Pretxt Emerges from Stealth to Pioneer AI Supervision, a New Category for Enterprise AI
Security veterans introduce purpose-built runtime technology to supervise AI agents as they work, guiding them back on
Press Release Disclaimer: This is a press release distributed through the XPR Media network. It has not been independently verified by our newsroom.
![]()
Pretxt emerged from stealth today with $5 million in seed funding from Deciens Capital, Howdy Partners, Service Provider Capital and NSV Wolf Capital to introduce AI Supervision, a new category of enterprise technology designed to keep AI agents accountable to business expectations while they work. Pretxt’s purpose-built runtime supervision layer evaluates agent behavior in the context of the task, guides agents back when they deviate from approved processes and identifies opportunities to strengthen supervision over time, subject to human review. By addressing failures that security and monitoring alone cannot prevent, Pretxt aims to help enterprises entrust AI agents with increasingly consequential work.
“You don’t need a prompt injection attack for an AI agent to do something it shouldn’t,” said Brian Soby, co-founder and CEO of Pretxt. “Security remains essential, but the next question is whether enterprises can trust agents to carry out real work according to their standards. You can’t supervise a workforce of AI agents with hope, a rulebook and reviews after the fact. AI needs supervision while it acts. That is the category Pretxt is here to establish.”
The problem: Agents can act, but they don’t reliably follow the rules
AI agents have moved beyond generating content to performing consequential work: Talking with customers, accessing other tools and systems, processing transactions and operating business processes. But agents don’t always do what they’re told. They can skip required steps, invent exceptions, make unauthorized commitments or overlook approval requirements. A customer-service agent promising an unapproved refund or a financial-services agent bypassing verification can create immediate operational, financial and reputational exposure. And these failures don’t require a malicious attacker.
“Supervising an AI agent shouldn’t mean becoming an expert in everything it touches, or in how it’s built,” said Aaron Costello, co-founder and CTO of Pretxt. “Every platform has its own permission model. Every agent has its own tools and its own way of working. Nobody should have to master all of that before they can trust an agent with real work. Pretxt takes that complexity off the table. The agent already knows its job. Teams tell us, in plain language, how it should behave and where the boundaries are, and we supervise and steer the agent to stay within them. No untangling permissions, no digging into the agent’s internals. That’s a big win for trust, and for outcomes.”
Security, governance, guardrails and observability address important parts of enterprise AI risk, but they don’t establish that an agent is carrying out its assigned work according to business or regulatory standards. Policies and prompts express expectations, but production behavior can still diverge. Some guardrails block prohibited actions, leaving the in-flight process stalled and in need of human intervention, and monitoring tools help teams see what happened. Pretxt addresses the distinct operational challenge of understanding what an agent is supposed to accomplish and guiding it when execution drifts without impeding the work. The consequences are tangible: 83% of U.S. consumers say brands should be held accountable when an AI-powered interaction goes wrong. As agents take on more critical tasks, enterprises need supervision while the outcome can still change.
Chris Steffen, VP of Research at Enterprise Management Associates, said: “Enterprises have invested heavily in tools that ask whether an agent is secure, compliant or observable. Far fewer can answer a more basic but critical question: Is the agent correctly doing the job it was given? That’s important because organizations that deploy AI own their outcomes, no matter who’s asking: regulators, insurers or customers. Most tools can only report what went wrong after the fact, when it’s too late. Pretxt intervenes while the outcome can still change to get agents back on track, creating a new category for AI supervision the market has been missing at the moment enterprises need it most.”
Introducing AI Supervision: The Missing Operational Layer for Enterprise AI
With today’s launch, Pretxt introduces AI Supervision as a distinct enterprise discipline: Continuous oversight of what AI agents do during execution, measured against the organization’s own processes, intent and procedures. Unlike controls focused primarily on threats, prohibited actions or post-event analysis, supervision is designed to recognize when an agent drifts from its assigned job and steer it back before the outcome reaches a customer or system of record. The premise is familiar: Organizations don’t entrust people with important work using a rulebook alone. They supervise the work.
“Our clients want to deploy agents in workflows where the business outcome matters,” said Andrew Sweet, VP of Emerging Technology & Strategic Accounts at Pretxt channel partner, EverSec. “Pretxt gives us a way to connect written operating standards to what agents actually do in production.”
Pretxt delivers fast, consistent, near-deterministic supervision. By design, Pretxt does not use LLMs as a judge, eliminating unnecessary latency, cost, and unreliable and inconsistent judgement calls, and providing AI supervision 60 times faster. Interactions that satisfy business standards proceed with millisecond-scale supervision, while proposed actions that cross the organization’s boundaries are identified and resolved by Pretxt before they reach a customer or system. In tests measured on Zapier’s AutomationBench, which represents real-world support workflows, Pretxt reduced agent actions that violated company policy from nearly one in five, down to zero.
“The ROI on AI agents is limited by how much real work a company is willing to hand them,” said Nate Pacer, Partner at NSV Wolf Capital. “In our view, the entire AI revolution hinges on precisely this point. Pretxt expands that limit. When every action is held to an organization’s own standards as it happens, teams can move agents from simple tasks to the work that drives revenue and carries real risk. That’s a large and growing market, and Pretxt is defining it.”
Supervision also improves through experience. Pretxt analyzes real-world agent behavior to identify gaps in guidance and operating boundaries, then proposes refinements to the supervision plan. Enterprise teams define expectations in plain language, test proposed plans against their own examples and approve changes before they take effect. The result is not an agent rewriting its own rules, but an operating discipline that helps organizations improve how agents perform over time. Supervision also enables an organization to confidently grant their agents access to the tools needed to take meaningful action, even in the absence of granular access controls.
“AI agents are moving into workflows where mistakes have real consequences,” said Vishal Rana, Partner at Deciens Capital. “We invested in Pretxt because we believe supervision at runtime will become a foundational layer of enterprise AI infrastructure. The opportunity is not just to identify a bad outcome after it happens, but to help organizations change the outcome while the work is still underway.”
Built for the work that matters most
Pretxt is designed for consequential enterprise workflows, particularly in regulated industries where an agent’s misstep can carry financial, operational, reputational or compliance consequences.
Founders and team
Pretxt was founded by enterprise security and technology veterans whose experience spans building market-leading companies and world-class SaaS platform and AI solutions, vulnerability research and the investigation of unexpected behavior in complex systems and environments. With leadership co-located in Aspen, Colorado and Dublin, Ireland, the distributed team brings a strong professional background from Salesforce, MITRE, SAP, Deloitte, VMware, HackerOne and AppOmni:
- Brian Soby, co-founder and CEO, a technology and security professional with more than 25 years of experience. With an enterprise AI, SaaS security and platform-building pedigree, he was formerly co-founder and CTO of AppOmni. Soby has worked in the defense and commercial industries in both large and small companies, serving in roles including Partner at FreeFly Security, director of Product Security at Salesforce, lead information security engineer at MITRE, and network security engineer at Raytheon.
- Aaron Costello, co-founder and CTO, formerly chief of security research at AppOmni, where he discovered more than 20 configuration risks and multiple zero-day vulnerabilities in Salesforce Industry Clouds, a flaw in ServiceNow AI Agents dubbed the most severe AI security vulnerability uncovered at the time, and 1,000+ ServiceNow instances exposing data. Through meticulous analysis and a commitment to transparency, he’s been pivotal in shaping best practices for key technology platforms.
Pretxt availability
Pretxt is available now through an early access program.
Read the complete announcement at https://pretxt.ai/news/pretxt-emerges-from-stealth/.
About Pretxt
Pretxt is The AI Supervision Company, pioneering a new category of enterprise technology designed to help AI agents do the jobs they are assigned, the way businesses intend. Its purpose-built runtime supervision layer evaluates agent behavior against business standards while work is happening, guides agents back when they drift and recommends improvements to supervision plans for human approval. Designed for fast, near-deterministic operation at machine speed and AI scale, Pretxt helps organizations entrust AI agents with increasingly consequential work. AI does not just need security. It needs supervision. Visit http://pretxt.ai/
View source version on businesswire.com: https://www.businesswire.com/news/home/20261117322271/en/
Media gallery


