AI Business · · 4 min read
Can Pretxt Really Stop Your AI Agents from Going Rogue?
We break down how a new breed of AI supervision aims to keep your automated workforce in check and what it means for your business.

Imagine an AI agent handling your customer service, processing transactions, or even managing critical supply chain workflows. It sounds efficient, doesn't it?
But what happens when that agent, in its digital zeal, skips a crucial step, hallucinates an exception, or makes an unauthorised commitment on your company's behalf?
It's a growing concern as artificial intelligence takes on more consequential tasks. That's why a startup named Pretxt, co-founded in Ireland, recently secured $5 million in seed funding, announced on November 17, 2026, to tackle this very problem.
This isn't just about preventing malicious 'prompt injection attacks' where someone tries to trick the AI. As Pretxt CEO Brian Soby puts it, an AI agent doesn't need to be attacked to 'do something it shouldn't.' It can simply deviate.
The real story here isn't just Pretxt's funding round. It's about a fundamental shift in how businesses must think about managing autonomous AI. How do you trust a digital workforce that can operate independently, yet still needs to adhere to your strict rules?
That raises three questions.
Question 1
What Problem Does Pretxt Aim to Solve for Businesses?
Across sectors, AI agents are increasingly handling tasks that were once solely human domains. They're talking to customers, processing transactions, and operating complex workflows. This promises huge efficiency gains, but it also introduces new risks.
These agents can sometimes skip steps in a process, 'hallucinate' (make up) exceptions that don't exist, or even make commitments that go beyond their authority. These aren't minor glitches; they can have real business consequences, from customer dissatisfaction to financial liabilities.
Pretxt, co-founded by security professionals Brian Soby (CEO) and Aaron Costello (CTO), is emerging from stealth mode to address this. The company, co-located in Dublin and headquartered in Colorado, raised $5 million in seed funding from Deciens Capital, Howdy Partners, and Service Provider Capital, specifically to develop a solution for these 'off-script' AI behaviours.
For context, Brian Soby previously co-founded AppOmni, a network security firm, and Aaron Costello served as its chief of SaaS security research. Their background highlights the security-first approach to managing AI agents. As Soby notes, 'You can't supervise a workforce of AI agents with hope, a rulebook and reviews after the fact.'
Question 2
How Does Pretxt's AI Supervision Layer Actually Work?
Pretxt's solution is a 'supervision layer' that sits atop your existing AI agents. Think of it like a quality control manager at a large manufacturing plant. This manager doesn't build the products, but continuously observes the production line, ensuring each step is followed correctly and intervening if something goes awry.
The platform aims to 'contextually evaluate AI agent behavior' in real time. This means it understands what the agent is *supposed* to be doing in a given situation, not just checking against a static list of rules. When an agent starts to deviate from approved processes, Pretxt's layer steps in to guide it back on track.
Aaron Costello, Pretxt's CTO, explains that this approach means you don't need to be an expert in every detail of the AI agent's construction or every task it performs. The system handles the heavy lifting of observation and intervention. Pretxt claims its platform can provide this supervision '60-times faster' than manual methods, and importantly, it does so without impeding the actual work the AI agents are meant to do.
And that brings us to the second question.
Question 3
What Does Real-Time AI Agent Supervision Mean for Your Business?
This shift from reactive oversight to proactive, real-time supervision is significant. Businesses often deploy AI agents with the hope they'll follow instructions, then rely on human reviews after the fact to catch mistakes. But as Soby observed, 'AI needs supervision while it acts.'
Vishal Rana, a partner at Deciens Capital, an investor in Pretxt, underlines this point: 'AI agents are moving into workflows where mistakes have real consequences.' He believes that 'supervision at runtime will become a foundational layer of enterprise AI infrastructure.'
For your business, this means potentially unlocking new levels of trust and scalability for AI agents. You could deploy agents into more critical roles, say, managing customer financial data or sensitive logistics, with greater confidence that they will adhere to compliance standards and operational protocols.
It's about building guardrails that allow innovation. Instead of fearing what an autonomous AI might do, businesses can use systems like Pretxt to ensure that their digital workforce operates within defined boundaries, reducing operational risk and enhancing overall efficiency.
The Bigger Picture
The march towards more autonomous AI agents in the workplace is inevitable. They promise to transform how we operate, freeing up human talent for more strategic work and accelerating processes.
But the core challenge isn't just building smarter AI; it's building AI that can be trusted. The emergence of companies like Pretxt highlights that trust isn't a passive outcome of good programming, but an active, continuous process of supervision and guidance.
It's about ensuring that as AI agents become more independent, they remain accountable to the businesses they serve. The future of AI in business isn't just about intelligence; it's about intelligent control.
Your AI agents are watching, and now, someone is watching them too.
Quick Answers
What problem does Pretxt solve for businesses?
Pretxt addresses the issue of AI agents deviating from approved processes, skipping steps, or making unauthorized commitments when handling critical business tasks. Their platform ensures AI agents stay within defined operational boundaries.
How does Pretxt's AI supervision layer work?
Pretxt's supervision layer contextually evaluates AI agent behavior in real time. If an agent deviates from approved processes, the platform guides it back on track, providing active oversight without impeding the agent's work.
What does real-time AI supervision mean for businesses?
Real-time AI supervision allows businesses to deploy AI agents in more critical roles with greater confidence. It reduces risks associated with errors or unauthorized actions, making AI agents more reliable and scalable for complex workflows.
Sources
- Business Post17 Nov · 17:38 IST
- Silicon Republic17 Nov · 13:32 IST
Facts are credited to the outlets above. The analysis, calculations and Signal Check are by the Finnect AI Insights desk.