Autopilot: Automating the Dealmaking Process

The rapid advancement of machine learning capabilities means that many junior-level tasks in the dealmaking process can be automated.

Recently, it’s been reported[1] that Goldman Sachs has stepped up the presence of their famous, or perhaps infamous, “strats” within their investment banking teams. These “strats”, as they’re called internally, are a team of programmers that are deployed to increase efficiency across various areas of the bank, usually through various forms of automation. Less optimistically, they’re also known for drastically reducing headcounts, by effectively coding jobs out of existence – notably, they’re the people behind the drastic reduction of trading positions available at the bank. At first glance, investment banking and dealmaking seem less able to be automated; after all, deals are highly unique, with many moving parts and specialized circumstances. While this is true, the recent rapid advancement of machine learning capabilities means that many junior-level tasks are able to be completed by a computer. Having seen what happened to their brethren in junior trading positions, some junior bankers fear for whom the bell tolls, and are asking some important questions regarding this increasingly popular trend…

What Can Be Automated?

Obviously, the first question people ask when their jobs are threatened by automation is what parts of their day-to-day life are at risk. For M&A teams, presentation building is at the forefront – it’s a grueling task, one that takes hours, and, while vitally important, doesn’t require the most complex thought processes, making it the perfect candidate to be automated. Given the ability to decide on visual details and insert custom-made slides when necessary, getting a computer to complete these processes would free up staff to add more value elsewhere.

Balance Sheet, Income Statement, and Cash Flow projections lend themselves very well to machine learning techniques. By being trained on the results of a multitude of past deals, a properly tuned algorithm would be able to predict base-case, downside, and upside scenarios with greater accuracy and precision than a human could. Going one step further than determining proper ranges for critical variables in models, machine learning can be used to select the best valuation technique to use for a deal, given the results of previous deals. Certain sectors lend themselves better to particular valuation methods more than others, but a properly trained machine could pick up on subtleties on nuances that human workers may miss. Both of these would lead to minimizing situations where money is left on the table, while also reducing the risk of overpayment. Considering both of these scenarios can lead to serious reputational damage for a dealmaker, there is considerable value to be added here with the implementation of machine learning techniques.

The process of finding a buyer can benefit from automation, as well. An algorithm could scour annual reports, press releases, comments by executives, and many more sources to determine whether or not particular companies were not only looking to acquire company, but also help determine if there would be a good cultural fit by analyzing the language used. Needless to say, a computer can run this task far more quickly than a human being, and at any hour of the day.

What Can’t Be Automated?

While most deals have common elements, every company has unique factors and identifiers that still require human input when it comes to the valuation process. Company specific technologies, processes, products – any kind of “moat” of value for a business – is currently still best handled by a human being. It is still 100% necessary to have someone who can predict the effects a company’s intellectual property will have, from juicing sales growth numbers to cost saving synergies.

The biggest, most glaringly obvious part of the M&A world that can’t be solved programmatically, though, is the cultivation of relationships between dealmakers and the businesses they work with. These interactions transcend simple transactional relationships, and many dealmakers end up knowing their buyers and sellers incredibly well. People today are more comfortable with computers than ever before – think of the explosion in the use of automated ordering systems at restaurants, ATM’s at bank branches, even Siri in cell phones – but we are not yet at a point in the development of artificial intelligence to place the level of trust required for something as large and complex as an acquisition solely with a machine. This is where a considerable piece of the intangible value is added, and can only be added by human capital… for now, anyway.

Meeting The Challenge of Automation in the Dealmaking Process

Having seen the writing on the wall, what can be done to protect one’s self from this incoming seismic shift in the industry? The M&A job market has always been ruthlessly competitive and this looks to ramp up considerably as machine learning takes hold. However, as this trend is still in its early phases, there are opportunities to be exploited.

First and foremost, growing your competence in relationship building is absolutely key, even moreso than in the past. The value of human capital in this aspect is more likely to hold in coming years, as opposed to the value of raw work hours able to be put in. Doing this quickly through networking, professional associations, public speaking practice, and other endeavours should be a priority for anyone hoping to remain in the industry.

Secondly, everyone should at least have some amount of coding skills in their repertoire. Programming is quickly becoming the literacy of the 21st century – either you learn it or you get left behind. Now, to be fair, someone with a business background will probably not be able to compete with someone who has a PhD in Computer Science from MIT when it comes to programming skills, but that isn’t what’s expected. Being able to converse with the tech teams as they grow increasingly integrated with the deal teams and liaise between the two sides is going to be a very valuable skill. Many a business leader has lamented the seemingly different languages their technology and business teams speak and how this leads to sub-optimal outcomes. Being able make sure important topics aren’t lost in translation adds considerable value.

Lastly, it’s important not to shy away from these technologies as they’re implemented but instead to embrace them. Head to head, over the long term, the likelihood of being able to consistently beat ever-improving machine-learning algorithms is very slim. However, learning to leverage the output they generate in new, creative, and value-adding ways is what separates us from the machines. With increased automation in the field, the opportunities for “extra-miling” is poised to jump significantly.

In Closing

Bill Gates recently suggested[2] designing a framework for the taxation of increasing automation. While seemingly at odds with the desire to increase efficiency, Gates makes the argument that the economy is not yet prepared for how quickly and massively the changes will be as machine learning is implemented in full. Using the proceeds from the tax to fund retraining for workers, Gates says, would help to stem the tide of workers who are left behind, and make this paradigm shift more easily manageable. Given the pushback from workers we are beginning to see now with respect to automation, it is also preferable to manage the inflow of new technologies into the economy than a full-out luddite movement. Make no mistake, however: this change is coming soon, and it is unstoppable. It is best to meet it head on and shape its outcome, instead of being left in the dust.



Kevan Hartford

Kevan Hartford is a Toronto-based finance professional working in asset management.