That $13.42 AI Marketing VP is a Steal—Until It Goes Rogue on Compliance

Field service scheduling is broken. Everyone knows it, but Jason Lemkin over at SaaStr recently put numbers to the cost side of the AI marketing equation. His breakdown of SaaStr’s AI VP of Marketing—dubbed 10K—is eye-opening. For $13.42, 10K executed 125 actions and read 2,463 lines of code in 61 minutes. That’s a work rate no human can match at a price no human can legally be hired for in California.

But focusing purely on the economics misses the other side of the coin. Of the 47 marketing-related problems we track, 8 fall squarely under compliance and legal risks, with an average severity of 4.1 out of 5. One problem, PROB-5321, is bluntly titled “Marketing AI Generates Non-compliant Content.” The issue isn’t just that agents might spew offensive material—it’s that they may also inadvertently violate GDPR, CCPA, or industry-specific regulations without oversight.

That’s a big deal if you’re pitching AI marketing agents to enterprises. Yes, you can run one for a fraction of a VP’s salary, but the hidden cost of a compliance failure could dwarf those savings. Lemkin’s article touches on the importance of clear specifications, but it doesn’t address the legal guardrails that need to be baked in from day one.

His point about architecture controlling cost is well-taken. Small models for simple tasks, caching over live API calls, and scheduled runs—all smart strategies. But here’s the twist: the same architectural thinking should apply to compliance. Why not a secondary model that audits agent output for regulatory red flags before it goes live? Or a human-in-the-loop step triggered only when the confidence score dips? Our explore page is littered with adjacent ideas—31 in marketing alone—but few tackle this head-on.

Some might argue that compliance is an edge case, that most SMBs won’t care. Our data suggests otherwise. The severity scores on these issues are higher than the average across all marketing problems (3.8), meaning they’re causing real pain. For vibe coders and indie hackers building the next generation of marketing agents, this is a feature, not a bug. Embedding compliance checks isn’t just a value-add; it’s a differentiator that opens doors to larger accounts.

That doesn’t take away from the core insight. Lemkin is spot-on that the bottleneck has shifted from cost to your ability to specify valuable work. But our data adds another layer: you also need the ability to specify what not to publish. Even the most imaginative and clear task-designer can be undone by a single automated tweet that runs afoul of a privacy law.

This is where the "."