Think about it. Whether you are applying for time off a mortgage or a driver’s license, the process is ruthlessly rule-based. Check the right boxes, provide the correct documents, follow the sequence, and you get the approval. Miss a step, and you don’t. This isn’t creative chaos; it’s the bedrock of operational efficiency. Most knowledge work, despite what the futurists claim, is about finding the right answer reliably, not generating a novel one.
Now, a thoughtful objection might arise here: Isn’t all knowledge work fundamentally probabilistic? People interpret, forget, misremember, and improvise. The deterministic fantasy was the expert system era, and it failed because work is messier than rules.
That objection is worth taking seriously. It is true that human knowledge has never been purely deterministic. The expert systems of the 1980s collapsed precisely because they tried to freeze-dry messy human judgment into brittle rule sets. And it is also true that agentic AI’s probabilistic nature could, in theory, handle ambiguity better than those old systems ever could.
But that is not an argument for abandoning determinism. It is an argument for knowing where each belongs. The fact that human work is messy does not mean payroll can be probabilistic. It does not mean regulatory filings can vary by mood. The real error of the expert system era was not determinism itself—it was incomplete rules. Today’s risk is the opposite: We have agents that are too flexible, running on too little accountability, deployed into environments where variation is not a feature but a liability.