AI Agents in the Economy, Organizations, and Science
AI agents now do work that once belonged to software, to staff, and to researchers. Many questions in operations management were built for one of these at a time; agents are often several at once, and they differ from all of them. The question is which ideas carry over and which have to be rebuilt.
The questions
Agents inside firms. An agent can be an algorithm a firm runs, an asset it owns, a feature it sells, or something closer to an employee. Which tools for managing each still apply? When should people and agents complement each other, divide the work, delegate, or compete? Agents pursue the goals they are given rather than pay or leisure, so how should they be organized, evaluated, and checked, and how does their presence change what their human colleagues do?
Agents in markets. As market participants, agents behave differently from people and have different characteristics. When they buy, sell, and negotiate for people, or on their own account, which market rules and safeguards still work, and which need to be redesigned?
Agents in science. Agents carry out more and more research tasks, and research is now produced faster than people can read and review it. How do the roles of human researchers change? How should research be verified, evaluated, and credited? And what happens to science when its hardest open problems become benchmarks for showing what AI can do?
The economics of AI services. Using, training, and selling AI models trades cost against revenue at every step: when and how much to train, which model to use for which task, and how to price the result.