What happens when you put a fast, calibrated System One model next to an LLM planner in a public procurement assistant: latency, costs, and early insights from live use. Most of what an agent does between a user’s message and a useful answer is not deep reasoning. It is a string of small multiple-choice decisions: which lane does this message belong in, which of these passages actually answers the question, which of these CPV codes fits the purchase. We had been paying a frontier-class planner roughly 7.5 seconds and about two US cents for questions that already had a handful of possible answers. Since 24 September 2026, three of those…
System One Next to the LLM: One Week of First Insights in a Procurement Assistant
Advertisement