Download:
Abstract:
Mergers and acquisitions (M&As) offer opportunities for value creation through synergies but often result in value destruction due to imperfect managerial evaluation and decision-making processes. Recent advances in generative artificial intelligence (AI) and large language models (LLMs) present novel opportunities to support M&A evaluation processes. This study investigates the ability of artificial evaluators to discriminate between value-creating and value-destroying M&As. In particular, we focus on LLM agents, which can increasingly operate autonomously. Four evaluator conditions with varying levels of autonomy are examined: (1) a baseline model without step-by-step reasoning, (2) chain-of-thought reasoning whose steps are controlled by the human, (3) autonomous reasoning whose steps are controlled by an LLM agent, and (4) autonomous reasoning involving interactions among multiple LLM agents. Using anonymized announcements of 109 deals between U.S. public firms, our findings reveal that only the multi-agent condition demonstrates a substantive ability to differentiate between value-creating and value-destroying M&As. The top half of deals identified as most promising in the multi-agent condition corresponds to abnormal returns at the deal announcement totaling $571 million, while the bottom half with the least promising deals incurs losses of $1,273 million. These results underscore the potential of multi-agent systems in enhancing strategic decision making in M&As.
Citation
Mirzayev E., Testoni M., Vanneste B. Artificial agents and the evaluation of M&As. Available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5119314