George Williams works at Kognitos, where the work is deterministic automation for enterprise finance operations: procure-to-pay, accounts payable, reconciliations and the month-end close. He writes about where rule-following execution belongs in a business process, and where probabilistic AI does not.
The recurring argument across these essays is that enterprise AI has made individual tasks faster without making processes shorter, because a process is a sequence of handoffs rather than a pile of tasks. Reliability multiplies across steps instead of averaging, so a chain of individually impressive steps can still fail most of the time. Finance work that has to reconcile needs execution that returns the same result on the millionth run as on the first, and that is a different kind of system from the one that writes a passable memo.
What he writes about
Enterprise AI and intelligent automation, aimed at finance and operations leaders rather than engineers. Recurring subjects: the distinction between probabilistic and deterministic AI, the month-end close, accounts payable and procure-to-pay, controlled autonomy and where a process should stop and escalate, explainable execution as distinct from explainable models, and how AI is changing the economics of services and channel businesses.
Where to find the writing
Essays are published here first and are also available by email through Substack. This site is the canonical version. The same pieces also appear on Medium and LinkedIn.