Week 21

Enterprise AI Weekly: May 19–25, 2026

By Attila · May 23, 2026

This week belonged entirely to Google. Google I/O 2026 ran May 19–20 at the Shoreline Amphitheatre in Mountain View and delivered over 100 announcements across AI, Android, Cloud, and security. Rather than recap all 100, this post focuses on the six announcements that matter most for IT decision-makers and sysadmins — the ones that will show up in your organization’s AI roadmap conversations over the next 12 months.

1. Gemini 3.5 Flash — The Enterprise Workhorse Model Is Here

What happened: Google kicked off I/O by releasing Gemini 3.5 Flash, the first model in the new Gemini 3.5 series. It is positioned as Google’s primary model for agentic and coding workloads — tasks that require multi-step reasoning, tool use, and sustained context over long operations. Google claims it outperforms Gemini 3.1 Pro on key benchmarks while running at speeds comparable to the Flash line. It is available today via Google AI Studio, the Gemini API, and Google Cloud’s Vertex AI.

Alongside Gemini 3.5 Flash, Google announced Gemini Omni — a multimodal model designed to accept any input type (text, image, audio, video) and generate any output type. Gemini Omni is positioned as a leap in world understanding and editing, with video generation as the launch capability.

Why it matters to sysadmins and IT decision-makers: If your organization is evaluating AI models for infrastructure automation, log analysis, or scripting workflows, Gemini 3.5 Flash is now the Google reference model to test. The “frontier intelligence at Flash speed” positioning is a direct shot at OpenAI’s o3-mini and Anthropic’s Claude Sonnet — Google is saying you no longer have to choose between capability and cost.

For Google Cloud customers already on Vertex AI, Gemini 3.5 Flash is available now with no migration required. For organizations evaluating model vendors, this is the model to put in your benchmark suite.

2. Gemini Spark — Google’s First Persistent, 24/7 Enterprise AI Agent

What happened: Google announced Gemini Spark, a personal AI agent that runs continuously in the background on behalf of users. Unlike a chatbot that responds to prompts, Spark operates autonomously — it can run recurring tasks, learn user preferences over time, connect to enterprise systems, and execute multi-step workflows without being prompted each time. It integrates with Google Workspace, Microsoft SharePoint, OneDrive, ServiceNow, and other business systems via Gemini Enterprise connectors.

Critically, Spark asks for approval before taking high-risk actions such as sending emails or modifying shared documents. Google has been explicit that it runs in isolated virtual machines inside a managed Google Cloud runtime, with an Agent Gateway that enforces Data Loss Prevention policies. Spark is rolling out to Gemini Enterprise and Workspace customers in the coming weeks.

Why it matters to sysadmins and IT decision-makers: Gemini Spark is the clearest example yet of what “agentic AI” actually means in an enterprise context — and it raises governance questions your organization needs to answer before rollout, not after.

The SharePoint and OneDrive integrations mean Spark can read and act on your corporate document repositories. ServiceNow integration means it can interact with your ITSM platform. These are not consumer integrations; they are production system connections. Before Spark lands in your Workspace tenant, you need clear answers to: What data can Spark access? Who controls the connector permissions? How are its actions audited? What is the escalation path when it takes an action that wasn’t intended?

Google has built the governance framework into the product — DLP enforcement, isolated execution, approval gates — but you still need to configure it, scope the connectors, and document the policy for your organization. Start that conversation with your Google Workspace admin and your security team now.

3. CodeMender — AI That Finds and Fixes Vulnerabilities Autonomously

What happened: Google integrated CodeMender into its Agent Platform and opened API access to external developers and security teams. CodeMender, originally developed by Google DeepMind and first shown in October 2025, is an AI agent that does not just identify vulnerabilities — it generates patches, tests them using specialized critique agents, and submits them for human approval before deployment. Over the six months Google has been running it internally, CodeMender has already submitted 72 security fixes to open-source projects, including codebases as large as 4.5 million lines. Google also announced a companion AI Vulnerability Reward Program with bounties up to $30,000, and Secure AI Framework 2.0 — an updated standard for securing autonomous AI agents.

Why it matters to sysadmins and IT decision-makers: Autonomous vulnerability remediation is a significant operational shift. The current workflow at most organizations involves a scanner finding a vulnerability, a ticket being created, a developer triaging it, writing a fix, getting it reviewed, and deploying it — a process that can take weeks or months for lower-severity findings. CodeMender compresses that workflow by handling everything up to human approval automatically.

The human-in-the-loop design matters here. Google has been deliberate that CodeMender surfaces patches for approval rather than deploying them autonomously, which makes it more realistic for regulated environments. The open-source track record (72 fixes already upstream) also gives security teams something concrete to evaluate rather than just marketing claims.

If your organization runs infrastructure on open-source software — and most do — CodeMender will eventually be relevant to your vulnerability management program. It is worth monitoring the Gemini Enterprise preview list.

4. Antigravity 2.0 — Agent Orchestration for Enterprise Development Teams

What happened: Google expanded Antigravity from a coding assistant into a full agent orchestration platform. Antigravity 2.0 includes a desktop app, a CLI, native Android vibe coding support in AI Studio, Google Workspace integrations for AI Studio-built apps, and a Managed Agents API — a single API call that spins up a full agent with persistent state. The platform is now available to Google Cloud customers via Agent Platform, not just Google’s internal engineering teams. Google DeepMind CTO Koray Kavukcuoglu described the expansion as: “turning it into a platform to develop and manage teams of autonomous AI agents.”

Why it matters to sysadmins and IT decision-makers: If your organization has developers building internal tooling, Antigravity 2.0 is now the Google-native alternative to Microsoft’s GitHub Copilot Workspace or Anthropic’s Claude Code. The Managed Agents API is particularly relevant for platform teams — it lets you deploy persistent agents into your internal infrastructure without managing the runtime yourself; Google Cloud handles that via Agent Platform.

The practical implication: development teams in your org will start asking about Antigravity access. That is a Vertex AI and Google Cloud billing question, not just an engineering decision. Get ahead of the procurement conversation by understanding how Agent Platform is licensed under your existing Google Cloud agreement.

5. Gemini Intelligence on Android — What It Means for Your Device Fleet

What happened: Google confirmed that Gemini Intelligence — the deep Android AI integration announced at the Android Show on May 12 — is rolling out this summer, starting with Samsung Galaxy and Google Pixel devices. Gemini Intelligence allows the AI to act across apps, understand screen context, fill forms using personal data (opt-in), place food orders, book reservations, and create custom widgets from natural language descriptions. Android 17 stable release is expected in June 2026. Googlebooks — a new category of Android-native laptops from Acer, ASUS, Dell, HP, and Lenovo — are launching this fall with Gemini Intelligence built in.

Why it matters to sysadmins and IT decision-makers: This is the one with the most direct impact on device management teams. Samsung Galaxy devices are in scope from day one, and Samsung is one of the most common corporate Android fleets globally. Gemini Intelligence gives an AI assistant the ability to act across apps on a managed device — reading screen content, submitting forms, and taking actions in Workspace apps.

Google has stated that users can disable individual Gemini integrations per app, and that sensitive actions require manual confirmation. But those are on-device user settings, not MDM-enforced policy controls. Before Gemini Intelligence rolls out to your corporate Android fleet, review your MDM configuration (Intune, Jamf, Android Enterprise) for any controls that scope or restrict AI assistant behavior on managed devices. If those controls do not exist yet, this is the right time to raise it with your MDM vendor.

Googlebooks adds another layer: Android-native laptops with Gemini Intelligence will eventually show up in employee device requests. Get ahead of it by understanding what Android Enterprise management looks like for laptop form factors before procurement starts.

6. Project Aura Smart Glasses — Coming This Year, With Real Enterprise Implications

What happened: Google and Xreal confirmed that Project Aura — XR smart glasses running Android XR with Gemini built in — will launch globally in 2026, with no specific date or pricing announced yet. The glasses feature a 70-degree field of view, three-camera hand tracking, and run on a Qualcomm processor combined with Xreal’s X1S spatial computing chip. They require a smartphone connection to operate. Google also separately announced audio-only AI glasses in partnership with Warby Parker and Gentle Monster — voice-first devices without a display, positioned as direct competitors to Meta’s Ray-Ban AI glasses.

Why it matters to sysadmins and IT decision-makers: Smart glasses with always-on cameras and AI assistants create a new category of endpoint management and policy questions that most IT teams have not addressed yet. The same concerns that arose with Google Glass in 2012 — recording in secure areas, access to ambient conversations, camera in meetings — are back, and this time the devices are actually going to ship at consumer scale.

Two questions worth raising with your security and HR teams now, before employees start showing up with these:

First, does your acceptable use policy cover AI wearables? Most corporate device policies cover phones and laptops but are silent on glasses. Second, does your physical security policy address camera-capable wearables in secure areas? If you have clean rooms, NOCs, server rooms, or client-facing spaces with confidentiality requirements, this is the year to update that policy.

The technology itself is impressive and the enterprise use cases (hands-free field support, real-time translation, AR-assisted infrastructure work) are real. But the governance framework needs to exist before the devices arrive.

The Bigger Picture

Google I/O 2026 was less a product launch event and more an architectural announcement. The thread connecting all six stories above is the same: Google is embedding AI into every layer of its stack — models, agents, orchestration, security, devices — and making that stack available to enterprise customers through Gemini Enterprise and Google Cloud.

The practical implication for IT teams is that AI is no longer something your organization evaluates in isolation. It is arriving through the tools you already own: Workspace, Android, Chrome, GCP. The governance frameworks, identity controls, and acceptable use policies need to be ready before the features land in your tenant — not after.

Google has confirmed infrastructure investment approaching $190 billion this year. That is not a company hedging on AI. Prepare accordingly.

Next edition publishes May 30. If your organization is navigating any of these announcements — Spark governance, Gemini Enterprise procurement, Android fleet policy updates — feel free to reach out.

More Enterprise AI Weekly coverage:

Enterprise AI Weekly: May 12–18, 2026 — Week 20
Enterprise AI Weekly: May 26–30, 2026 — Week 22