Dome Systems: The operational control plane for the agentic enterprise

Every major shift in enterprise computing — virtualization, cloud, containers, microservices — has produced a moment in time where the rate of adoption outpaced the infrastructure available to govern it. In every case, enterprises that built an operational layer early gained a structural advantage, not just in security, but in the speed at which they could say ‘yes’ to innovation and survive the shift.

Enterprise AI adoption has arrived at that moment — and we're seeing it across our own portfolio. Legora’s AI platform now serves over 1,000 law firms and in-house legal teams across more than 50 markets, and Ramp shipped finance agents that enforce expense policy with 99% accuracy across 50,000 businesses. Agents are becoming load-bearing infrastructure across commerce, finance, and legal operations — and every enterprise function is building or evaluating them. Yet, most enterprises are deploying agents in a vacuum. Not because the technology isn't ready, but because the infrastructure to govern it doesn't fully exist yet.

AI is a new era and demands new infrastructure 

The enterprise toolkit we've relied on for decades, such as RBAC, API gateways, audit logs, and network policies, was largely designed to control access, but agents require control over decisions. The shift from deterministic to probabilistic execution changes the very nature of what governance means in this era. Governing who can invoke an agent is necessary but no longer sufficient when the agent itself decides what to do with its permissions. 

Fragmentation compounds the challenge. Agents are appearing simultaneously across AI platforms, cloud providers, data platforms, SaaS tools, and developer environments. Each surface is building its own partial answer to governance, while none provide consistency and coherence across the others. The result is an agentic estate where teams face five surfaces, three components per agent (code, model, and tools), and no unified system of control.

At Bessemer, we've long believed that the infrastructure layer for each computing era follows a predictable arc: capability proliferates, governance lags, and a platform layer eventually closes the gap. The companies that defined those layers — directory services for networking, API gateways for distributed applications, container orchestration and secrets management for cloud — became foundational. We see the same structural opportunity in the agentic era, with even higher stakes this time around.

Controlling the agentic estate with Dome Systems

Dome Systems is building the agentic infrastructure platform: one system of control across every agent, every runtime, every cloud. Dome is purpose-built for reasoning systems from the ground up, not adapted from tooling designed for deterministic ones.

At its core, Dome recognizes that every agent, regardless of how it's built or where it runs, is composed of three things: code (the runtime and orchestration), a model (the reasoning engine), and tools (anything the agent calls to interact with the external world). Each component introduces its own governance surface. Dome addresses all three through a workflow with three operational principles:

• Dome first connects to every system to ensure every agent is known and every tool is accessible through a governed path — assigning identity, surfacing permitted tools filtered by policy, provisioning credentials, and integrating via SDK, sidecar, or gateway regardless of runtime
• Dome then evaluates every action against policy before it executes, using deterministic authorization at the tool-call boundary, the point where an agent's reasoning crosses into the external world, securing the agent for use within the enterprise
• Dome also operates and records every governed decision with full context (who, what, under which policy, with what result) and streams events into the enterprise's existing infrastructure rather than replacing it

Built by a team that’s done this before 

Platform teams want to enable their stakeholders to innovate without sacrificing the governance their enterprise requires. Dome Systems co-founders Dave McJannet and Marc Holmes have seen and solved this problem before, and we saw their success firsthand during the cloud-era when we were privileged to back them at HashiCorp.

Dave, as CEO of HashiCorp, took the company from a collection of early-stage open-source projects to a Cloud 100 mainstay, an IPO, and ultimately a significant acquisition by IBM. He did so by understanding a truth that repeats with every platform shift: enterprises need a neutral, consistent system of control before they can adopt new infrastructure at scale. HashiCorp's Terraform, Vault, and Consul became the governance layer for the cloud era — not by replacing AWS, Azure, or GCP, but by sitting above all of them and providing consistency and control. Marc, as CMO, helped architect the developer-first go-to-market engine that made HashiCorp's adoption model a case study in how infrastructure products achieve ubiquity alongside developer love. 

HashiCorp gave platform teams a consistent system of control for cloud infrastructure. Dome Systems will do the same for the agentic estate, which is one of many reasons why we’re thrilled to be backing Dave and Marc again by co-leading Dome Systems' seed round.

Saying ‘yes’ to AI deployments with confidence and control

We're still in the early days of the AI era, but the window for establishing consistent operations is narrowing. The gap between deployment velocity and governance capability widens every week. Enterprises that implement a governance layer now will be the ones that deploy AI systems and capabilities fastest and most confidently. Those that hesitate will face the same catch-up they faced with containers and cloud, but at the current pace of AI, it’s unlikely they’ll survive.

Dome Systems is currently hiring engineers who want to build foundational infrastructure from scratch, alongside an exceptional team that's done it before. Learn more here to join them.