Highlights from Google Next: How AI is becoming an active player in our workplace
Google Next 2026 in Las Vegas highlighted just how rapidly the world of AI has evolved since last year’s conference and how companies will operate in the coming years. The central theme of the entire conference was clear: We are moving from AI as a tool to AI as an active agent. After their visit to Las Vegas, our colleagues Jerry, Antonio, and Michael explored what this means for businesses and our future.
From isolated AI features to multi-agents
“From AI that answers to AI that acts.” In the future, the focus will shift from individual AI features and isolated chatbots to autonomous AI agents that can plan, act, communicate with one another, coordinate, and interact with tools. These agents are the new microservices—with identities, clearly defined permissions, and memory storage—that communicate via standardized protocols (A2A, Agent Registry).
The true strength of the future, therefore, does not come from a “super-agent,” but from multi-agent systems: specialized agents collaborate through a division of labor and only form a robust system when working together.
Multi-orchestration is becoming an integral part
This points to another significant shift: models are becoming interchangeable; it’s the platform surrounding them that makes the difference.
In the future, companies will make decisions based on platforms, not models. Depending on the use case, governance requirements, costs, or performance, the appropriate model is deployed—orchestrated by a common platform.
Competition is thus shifting noticeably: less focus on individual models—more focus on orchestration, governance, and execution control. AI is becoming a full-platform layer embedded in every layer of the stack: from cloud infrastructure, through productivity units and tools, to analytics and customer service, as well as cybersecurity and operations. This clearly shows a shift away from individual AI projects toward the fundamental integration of AI into daily systems and processes—AI is becoming an integral part of IT. Google is positioning itself as the “operating system” for the agent-driven corporate world—an exciting strategic positioning.
Data remains the foundation of every AI architecture
No data, no agents—and no good data, no good decisions.
But here, too, a profound integration is emerging: data architectures are being designed with AI in mind and are intended to bring together data, context, and action within a single system.
The paradigm shift:
- Not just storing or analyzing data
- but making it directly usable for autonomous actions
For companies, this translates into very specific new architectural questions:
- How do we manage read and write permissions across cloud storage, S3, and ADLS?
- How do we prevent uncontrolled growth in hybrid and multi-cloud data paths?
- How do we ensure that agents see only what they are supposed to see—and nothing more?
Agents as identities with enhanced security measures
One area that many companies tend to underestimate is Identity & Access Management for agents. Here, we see strong parallels to the UEM world: What used to apply only to endpoints and users must now be extended to include agents—because they must be controlled, monitored, and restricted just as clearly. More info on IAM here.
Security considerations must be addressed to:
- How do we model an agent in IAM? A service account per agent or an agent pool?
- How do we handle credential rotation for long-running agents?
- Data paths: Cross-cloud lakehouse requires clear access rules for S3, ADLS, and cloud storage.
- How do we set hard caps and soft caps to prevent cost explosions with agents scaling in parallel?
Secure-by-design is becoming essential
As companies expand into complex hybrid and multi-cloud environments and feed AI with sensitive corporate data, attackers are also taking advantage of these opportunities and have managed to reduce the time between initial system access and secondary exploitation from 8 hours to just 22 seconds. Security must also form the foundation for AI initiatives and be considered from the ground up for every product, built in directly, and implemented not just AI-assisted but AI-led. At Next 2026, too, the topic carried far greater weight than in previous years. Google even speaks of a transition to AI-led cyber defense, monitored by humans.
Whereas generative AI tools used to summarize warning messages, agents will actively collaborate in the Security Operations Center in the future. For example, a Detection Engineering Agent that identifies vulnerabilities and writes new rules on its own when necessary. Or Third Party Context Agents that enrich security workflows with external threat intelligence.
Practical experience shows that this approach is already having a measurable impact today: Using the AI model Claude Mythos Preview, the Firefox team was able to identify and close 271 security vulnerabilities within a single update cycle—a scale that would have been virtually impossible to achieve with traditional manual methods. What matters here is not so much that AI finds “superhuman” vulnerabilities, but that it thinks at the level of experienced security researchers—only at a speed and scale that significantly reduces the advantage attackers have.
For admins and security teams, this means: less manual maintenance, faster response times, and better coverage of large code and system landscapes. At the same time, this very example shows that the race between AI used for attacks and AI used for defense does not lead to less responsibility. On the contrary: when vulnerabilities can be detected much faster by machines, the pressure increases to consistently further develop security processes, governance, and monitoring—because the same technology is, in principle, also available to attackers.
AI-Assisted Development: Productivity That Scales
Another clear signal from Next 2026: Software development will be permanently supported by AI. AI development tools take on time-consuming tasks and effectively function like a senior engineer who is always available:
- Writing code
- Finding bugs
- Optimizing infrastructure
- Analyzing performance
At the same time, low-code and no-code tools are opening the door for business departments. Teams across the company can develop applications for their specific use cases. A key skill for the future is the ability to formulate clear goals and provide the right context.
Google’s figures underscore this:
- About 75% of Google Cloud customers use AI products to strengthen their business
- In the last 12 months, 330 Google Cloud customers have processed more than one trillion tokens—35 customers even processed 10 trillion tokens
This isn’t a pipe dream—it’s reality. This is the new competitive landscape.
Google’s path in this new reality
Google appeared extremely focused at Next 2026, is moving faster than in previous years, and is determined to stay ahead in the field of enterprise AI. The combination of infrastructure, the Gemini ecosystem, agent tools, partnerships, and investments in security makes the company a very strong player.
CEO Thomas Kurian has clearly structured Google’s strategy:
1. Agentic Foundation
Gemini models, TPU infrastructure
2. Agentic Platform
Gemini Enterprise, Agent Designer, Inbox, ADK
3. Agentic Applications
Pre-built agents for CRM, IT operations, marketing
Particularly notable was the strong presence of Wiz, an AI-first security platform. This demonstrates how strategically important cloud security is to Google. Google’s acquisition of Wiz signals its intention to strengthen the Google Cloud security portfolio through modern, cloud-native security solutions, security posture management, threat detection, and multi-cloud visibility.
Our Conclusion
- The business impact of AI is real and growing stronger—now ranging from microservices to multi-service agents with automated end-to-end processes.
- AI agents have evolved from a concept to a corporate reality: They are viewed as new “users” and must be considered when orchestrating data and infrastructure.
- Security must be part of the architecture from the very beginning—and, as a foundation, plays an enormously important role in the AI context.
- The role of IT security and platform teams is evolving:
– away from manual triage
– toward management, governance, and control of AI agents
The major challenge for companies will not be introducing AI, but maintaining control—both technically and organizationally. The Agentic Era has begun—and is no longer a topic of the future.
Or, in Google’s words: AI agents are moving from concept to enterprise reality—and security is moving with them.
Contact us
As an IT service provider, we see our role precisely there: at the intersection of platform, IT & endpoint security, and day-to-day operations. We focus on both developing our own AI agents and managing their identities within a UEM framework to keep your infrastructure secure.
