Herbert Simon published The Sciences of the Artificial in 1969. He had already helped found the field of artificial intelligence, and he went on to win the Nobel prize in economics in 1978. Few people have felt equally at home in both worlds.
The book has a famous small scene. An ant crosses a beach toward its nest, and its path twists around pebbles and ripples in the sand. Sketch the path and it looks complicated. Simon’s point is that the ant is simple. Most of the complexity in the path belongs to the beach.
Look at the beach first
When an AI agent behaves erratically in a business process, teams usually blame the model and start rewriting prompts. Simon would look at the environment first, at the half-documented systems the agent touches and the exceptions nobody ever listed.
He also coined satisficing: settling on the first option good enough for the need. Agents told to find the best answer will burn budget hunting for perfection a person would have accepted ten steps earlier, unless someone writes the bar into the instructions.