Google Cloud Next '26 is right around the corner (April 22–24) at Mandalay Bay, Las Vegas and while the buzz is all about Agentic AI, I think the more important conversation happening on that expo floor will be about something less glamorous: governance.
We built agents. Now what?
Over the past two years, most enterprise teams have been in a race to deploy AI agents. Automate this workflow, connect that data source, build a copilot for this team. The velocity was impressive.
But ask any engineering or ops leader honestly, and you'll hear a version of the same story: agents got built in silos. Different tools, different stacks, no shared visibility into what the collective system is doing. The term practitioners are quietly using for this is 'agent sprawl' and it's one of the trickier problems that the industry is grappling with in 2026.
Agent sprawl creates real compliance exposure. It makes ROI measurement almost impossible. And it means that the thing you built to make decisions autonomously is doing so without adequate oversight.
What makes this year's Next different
In the previous years, Google Cloud Next was largely about what AI can do. This year, based on what Google and its partner ecosystem have been signaling, the conversation is shifting to what AI can do reliably, at scale, inside regulated enterprises.
That's a meaningfully different question, and it's one that developers and architects need to be a part of, not just C-suite buyers.
A few themes worth tracking at the event:
1. Unified data as the governance foundation. BigQuery's ability to unify structured and unstructured data in a single governed layer is increasingly the difference between AI that can be audited and AI that can't. Watch for announcements around data estate governance and sovereign data compliance.
2. Native observability over bolt-on monitoring. There's growing recognition that AI governance tools stitched across multiple platforms create their own fragility. The most defensible architectures are the ones where the observability is native to the platform, not added afterward.
3. Agent lifecycle management. Deploying an agent is table stakes. The harder problem is managing it across its full lifecycle: versioning, rollback, permission changes, deprecation. Expect this to surface in sessions around Vertex AI and Gemini Enterprise.
The question I'd bring to the event
If you're attending Next '26, then here's the lens I'd apply to every demo and announcement:
"Does this make the AI more capable, or does it make it more trustworthy at scale?"
The most exciting thing happening in the tooling and architecture patterns that let organizations trust autonomous systems to run inside their business with audit trails, accountability structures, and the ability to intervene when something goes wrong.
That's the story I'm most interested in at Mandalay Bay this April.
If you're heading to Google Cloud Next '26 and want to dig into the governance and agentic AI angle, then you must read about this team that's participating in the event.
Some teams attending are specifically focused on the 'production-scale AI governance' conversation. It'll be worth connecting during the event if that's on your radar too.
Top comments (0)