Automation Anywhere announced EnterpriseClaw at its Imagine 2026 conference on May 19, in collaboration with Cisco, NVIDIA, Okta, and OpenAI. The launch addresses one of the most persistent blockers to enterprise AI adoption: how to actually run AI agents inside complex, multi-system enterprise environments without creating security and governance nightmares.

The name references the OpenClaw movement — open-source agent protocols that have made it easier to build AI agents but harder to control them at scale. EnterpriseClaw is the enterprise response to that: a framework that brings agent capability into production environments while adding the identity, security, and monitoring controls that IT and security teams require before signing off on deployment.

What EnterpriseClaw Actually Does

At its core, EnterpriseClaw is a runtime environment for autonomous AI agents. It lets organisations deploy agents across cloud platforms, physical desktops, on-premises systems, and secured enterprise networks — all from a single orchestration layer.

Each of the four partner organisations contributes a distinct piece of the stack:

Cisco provides AI Defense, a security layer built specifically for agentic systems. Unlike traditional perimeter security, AI Defense monitors agent behaviour, detects anomalous actions, and enforces policy at the agent level rather than just the network level.

NVIDIA contributes two components. The first is OpenShell, an open-source runtime designed to build and deploy autonomous, self-evolving agents more safely. The second is NVIDIA NIM microservices, which run Nemotron open models for enterprises that want to keep AI processing fully on-premises rather than sending data to external cloud services.

Okta handles identity. Cross-agent identity management means each agent gets authenticated credentials, and access controls are enforced consistently as agents interact with enterprise systems. This matters because multi-agent architectures can quickly become an identity sprawl problem — agents calling other agents, each with different permissions, with no clear audit trail.

OpenAI enables EnterpriseClaw agents to run on GPT-5.5, OpenAI’s current flagship model, for organisations that want frontier model capability within the governed framework.

The Governance Layer That Was Missing

Alongside EnterpriseClaw, Automation Anywhere announced AI Evaluations — a capability for testing agent performance at both design time and runtime. Organisations can now assess whether agents achieve the correct outcome, use the right tools, and follow appropriate execution paths before and during deployment.

This is not a minor addition. One of the biggest reasons enterprises have stayed in AI pilot mode is the inability to answer basic governance questions: Did the agent do what we expected? How do we know? What happens when it does not? AI Evaluations gives teams a structured answer to those questions rather than relying on manual monitoring.

EnterpriseClaw is available in preview now, with general availability expected later in 2026.

Context: AI Booking Numbers Tell the Real Story

Automation Anywhere reported that 61% of its Q4 software bookings came from AI. That is not a product line at the edge of the business — it is the majority of what enterprise customers are buying from them. The Imagine 2026 announcements are a signal that the company is betting its next chapter on agentic automation running inside enterprise systems rather than alongside them.

The partnership composition is also deliberate. Cisco brings network security credibility. NVIDIA brings the infrastructure layer, including the on-premises story that regulated industries require. Okta brings identity governance that enterprise IT teams recognise. OpenAI brings model capability. Together they form a coalition that speaks to the four most common objections enterprise security and IT teams raise when AI deployment proposals land on their desks.

What This Means for Business

If you are an enterprise IT or operations leader who has been watching AI agents from a safe distance, this announcement reframes the conversation. The argument used to be: we cannot deploy AI agents because we cannot govern them. EnterpriseClaw and the associated governance tooling are a direct response to that argument.

A few things worth thinking through as you evaluate this:

On-premises is now a first-class option. The NVIDIA NIM integration means organisations in healthcare, finance, or government that cannot send data to external cloud services have a viable path to enterprise-grade AI agents. That removes one of the most common blockers for regulated industries.

Identity is the underestimated problem. Most AI deployment discussions focus on model capability. The Okta integration signals that Automation Anywhere understands where most enterprise deployments actually break down: not at the model level but at the permissions and accountability level. Who did the agent act as? What was it allowed to access? Who approved that?

Preview availability creates a planning window. GA is not until later in 2026, which means now is the right time to assess whether your existing automation infrastructure is ready to add agent orchestration on top of it, rather than scrambling once it ships broadly.

The organisations that will get the most out of agentic automation are not the ones that wait for the technology to mature further. They are the ones that use the preview period to answer the harder questions: which processes are genuinely better with autonomous agents, where is human oversight still required, and what governance infrastructure needs to be in place before the first agent goes live in production.

---

Enterprise DNA helps business leaders assess, deploy, and govern AI in real enterprise environments. If you are evaluating agentic automation for your organisation, start with a discovery call.