Pretxt Exits Stealth With $5M Seed to Supervise AI Agents at Work
Pretxt exits stealth with $5M seed to supervise AI agents mid-task, claiming tests cut policy-violating actions from one in five to zero.
By Amara Osei
4 min read

What's News
- Pretxt raised $5 million in seed funding from Deciens Capital, Howdy Partners, Service Provider Capital and NSV Wolf Capital.
- The company claims its supervision runs 60 times faster and cut policy-violating agent actions from nearly one in five to zero in Zapier's AutomationBench tests.
- Co-founders Brian Soby and Aaron Costello previously worked together at AppOmni.
- The platform deliberately avoids using an LLM to judge agent behavior.
- The company operates from Aspen, Colorado, and Dublin, Ireland, via an early-access program.
Pretxt has raised $5 million in seed funding to build an enterprise platform that supervises AI agents while they perform business tasks. Deciens Capital, Howdy Partners, Service Provider Capital and NSV Wolf Capital participated in the round. The company, based in Aspen, Colorado, and Dublin, Ireland, is offering the platform through an early-access program.
The startup brands its product category as AI Supervision. Its runtime evaluates an agent's behavior in the context of an assigned task, spots deviations from approved business processes, and steers the agent back to an acceptable course of action.
What problem does Pretxt target?
The company is chasing operational failures that occur even when an agent hasn't been compromised. CEO Brian Soby puts it bluntly: "You don't need a prompt injection attack for an AI agent to do something it shouldn't."
AI agents increasingly talk to customers, access enterprise applications, process transactions and run multistep workflows. An agent can still skip a required procedure, create an unauthorized exception, make an unapproved commitment, or act without a necessary authorization. Pretxt cites examples such as a customer-service agent issuing an unauthorized refund or a financial-services agent bypassing identity-verification requirements.
Those failures carry financial, compliance, operational and reputational consequences. Pretxt argues that existing security, governance, observability and guardrail tools don't determine whether an agent is completing its assigned work according to procedure. Monitoring tools flag problems after the fact; a guardrail that blocks a prohibited action can leave the workflow incomplete.
How fast is it, and what do the numbers say?
Pretxt claims its supervision runs 60 times faster than the alternative approach. In testing with Zapier's AutomationBench, which models real-world customer-support workflows, the company said the technology cut policy-violating agent actions from nearly one in five to zero. Pretxt reported those results itself, and they may not hold in every production environment.
The platform deliberately avoids using a large language model to judge another AI agent's behavior. The company says that approach avoids unnecessary cost, latency and inconsistent decisions, and delivers fast, near-deterministic supervision at machine speed. Actions that satisfy an organization's standards proceed; proposed actions that cross established boundaries get flagged before reaching a customer or system of record.
Enterprise teams define expectations, procedures and boundaries in plain language, without needing to understand the agent's permission model, architecture or tools. CTO Aaron Costello said: "Supervising an AI agent shouldn't mean becoming an expert in everything it touches, or in how it's built. Pretxt takes that complexity off the table."
The platform also reviews real-world agent behavior to find gaps in instructions and propose refinements. Human reviewers must approve any changes before they take effect, and the structure doesn't let an agent rewrite its own rules.
Who is behind the company?
Co-founder and CEO Brian Soby has more than 25 years of experience across enterprise technology, AI, SaaS security and platform development. He co-founded AppOmni and served as its CTO. Earlier roles include partner at FreeFly Security, director of product security at Salesforce, lead information security engineer at MITRE, and network security engineer at Raytheon.
Co-founder and CTO Aaron Costello was chief of security research at AppOmni. He identified more than 20 configuration risks and multiple zero-day vulnerabilities affecting Salesforce Industry Clouds. He also found a vulnerability in ServiceNow AI Agents described at the time as the most severe AI security flaw identified, plus more than 1,000 ServiceNow instances exposing data. The distributed team includes alumni of Salesforce, MITRE, SAP, Deloitte, VMware, HackerOne and AppOmni.
Why investors and analysts care
Pretxt is targeting consequential enterprise applications first, especially in regulated industries where an agent's mistake could cause financial losses, compliance violations or customer damage. The platform could also let companies hand agents the tools for more substantial work, including in environments lacking granular access controls.
Chris Steffen, Vice President of Research at Enterprise Management Associates, said: "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."
Nate Pacer, Partner at NSV Wolf Capital, frames the bet in ROI terms: "The ROI on AI agents is limited by how much real work a company is willing to hand them. Pretxt expands that limit." Vishal Rana, Partner at Deciens Capital, added that his firm invested because supervision at runtime "will become a foundational layer of enterprise AI infrastructure."
Andrew Sweet, Vice President of Emerging Technology and Strategic Accounts at EverSec, said his clients want agents in workflows where the business outcome matters, and Pretxt connects written operating standards to what agents actually do in production. The near-term test: converting self-reported benchmark results into trusted supervision across live enterprise environments.
Source: pulse2.com
More from Amara Osei
Show full bio
Senior reporter covering companies, deals and strategy in entrepreneurship, startups and venture funding at Business Bearings.
2,175 articles