Xurrent says 91% of its customers run AI in production, outpacing industry average.

Xurrent Ran AI in Production at 91% Before Anyone Else. Now It Wants to Sell You the Governance Layer.

Xurrent, a Santa Barbara-based IT service and operations management platform, announced this week that it is extending its AI fabric with autonomous agents and an open Model Context Protocol server. The announcement is being covered as a governance story. It is not. It is a market timing story.

Here is the number the coverage is missing: 91 percent. That is the share of Xurrent's customers running its Sera AI system in production today, handling ticket triage, knowledge work, and routine resolution without human initiation. The industry average for enterprises deriving value from AI deployments is somewhere between 19 and 29 percent, depending on which analyst number you prefer. Xurrent did not just beat that number. It beat it by a factor of four to five, and it has been doing so for years, according to PR Newswire.

The press release announcing the new agents quoted CEO Brian Wenngatz saying his company's agents "move from assisting the IT team to joining it." That is the line Xurrent's PR team wanted in the story. It is also the wrong lede, because it treats this as a product announcement when it is actually evidence of a market inflection that nobody has named yet.

The unnamed thing is this: the companies selling AI autonomy have been solving the wrong problem. They have been asking whether AI is good enough to close tickets without a human in the loop. The enterprises actually running AI in production at scale have been asking a different question: what governance structure makes AI auditable, accountable, and legally defensible when it acts? Xurrent's 91 percent number suggests it found an answer that works. Everyone else is still arguing about the question.

The new MCP server is Xurrent's attempt to turn that answer into a product. The protocol lets customers plug any external AI model — OpenAI, Google, Anthropic, or something built in-house — into Xurrent's workflow layer. The pitch is that external models inherit the same Shared Policy and Data Layer that governs Xurrent's native agents: the same audit trail, the same policy enforcement, the same visibility. "No matter where the intelligence comes from, Xurrent controls how it behaves," in the words of CPO Phil Christianson, per APM Digest.

That is the "governance as infrastructure" framing that several outlets have already written about. It is correct. It is also incomplete, because it treats the governance layer as a constraint on AI deployment rather than the product opportunity itself.

Here is the reframe: Xurrent is not selling governance because AI is dangerous. Xurrent is selling governance because governance is what made 91 percent production deployment possible in the first place. The causal arrow runs the other direction. The companies that built governance infrastructure first are the ones who solved the deployment problem. Now they are packaging that solution for anyone running heterogeneous AI models who has not solved it yet, per PR Newswire.

This is the ITSM industry quietly inserting itself into the AI stack. These are companies that spent a decade building ticketing workflows, policy layers, audit trails, and compliance frameworks for human operators. They are now arguing that the exact same infrastructure — built for a different era — is exactly what AI agents need to operate at enterprise scale. ServiceNow, Ivanti, and the other major ITSM platforms are making the same argument in their own language. The governance layer is becoming a product category, as Channel Insider reported.

The honest counter to this story is that Xurrent is a mid-size ITSM vendor making a familiar claim: that its platform is different because it was "built for the AI era from the start." That is what every vendor says. The 91 percent production figure is self-reported and unverified by a third party. The MCP server is new and has no independent benchmarks for whether policy inheritance actually works at the API level or is a marketing description of a checkbox. Until someone tests whether an Anthropic agent plugged through Xurrent's MCP endpoint behaves identically to a Sera agent under the same policy rules, the governance claim is an assertion, not a demonstration.

The 91 percent number is still the right place to start the story. It is an order of magnitude better than the industry average, and it has been hiding in plain sight as a background stat while the AI press covers foundation model releases and funding rounds. The story is not really about Xurrent. It is about what the companies that solved governance at scale know that the rest of the industry is still learning: the autonomous future requires a layer of accountability that nobody wanted to build, and it is now the most valuable thing in the stack.

Xurrent wants to sell you that layer. Whether it works the way the press release says it works is the next story.