Last week, 32,000 people came to Las Vegas for what’s next…turns out, quite a lot of it.

Zazmic had a team on the ground at Next '26 too — as a Google Premier Partner and one of Google's AI delivery partners, this is one we don't miss. A week back, heads still full of it, here's what we think you should know.

Everything You Need to Be an Agentic Enterprise

The pilot era is over

If last year's Next was about possibility, this one was about production. Google Cloud CEO Thomas Kurian said it plainly: the question is no longer whether to use AI (that ship has sailed). It's how fast you can move it across your entire organization.

Everything announced last week flows from that premise: the infrastructure is ready: the platform is mature. And it's the same stack Google runs internally for Search, YouTube, and Android. What was built for Google scale is now available to yours.

Execution is the only thing separating the companies already benefiting from AI from those still running pilots.

Build agents without a PhD

The centerpiece of the whole Next '26 was the Gemini Enterprise Agent Platform, a single end-to-end system for organizations to build, deploy, and govern AI agents at scale.

The capability is remarkable. But what's more remarkable is the completeness. From first build to full production, the entire lifecycle is covered:

• Agent Studio: build agents visually, no ML expertise needed. Export to Agent Development Kit (ADK) when you need full code control.
• Long-running agents: work autonomously in the background for days. No prompting, no babysitting.
• Agent Identity, Registry, Gateway: every agent is traceable, governed, and auditable. Full control, always.
• Agent simulation and evaluation: test before you deploy, score performance once you're live.

And it's already working across industries, at scale, in production.

Not ready to build from scratch? The Agent Gallery launches with 70+ pre-built agents from Google partners, including Zazmic — vetted, ready to deploy, and customizable for your environment.

Agentic commerce: AI that sells & serves

Customer-facing AI got its own dedicated track at Next '26, and it's further along than most people realize.

The Shopping agent and Food Ordering agents bring conversational sales and ordering across digital channels. The Omnichannel Gateway keeps context intact across web, mobile, and voice, so when a customer moves from chat to a phone call, the agent picks up exactly where it left off.

Already in production at:

• Home Depot: AI phone agent identifies customer needs in under 10 seconds
• Papa John's: unified voice-and-text ordering with memory of past preferences
• Macy's: multimodal shopping concierge with virtual try-on

If you're in retail, hospitality, or any consumer-facing business, these aren't demos. They’re running at enterprise scale today.

AI for every employee, not just developers

For employees who aren't developers, the Gemini Enterprise app is where agentic AI becomes real and usable.

Anyone can now build automated workflows without writing code. There's a central inbox to monitor what agents are doing and where they need input. Teams can create repeatable "skills" for standard processes and share them like a document.

The most practical shift: AI stops being something you go to and becomes something that's already running inside the tools your team uses every day.

Workspace Got Smarter Across the Board

Workspace Intelligence is Google’s semantic layer across Google Workspace. It doesn’t just understand content, it understands context: who you work with, what you’re working on, and how your organization operates.

Key capabilities:

• Ask Gemini in Chat: issue goals, Gemini executes across Workspace
• Drive Projects: organizes Drive around active work, not folders
• Workspace MCP Server: exposes Gmail, Drive, and Calendar to external AI apps and agents

Across core apps:

• Gmail: surfaces priority information and answers queries in search
• Sheets: connects live enterprise data to build dashboards in place
• Meet: auto-generates notes and action items across all meeting types
• Vids: branded AI avatars for scalable content creation

For Microsoft 365 organizations:

Migration is now up to 5x faster, with improved interoperability tools that mean your teams can collaborate with Office users without disruption during or after the move.

Agentic Security in the AI Era: Google + Wiz

As AI expands your attack surface, manual security can't keep up. Google's answer is making defense agentic too. Its acquisition of Wiz is complete, and the combined security platform is already more capable than either was separately, giving you visibility across your entire environment, however and wherever your teams build.

What's live or coming shortly:

• Inline AI security hooks: guardrails run before code is committed, not after
• AI-Bill of Materials: every AI framework, model, and IDE extension in your environment, mapped automatically
• Shadow AI reporting in Chrome Enterprise: see exactly what unsanctioned AI tools your team is using
• Threat Hunting, Detection Engineering, and Third-Party Context agents: autonomous security operations that work at the speed of the threats themselves
• Model Armor: built-in protection against prompt injections, tool poisoning, and data leakage across agent workflows
• Secret Manager + ADK integration: prevents password leaks and prompt injection risks at the development stage
• Google Cloud Fraud Defense: tells apart bots, humans, and agents in real time
• Post-quantum cryptography: quantum-safe key imports for organizations planning ahead
• Confidential Computing for G4 VMs: hardware-level protection for sensitive AI workloads
• IAM updates: streamlined least-privilege roles, simpler to configure and enforce

The Infrastructure to Run It All

Agents at scale need serious compute behind them. And this is where Google made a very deliberate move forward.

Purpose-built compute for the agentic era

For the first time, Google split a TPU generation into two distinct chips:

• TPU 8t: built for training.

Scales up to 9,600 chips in a single superpod, delivering nearly 3x the compute performance of the previous generation. This is what enables faster model development and iteration at enterprise scale.

• TPU 8i: optimized for inference

This is the part that actually runs your agents in production. It delivers up to 80% better performance per dollar, which becomes critical when you’re operating thousands (or millions) of agent interactions daily.

The economics of AI inference just improved meaningfully.

Data movement is no longer the bottleneck

Running agents isn’t just about compute, it’s about moving massive amounts of data, fast.

• The Virgo Network doubles connectivity for large-scale AI workloads
• Google Cloud Managed Lustre now supports up to 10 TB per second throughput

For organizations working with large datasets, real-time analytics, or multi-agent systems, latency and throughput are often the hidden constraints. Those constraints just loosened.

The data layer for agentic systems

An agent without context is just guesswork. That's the problem the Agentic Data Cloud solves.

The Knowledge Catalog builds a dynamic map of your entire business — automatically tagging and connecting data across your organization so agents understand your context, your terminology, and your data. The Cross-Cloud Lakehouse means agents can query data wherever it lives — including AWS and Azure — without moving it. No friction, no migration, no compromise.

The bottom line: better data access means smarter agents. This is the layer that makes everything else actually work.

Built for regulated & sovereign environments

With Google Distributed Cloud, Google continues to expand support for:

• Air-gapped deployments
• Sovereign cloud requirements
• Strict data residency controls

Industries like financial services, healthcare, and the public sector can now adopt advanced AI without compromising compliance or data control.

What We're Telling Our Clients

Back from Las Vegas, these are the conversations we're having with clients right now:

Your Workspace licenses

Many of the Gemini and Intelligence features above aren't on by default. They depend on your add-on tier. The first step is knowing what you actually have access to.

Your Chrome Enterprise setup

Shadow AI reporting and AI-aware threat detection are available now. If you don't have visibility into what AI tools your team is using, that's worth fixing quickly.

Get one real agent in production

Pick a workflow (meeting follow-ups, invoice review, customer routing, etc.) and build a properly governed agent around it. That's how this becomes real for your business.

Your security posture

The Wiz integration changes what's possible for cloud and AI security. If it's been a while since you looked at this, now's the right time.

The gap is growing

Thirty-two thousand leaders in one room, all asking the same question: when does AI start paying off for us?

The ones who'll have answers by Next '27 are the ones who start building now. The infrastructure Google put on the table this week is the most complete enterprise AI stack we've seen. The question is whether you move with it now, or spend next year catching up.

We're here to make sure you're on the right side of it.

Let's talk.