Pretxt, a newly launched enterprise AI company focused on runtime oversight of AI agents, has stepped out of stealth mode with $5m in seed capital to build out a technology category it calls AI Supervision.
Deciens Capital, Howdy Partners, Service Provider Capital and NSV Wolf Capital backed the round. Pretxt plans to use the capital to bring its supervision layer to enterprises that want AI agents to handle higher-stakes work without drifting from approved procedures.
AI agents no longer simply produce text. They now speak with customers, connect to other systems, handle transactions and run business processes. According to Pretxt, these agents do not always behave as instructed.
They may skip mandatory steps, create their own exceptions, make commitments nobody authorised or ignore sign-off requirements. A support agent offering a refund without approval, or a financial services agent that sidesteps identity checks, can expose a business to operational, financial and reputational damage, and none of this needs a hacker to trigger it.
Pretxt argues that security, governance, guardrails and observability tools each tackle part of the risk, yet none confirms that an agent is completing its task in line with business or regulatory standards.
Some guardrails halt prohibited actions and leave processes stuck awaiting human help, while monitoring tools mainly explain failures after they occur. The company cites figures showing 83% of US consumers believe brands should answer for AI interactions that go wrong.
Pretxt defines AI Supervision as continuous oversight of agent behaviour during execution, judged against an organisation’s own processes, intent and procedures. Rather than concentrating on threats or post-event review, the platform detects when an agent strays from its assigned task and redirects it before the result reaches a customer or system of record.
The company does not rely on large language models to judge agent behaviour, a design choice it says removes latency, cost and inconsistent decisions, and makes its supervision 60 times faster.
Compliant interactions pass through with millisecond-level checks, while actions that breach set boundaries are caught and resolved first. In testing on Zapier’s AutomationBench, which models real-world support workflows, Pretxt cut policy-violating agent actions from almost one in five to zero.
The platform also learns from live agent behaviour, spotting gaps in guidance and proposing updates to the supervision plan. Enterprise teams set expectations in plain language, test proposals against their own examples and approve changes before they go live.
Pretxt says this lets organisations give agents wider tool access even where granular access controls are absent. The company targets consequential workflows, especially in regulated industries where a single agent error can carry financial, operational, reputational or compliance fallout.
Pretxt co-founder and CEO Brian Soby said, “You don’t need a prompt injection attack for an AI agent to do something it shouldn’t. 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.”
Pretxt co-founder and CTO Aaron Costello added, “Supervising an AI agent shouldn’t mean becoming an expert in everything it touches, or in how it’s built. 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.”
Copyright © 2026 FinTech Global









