When you deploy an AI agent inside your business, you introduce a problem that most enterprises haven’t thought through yet: how do you know what it’s doing?

That question is what groundcover is betting $100 million on. The Israeli observability startup closed a Series C round led by One Peak on July 29, bringing its total funding to $160 million and valuing the company at $500 million. Existing investors Zeev Ventures, Angular Ventures, and Heavybit participated, alongside new investor Morgan Stanley Expansion Capital.

The round came after a year in which groundcover tripled annual recurring revenue, doubled its workforce, and closed multiple seven-figure contracts with enterprise customers, including Fortune 5 companies. It now has more than 250 paying customers.

The Problem Traditional Monitoring Tools Can’t Solve

Most enterprise observability platforms, the tools that track what’s happening inside software systems, were built for a world where engineers wrote deterministic code. A request comes in, it follows a defined path, a result comes out. The system either worked or it didn’t, and the logs told you which.

AI agents don’t work that way.

An AI agent might take a different sequence of steps each time it handles the same task. It makes judgment calls. It calls external tools in unpredictable orders. It generates outputs that vary based on context. The data volumes generated by a fleet of AI agents handling real business workflows can be orders of magnitude higher than what traditional monitoring was built for.

groundcover argues that platforms like Datadog, built for the previous era, weren’t designed for this environment. The company’s approach keeps all agent telemetry inside the customer’s own cloud, rather than routing it through a third-party SaaS platform. This matters for security-sensitive enterprises that need to know their agents’ activity logs never leave their own infrastructure.

Rotem Gur, groundcover’s CEO, has been direct about the positioning: “We want to replace Datadog.”

Why This Matters Right Now

The timing of this raise reflects where enterprise AI actually is, not where it’s being discussed.

Most of the coverage of AI agents focuses on what they can do: handle customer queries, process documents, generate code, run workflows. That’s the capability story. The operational story, the one that determines whether businesses can actually rely on AI agents at scale, is about what happens after deployment.

How do you know an agent is doing what you intended? How do you spot when it’s drifting, failing silently, making errors that compound across thousands of interactions? How do you debug an AI system when the logs are massive, unstructured, and non-deterministic by design?

These are infrastructure questions, and they’re not being answered well yet. The $100 million flowing into groundcover is a signal from institutional investors that this layer is real and imminent, not theoretical.

According to the company, traditional monitoring approaches break down specifically because of the data volumes generated by AI systems. Cloud-native and AI-driven environments produce far more telemetry than legacy platforms were designed to ingest and analyze in real time.

What This Means for Business

If you’re deploying AI agents in your business, or planning to, observability is not optional. It’s the difference between AI you can trust and AI you’re just hoping works.

Right now, most enterprises deploying AI agents are in a period where outputs look acceptable in demos and early production. The failure modes emerge at scale, when agents are handling thousands of interactions per day across diverse scenarios, when the edge cases accumulate, when small errors in reasoning compound across a workflow.

The organizations that will get real sustained value from AI agents aren’t just the ones that deploy them, they’re the ones that build the feedback loops to understand performance, catch problems early, and continuously improve. That requires visibility.

A few things every business starting to run AI agents in production should be thinking about:

What can you see? Can you trace a specific agent interaction end to end? Can you identify which step in a workflow failed, and why? If the answer is “we check outputs manually sometimes,” you don’t have observability, you have sampling.

Where is your data going? Agent telemetry can contain sensitive information about your operations, your customers, and your internal processes. If it’s flowing through a third-party platform, understand exactly where it’s going and what the retention policies are.

Can you measure improvement? If you can’t compare agent performance across versions and timeframes with real data, your AI deployment is running on intuition. That works when the stakes are low and volumes are small. It breaks down at scale.

The operational infrastructure for AI agents, the layer that tracks, monitors, diagnoses, and improves them in production, is becoming a distinct discipline. The funding going into companies like groundcover signals that enterprises are starting to treat it that way.

The Broader Signal

groundcover is not alone. The observability and AI operations category has attracted significant venture attention in 2026 as the “deploy and pray” phase of enterprise AI gives way to something more rigorous.

What’s notable about the groundcover raise specifically is the customer profile. The company says it is serving Fortune 5 enterprises with production AI agent deployments, which means the largest companies in the world are paying real money for this capability. That’s not a proof-of-concept market.

The shift from experimental AI to operational AI is the most important transition happening in enterprise technology right now. The companies building the infrastructure layer of that transition, monitoring, governance, security, cost management, are the ones capturing the most durable value.

For businesses that have deployed or are deploying AI agents: investing in the operational layer alongside the agents themselves is not overhead. It’s how you make the agents work in the long run.

If you’re planning an AI agent deployment and want to build it on a foundation that includes operational maturity from day one, Omni Ops helps businesses design and deploy AI agent workforces with real production thinking built in. Or if you want a strategic view of what your AI infrastructure should look like, Omni Advisory can help you map the full picture before you build it.