AI agents and the boundaries of the firm
Transaction costs are back as a growing keyword in the management journals, while agentic AI and language-model agents rise in AI. Coase argued that firms exist because using markets is costly. If AI agents can search, negotiate and monitor contracts cheaply, do firms buy more and make less?
What prompted this
In the management journals, "transaction costs" had 6 papers in July–September 2026 against about 1 expected (see the management dashboard); in the AI journals, "agentic AI" (11 against 5) and "LLM agents" (14 against 8) are growing. Since October 2025, 170 papers carry a transaction-cost keyword (65 in finance journals, 49 in economics, 37 in management), and only 7 of them carry an AI keyword. "Agentic AI" appears on 41 papers, 27 in AI journals and 11 in management.
Questions
- Which transaction costs do AI agents lower most (search, bargaining or monitoring), and which do they raise (verification, security)?
- Do firms that adopt AI tools for procurement and contracting outsource more and integrate less?
- Does the change favour small firms, which can now buy what they could not organize, or large firms with more data?
Data that could help
Firm-level purchases of goods and services (procurement records, firm-level input data), AI adoption measures (job postings, software use) and supplier counts.
Next step
Track the number and type of suppliers per firm before and after adoption of AI procurement software, for firms whose adoption dates vendors have published.