Watson’s high-profile Jeopardy win in 2011 seemed to be the eye-catching opening number IBM needed to introduce their AI venture into the healthcare space and turn a game show trick pony into a formidable business offering. Over the next decade, IBM fought to live up to the initial wave of interest in Watson, which experts at the time were calling an “inevitable” human replacement and a “vindication for the academic field of artificial intelligence.”IBM also seemed to forecast the current explosion of activity in the M&A world as it funneled well over $4B into the acquisition of healthcare data and analytics businesses, including Truven Health Analytics, Phytel, and Merge Healthcare, in order to build a wide range of capabilities around their core AI technology. But after nearly a decade of disappointment after disappointment, IBM announced earlier this year that they are looking to sell Watson Health.Part of Watson Health’s downfall lies in the technology itself. For instance, Watson’s performance in Jeopardy and its application in the medical field is based on natural-language processing (NLP), a branch of AI that allows a computer to quickly analyze language-based data. In theory, Watson would use NLP to process free-text patient medical records and medical literature to find patterns in data that physicians could not. In practice, it was unable to pick up on subtleties in notation and language, compromising its clinical support function.And while Watson’s executives touted its success, telling reporters that the rollout and development of Watson was “going fabulously,” internal slide decks presented by IBM Watson Health’s deputy chief health officer revealed that Watson’s oncology software offered recommendations that were erroneous and unsafe. Physicians at the hospitals piloting the AI continued their advertising campaigns featuring the cutting-edge “Dr. Watson” in order to drive up revenue while quietly telling IBM that their technology was anything from subpar to a “piece of s—.”But while gaps in the technology could be rectified, what ultimately sunk the Watson ship was the inability to successfully integrate acquired technologies and businesses. Watson effectively “bundled a wide range of assets around one tech stack” while failing to produce one technology that held greater value than the sum of its parts.After IBM acquired Phytel, Explorys, and Truven in 2015 and 2016, they had more patient data and customers than ever before, with the acquisition of Phytel alone bringing in 150 new clients and the acquisition of Explorys adding more than 315 billion patient data points. Their first step was bluewashing, a phrase used internally to describe the IBM-ification of the new acquisitions. For almost an entire year, engineers on the team worked on merging databases from Phytel and Explorys, putting the core healthcare database functionalities of both technologies on pause. While product managers at IBM treaded water, trying to figure out how to integrate Phytel and Explorys into a better combined product, smaller companies moved quicker. Even after they previewed the new product—with AI integration noticeably missing—customers even began asking for the original Phytel offerings.According to the acquisition announcements, Watson’s AI capabilities would uniquely be able to leverage and “surface new insights from the massive amount of personal health data being created daily.” The bold promise rang empty - even with troves of new clinical, research and social health data, Watson led with the promise of insights, even before realizing those insights in the first place. This combination of putting the cart in front of the horse and losing momentum during the lengthy database merging process made IBM’s approach more similar to a haphazard Pokémon card collection strategy than a long-term corporate strategy. Now, as M&A activity in healthcare is expected to increase in 2021, with a total of 162 deals of more than $10M each and a total value of $63.4B emerging in 2020 alone, understanding how to use acquisitions as an effective method of scaling an existing business will become only more important. So what can we learn about platform building strategy from Watson?First, and most fundamentally, companies need to make sure that their core offering (and team) are mature enough to turn an M&A deal into one stronger and cohesive offering, as opposed to multiple verticals that are united only in name. Watson put $4B towards acquisitions instead of towards internal research and development, even though no clinical studies proved that AI actually improved patient outcomes or clinical decision making. As a result, no acquisition could offset the surmounting doubt that Watson (and AI in general) could not be a serious clinical support offering.Next comes an understanding of how products integrate with each other. Companies and their product management teams need to identify the synergies between their product and any potential acquisitions before the deal is signed. Though the increasing and dominating presence of private equity firms and the desire to snag smaller companies hit hard by the COVID pandemic might add to the pressure to jump the gun and quickly enter into a M&A agreement, Watson shows us that the first step is to realize the integration vision. Without a core vision of the role Watson was going to play in the digital health space, it was difficult for IBM to identify the way it would integrate with other platforms, even if they were the best in the market, like Phytel. The premature debut of Watson on Jeopardy began a period where Watson’s public perception was always two steps ahead of its actual capabilities, and the company was forced to spend time and resources trying on a number of hats for the platform, from pathology to oncology. As a result, even after Watson acquired troves of new data, technical infrastructure, and customers, the group was unable to use them to make Watson AI a better offering.Finally, Watson’s failure underscores the importance of cultural and organizational fit when it comes to acquisition. As a comparatively slow-moving company, IBM’s method of bluewashing was frustrating to employees from smaller companies who were used to accelerated decision making. Engineers also cited misalignment with having to work under new management without technical backgrounds who often didn’t understand how to take full advantage of Watson’s AI technology, as well as internal competition between IBM’s different healthcare teams. IBM’s integration shortcoming isn’t specific to AI technology, nor is it exclusive to healthcare and digital health. “I’ve seen many acquisitions happen at big companies like Intel, IBM, even Google. Unfortunately, most of them fail,” said Leon Tsirkel, former Head of Biosensing at Intel. “The reason is that when you have a big corporation try to acquire a smaller entity, there’s a fundamentally huge culture clash.”Learning how to effectively integrate acquired teams and technologies is an ongoing and iterative process for companies and their managers. The industry is now pointing towards a hands-off approach, at least in the initial stages. “For the first year after the acquisition, we asked the ‘mothership’ to be completely hands off, and allow the acquired entity to continue executing toward their commitments and plans. Many companies spend lots of money, but don’t get the key value from what they acquire and many times end up spinning-out or selling the acquired entity,” said Tsirkel. The best practice is to allow the acquired company to leverage the larger mothership capabilities they need, as opposed to the other way around. “This is the new mentality and it’s more successful,” concluded Tsirkel.Ultimately, the “for sale” sign in front of Watson Health’s door is unlikely to deter future waves of M&A activity in the digital health space. But it certainly stands as a forewarning of what happens when companies don’t look before they jump into the deep end and acquire before their product and their management are ready.