Last year, a venture-backed startup raised $17 million to build what they called the world’s first AI-native investment bank. It was a shock and awe moment for the industry where buyers, sellers and M&A advisors alike, all found themselves wondering if AI can actually close deals better than humans.

I’m not writing this to dive into a competitor. I’m writing it because that gap…between launched and closed…is the most important thing a founder needs to understand about AI in M&A right now.

AI technology in M&A is real. I know this because we’ve built some of it. Our company worked on creating proprietary AI-driven capabilities for certain portions of the M&A process. We originally planned to take it to the point of nearly autonomous deal execution, however we realized that the data-driven requirements of M&A are no longer the complex part of our role. AI had reduced the proprietary data barrier to a constant across all industry players. Rather, the ability to build trust, intimately understand the needs of our clients while protecting their psychological safety in order to get a deal done is paramount in order to deliver successful M&A outcomes in this increasingly complex and noisy marketplace.

AI is genuinely compressing the analyst layer of mid-market M&A. Buyer universe generation, document automation, financial models, CIM drafting, data room organization…these are tasks that used to take junior bankers weeks and now take hours. Any advisor who tells you AI isn’t changing their workflow is either lying or falling behind. It’s made us faster, more effective and capable of focusing on the human side of the transaction with more clarity. The firms that survive the next decade will be the ones who use it as infrastructure to do exactly that.

But AI infrastructure isn’t a transaction. And this is where the narrative around AI-native M&A advisory breaks down.

The challenge with the notion that mid-market M&A is a logical, information-driven and binary process is that it completely ignores the complex realities that plague the mid-market. And most notably, the assumption that mid-market M&A can be boiled simply down to the pursuit of a highly well-run company by a highly sophisticated buyer; further compounds this issue. The stark reality is that many mid-market companies are built around the whims of their founder, and that creates a complex operational layer that an AI program cannot easily decipher. For example, we advised on a sizeable mid-market transaction in a town of 5,000 people about three hours from the closest urban centre. The company had serious positives about it and equally as many challenges. The data supporting the deal would not have supported the realities underpinning the attractiveness of the asset to the buyer — it would have been deemed too high growth, too speculative, with operations too decentralized to manage, all while being in a remote area that further negatively impacts valuations. In short, the face value of this deal was poor but the underlying schematics were very strong. And one would need to spend time in the company, meet its leaders, meet the people, touch and feel the products they produced and see its facilities in order to determine this.

In mid-market M&A, the seller sitting across from you made a decision that often took them twenty years (or more) to arrive at. They are not generally optimizing for speed or even low cost. They are optimizing for certainty and clarity…certainty that the buyer is right, that the structure is fair, that the life they built will be treated with care by whoever comes next. For the seller this is not a business decision, it is a life decision. No algorithm closes that gap. No buyer list of 2.5 million data points tells you that the founder across the table needs thirty more days not because of diligence, but because their spouse isn’t ready, or they are struggling with an internal conflict in the company they don’t wish to disclose until it’s resolved. These are the human elements that can be picked up in meetings…not those that can be picked up in emails or data.

Back to the thinking that mid-market M&A is logical and data-driven…the vast majority of deals I’ve seen fall apart at the one-yard line for reasons that had nothing to do with price, structure, or terms. For example, we once had a $35M+ deal to sell a glazing contractor to a very well capitalized PE fund. On paper, this deal was wonderful. However, there were three shareholders in the company: the founder, long since retired, who owned most of the company; the son-in-law who worked in the business and would be retiring shortly; and the general manager who owned a minority position but was the company’s lifeblood for sales. The general manager felt that the $5M he was receiving was not enough given he was being asked to stay on board for five years while the other shareholders were riding off into the sunset. As much as we tried to justify the numbers to him, he chose to blow up the deal because of the unfair internal dynamics he felt were occurring.

That was not a data problem. That’s a human problem, and it requires a human in the room.

But AI can be highly effective. Here is what AI does well in M&A: it can find buyers faster, it processes documents more efficiently, it removes friction from the parts of a transaction that are genuinely mechanical, it speeds up analyst teams, it allows us to run more scenarios. For a small, domestic, asset-light business in a commoditized sector, an AI-assisted process may well be the right fit. If you’re selling a small HVAC company with clean books and a motivated buyer pool, the algorithm might serve you fine. However, if you’re reading this newsletter, you’re likely not selling a small HVAC company.

The mid-market founder we work with has built something complex…cross-border relationships, key-person dependencies, proprietary processes, sector-specific buyer pools that require judgment to navigate, ingrained intellectual property...not just a database query. And most importantly the mid-market founder is overwhelmingly complex. Their transaction will surface business and personal issues that require creative structuring and an ability to handle personal challenges with empathy and rigor. Their buyer will need to be managed through uncertainty. Their advisors will need to sit in rooms where the conversation goes sideways and know how to bring it back.

As I write this, we are working on a sell-side transaction that has been plagued by delays. All parties are tired and managing the psychological health of our client and the buyer has become a paramount role in the transaction. This is the type of human-driven, empathetic relationship that is needed to ensure a deal gets done…even when all parties are tired, ready to walk, and can no longer see the light at the end of the tunnel.

The promise of AI-native advisory is speed and cost. Those are real benefits. But only 20% of companies ever sell, and speed is only valuable if you get to the finish line...cost is only savings if the outcome is the same, which is nearly impossible to measure in M&A other than closing deals. And speed is always limited. Anyone can send hundreds of emails to potential buyers, however the industry is going backwards to a preference for relationships and what they mean for all parties involved. Simply having the ability to send out mass emails is no longer a competitive advantage and is increasingly going unread.

Don’t get me wrong...I am not arguing that mid-market M&A advisors are immune to disruption. We’re not. The advisors who treat AI as a threat rather than a tool are going to lose ground to the ones who deploy it intelligently. But there is a version of this story being told in the market right now…one where AI can replace the judgment, the relationships, and the human infrastructure of a well-run M&A process, which is not supported by the evidence.

The algorithm is in the room. It just can’t close the deal.

The firms that will define the next decade of mid-market M&A are not the ones who resist AI, and they are not the ones who outsource their judgment to it. They are the ones who understand exactly where the line is...and who invest as deliberately in the human side of their practice as they do in their technology stack. That means advisors who understand psychology as well as process. Who can hold a founder’s confidence steady through months of uncertainty. Who know that the most important conversation in a transaction is rarely the one about price.

I’ve spent my career on that side of the line. And I’ve never once seen a deal close because of a database.

‍