting itself at the transaction layer. I/O 2026 extended that logic upstream. The research phase, the comparison phase, and the shortlist phase can now all happen inside Google's interface. A business that is not represented accurately in the data Google's agents can read may not appear in that process at all.

This has direct budget implications. Organic search was a near-zero marginal cost acquisition channel for many mid-market companies. As AI-synthesized answers reduce click-through to external sites, that volume does not disappear. It migrates to paid channels where competitors are also bidding harder, or it never leaves Google's interface at all. The CFO conversation about this belongs in the next planning cycle, not the one after it.

Antigravity 2.0 and A2A Change the Enterprise AI Build-vs-Buy Calculation

For CIOs, the most consequential announcement was not a model. It was the Agent2Agent protocol, abbreviated A2A. Google built it on open web standards: HTTP, JSON, and standard authentication tokens. That means an enterprise can build an AI agent workflow that spans Google systems, Microsoft systems, and third-party platforms without the orchestration layer being owned by any single vendor.

Proprietary agent protocols lock the orchestration logic to the vendor. A2A does the opposite. For a CIO signing multi-year platform agreements in 2026, the question worth asking before signature is whether the competing vendor offers equivalent interoperability commitments in writing, not just in a product roadmap.

Antigravity 2.0 pairs with A2A as the build environment. It is free, ships a desktop application and a command-line interface, and includes built-in sandboxing and credential masking. The practical implication for enterprise software teams is that the barrier to standing up a capable agent development environment dropped significantly at this event. Teams that have been waiting for the tooling to mature have less justification for that position after May 19.

Gemini 3.5 Flash at $1.50 per million input tokens makes high-volume agent use cases economically viable in ways that earlier pricing did not. Internal automation projects that previously struggled to show return on investment at $10 or $15 per million tokens look different at this price point.

What the Pricing Change Actually Signals

Google cut AI Ultra from $250 to a $100 entry price. Read that as a competitive signal, not a generosity signal. When the entry price for a capable AI subscription drops by 60%, it compresses the pricing leverage of every competitor in the mid-market. It also tells enterprise procurement teams that the model access layer is commoditizing, and differentiation is moving to the services, integrations, and agent ecosystem built on top.

For enterprise software buyers, this matters because it changes which features justify a premium. Storage, agent limits, and integration depth matter more than raw model performance when two or three platforms are all running frontier-class models at similar price points. Evaluation criteria for AI platform renewals in the second half of 2026 should reflect that shift.

What Enterprise Leaders Should Do in the Next 90 Days

For CMOs: run a direct attribution audit on organic search. Isolate what percentage of top-of-funnel pipeline currently originates from unpaid Google traffic. Model what a 30% reduction in that volume costs in paid channel substitution. If you do not have that number, you cannot assess the exposure.

Review how your product data is structured for machine readability. Universal Cart and Information Agents evaluate products based on what structured data they can access. Schema markup, product feeds, and pricing data accuracy are now customer acquisition infrastructure, not SEO hygiene.

For CIOs: evaluate Antigravity 2.0 against your current agent development environment. If your team is building on proprietary tooling, verify whether the vendor has published interoperability commitments comparable to A2A before the next renewal. Governance frameworks for agent deployments are significantly cheaper to build before production scale than after. Gemini 3.5's day-one free-tier availability means your employees are already using it, with or without a sanctioned deployment.

For both roles: the Gemini Spark and Daily Brief announcements signal that Google is positioning its AI as the operating system for knowledge work, sitting above individual applications. If your organization has a significant investment in a competing productivity suite, the competitive pressure on that vendor increased at I/O 2026. That is worth monitoring at the contract level.