You can vibe code features. You cannot vibe trust.

Forrest Brazeal runs Freeman & Forrest, a services company that works with enterprises and helps customers sell into them. He's watched the AI revolution lower the barrier to building software products. On paper, it's never been easier to start a company and move upmarket. In practice, the bottlenecks haven't changed.

Writing code is not the bottleneck. Policy is. Process is. Certifications are. All the things you need to be a trusted player in the enterprise can't be generated by an AI.

The pressure to move faster

Forrest sees the impossible position many developers are in. AI can build features quickly, so there's pressure from the rest of the business to make everything happen instantly. The CEO vibe-coded a landing page over the weekend, so why can't you ship this feature by Tuesday?

You're being given very little time to comprehend the ins and outs of making an app enterprise-ready. Authentication, authorization, multi-tenancy, security certifications—these require human judgment and deep expertise. But nobody's budgeting time for that expertise when AI makes the feature work look effortless.

Two choices

You can try to DIY everything and hope you don't miss anything. Or you can bring in partners whose job it is to have already solved these problems.

Forrest sees this pattern with companies that use WorkOS. They're scaleups moving quickly upmarket. Using a trusted product that handles unknown unknowns lets them move faster with more confidence. You can't vibe trust, you can't launder trust, but you can bring in partners to make the trust process quicker.

The IDP problem

Here's a specific example: you build SSO integration for one identity provider. You go to market. Then another enterprise customer shows up using a different IDP. Back to the drawing board. Every new customer potentially means new engineering work on slight variations of something you've already built.

This is the wheel-reinvention trap, but Boris prefers a different framing: don't unsolve solved problems. Auth isn't trivial. The odds that something you build quickly will be inadequate are high. And it has to be 100% right.

Pre-AI is an ethos

Forrest describes WorkOS as a "pre-AI company," which he means as a compliment. It implies both timing (this is infrastructure you need before AI features) and approach (rigorous human expertise applied to non-trivial problems).

Everyone's building agents and MCP servers. Are they adding proper guardrails? Are they exposing the right tools to the right people? Are they thinking about authentication and authorization in the AI layer? These questions require the same rigor that pre-AI infrastructure required.

The operating system metaphor

Forrest loves the name WorkOS because of what operating systems represent: routines for problems no one should have to solve again. That's what WorkOS does at the authentication and authorization layer. It's the foundation that lets you focus on what makes your product different.

This interview was conducted at AWS re:Invent 2025.