Enterprise AI vs Regulatory Reality
Stochastic tools, deterministic regulatory systems, and the path through the minefield
Welcome to the latest edition of Buy the Rumor; Sell the News. We’re closing in on 2,000 subscribers, including institutional allocators, venture capitalists, litigators, senior executives and entrepreneurs. Thank you to all who have subscribed!
In today’s post, I explore the tension between stochastic AI and deterministic regulatory requirements, and how startups and investors can successfully reconcile this tension.
If you want to connect with me directly, my contact information is at the end of this post.
Everyone’s talking about enterprise adoption of AI. What almost no one is talking about is this: Most enterprise systems are designed for determinism, and most AI systems are stochastic.
That tension will define the next five years of enterprise AI. It will kill startups, disorient investors, and confuse buyers. But for those who navigate it correctly, it’s also the wedge.
Let’s unpack this.
The core tension
Enterprises are not monoliths. Some functions, like branding, sales, or marketing, tolerate stochasticity1. They live in ambiguity. They experiment. Their outputs are true enough.
Other functions, including payroll, accounting, or compliance, operate under a fundamentally different epistemic regime: deterministic, auditable, legible. These teams don’t try things. They certify, reconcile, file, post. Their outputs must be replicable under audit.
The mistake many AI founders make is trying to sell stochastic systems into deterministic workflows without understanding that auditability and legibility are regulatory requirements, not bugs to be fixed by AI. This is why an LLM that generates marketing copy is a hit, but the same LLM writing policy summaries for legal or auto-posting accounting journal entries will be dead on arrival.