Does AI Offer an Advantage in Deals?

Can AI be effective in identifying mergers and acquisitions candidates and getting deals across the finish line? First and foremost, your data needs to be accurate and relevant, according to SMU Cox Professor of Practice in Strategy Frank Fan Xia and coauthor Professor Gordon Walker, the Bobby B. Lyle Endowed Professor of Entrepreneurial Studies at SMU Cox, analyze when AI works well in deals and when it does not. They found deal performance improved when firms had alignment in technology capabilities and were from a related industry.  

 “We're seeing companies, consulting firms, and investment banks using AI and machine learning (ML) for merger and acquisition activities,” Xia observes. AI has tremendous potential in mergers and acquisitions (M&A), he adds. Companies like Goldman Sachs, JP Morgan, and Blackstone have their own AI systems to analyze deals, or M&A projects. AI and ML can offer an advantage because of its ability to handle so many documents in a short period of time and update information dynamically. “This used to be really expensive and time-consuming if you wanted to hire professionals to do those tasks, taking much longer without AI," Xia says.

The foundation of this study stems from the notion that AI and machine learning mechanisms need relevant data to make efficient analysis and decisions. If you feed AI algorithms with irrelevant data, it's not just less efficient—it's actually costly.  “Using AI consumes electricity, time, and investment in the AI algorithm itself,” notes Xia. “In using off-the-shelf data, you may have false results, given the lack of domain expertise.” 

For the study, Xia manually searched all the financial reports of publicly traded companies in the U.S. He identified the companies and time periods when they actually utilized proprietary AI and machine learning technology or products. “That took me about half a year to finish because it’s a lot of reading," he relayed. He searched for terms such as “artificial intelligence,” “machine learning” and “deep learning” to determine firms’ usage of the technology.

Xia tried to use AI tools to sort through reports, but the results were not satisfying. “If you use ChatGPT, for example, it's very useful but unable to understand your specific situation,” he adds. Xia compares this to using Microsoft Office. Though efficient, it only offers strategic parity, not advantage, because other companies can use the same office software or AI model. To fully understand performance, Xia aimed at proprietary machine learning technology and products in the study.

The study examined manufacturing companies across the United States, excluding utilities and financial firms. The authors measured performance using return on assets (ROA)—a metric Xia considers "more realistic" than stock market measures. "You have to generate real profits to increase return on assets," he adds, referring to valuations that can be volatile." 

“Overall, when the acquirer had employed proprietary machine learning and AI before the acquisition, the M&A had a positive impact on performance, improving the acquirer's ROA," Xia reported. However, this benefit was concentrated in deals involving related industries. 

This performance boost was most evident in “within-industry” or related acquisitions. In fact, the more related the industries of the acquirer and the target firm were, the better the impact on post-M&A performance. Conversely, the impact would actually turn negative when the two industries were more distantly related. As Xia noted, "If you actually try to acquire a target in an unrelated industry, then performance was worse." A good example of this would be Coca-Cola trying to acquire Tesla, as the acquirer's machine learning tool would not be able to be fed with relevant data.

The best performance outcomes occurred when an acquirer with proprietary AI technology purchased a target with similar capabilities and were in related industries. Xia says this suggests a potential synergy exists by merging the data of the acquirer and target. By merging their data, the two relevant data sets are more complete, which offers an advantage in M&A performance. “That is something I didn't expect in the findings," Xia surmises.

As companies race to integrate AI into their M&A processes, this research suggests that technological capability alone isn't enough. Success depends on deploying AI where relevant data exists—and recognizing its limitations. For dealmakers, AI is a powerful tool, but best used with firms having aligned technology capabilities and industry synergies.

The paper “Machine Learning, Diversification, and M&A Performance,” is a working paper authored by Frank Fan Xia and Gordon Walker of Southern Methodist University’s Cox School of Business.

Written by Jennifer Warren.