The math problem hiding in your AI agent strategy
When each step has 98% accuracy, a 20-step process drops to 70% reliability. That compound probability problem explains why AI agents excel at narrow tasks but need deterministic scaffolding for enterprise workflows—and why hybrid systems win.
The New Stack's analysis cuts through the AI agent hype with a simple calculation: when each step in a process has 98% accuracy, a 20-step workflow drops to under 70% reliability. That's the compound probability problem nobody talks about when promising that agents will replace enterprise software. The article walks through why agents excel at narrowly defined tasks—research, data retrieval, invoice processing—but fall apart on multi-step enterprise workflows without what they call "deterministic logical scaffolding." That's structured rules and workflows that guide AI decisions rather than letting probabilistic systems roam free.
The practical answer isn't waiting for perfect AI. It's building hybrid systems where deterministic logic handles known workflows and AI adapts within tight boundaries. This tracks directly with what I wrote about evaluation infrastructure—you can't deploy agents without knowing when they'll fail. It connects to the governance frameworks needed when agents act autonomously, and it reinforces why companies keep building autonomy faster than they can deploy it responsibly. For legal and product teams, this math matters because it defines where AI can work today versus where it needs human oversight, contracts, and liability frameworks.
Peter White
