AI Startups: Full-Stack Solutions and SaaS in Drug Discovery

This article, titled “AI start-ups: Don’t sell picks and shovels, dig with them,” compellingly argues that AI startups should not just sell their tools but build full-stack companies to directly leverage their technology. There are several reasons for this, but the two most important can be summarized as follows: Existing companies face various challenges in fully understanding and adopting AI technology (hence, startups can gain a relative advantage), and if AI technology can truly be a game-changer, it’s worth betting on it. The analogy is that it’s better to dig for gold yourself than to sell shovels, as the chances of striking gold are higher. For instance, instead of selling AI tools to assist with marketing, the suggestion is to start a marketing company that uses these tools directly. However, the author seems to soften their stance by stating, “All I suggest is that founders consider some out-of-the-browser alternatives.”

From my perspective, this could be interpreted as suggesting that developing tools for AI-driven drug discovery and selling them as SaaS might not be as effective as using the tools to discover new drugs directly. Back in 2019, many (if not all) of the numerous AI drug discovery companies likely had this thought. By 2024, many of these companies have shifted to a SaaS business model as far as I know. What accounts for this shift?

Ultimately, the key point is how closely the core value of the business model is linked to the value provided by AI technology. The goal of drug discovery is to provide preclinical candidates (along with sufficient patent support). If the intention was to use AI technology for drug discovery, then the AI must either provide preclinical candidates or play a crucial role in that process. If this premise does not hold, the gap between the core value of the business model and the value provided by AI technology will feel as wide as the (if any) inefficiencies of legacy companies.

To bridge this gap, there are two choices: either expand the value that AI technology can deliver (develop broader and more powerful technologies), or narrow the core value of the business model (focus on specific tasks within the drug discovery process). Given that startups should have an obvious problem to solve, choosing the latter option is usually more appropriate, but with sufficient resources, one might choose the former. Either way, there is no one-size-fits-all answer; however, attempting to do both or mixing goals with reality is clearly a wrong approach.

Starting or joining a startup means pursuing ‘high risk, high return’ by comparing the difficulty of a problem with the value it brings when solved. It’s important to have a realistic judgment explaining the gap between thinking about the enormous value of new drugs and actually working on just one task in the long journey of drug discovery and development. Risk and return are inevitably proportional; as problems become simpler, solutions also become simpler.

Recently, JUMP AI 2024 competition was announced (this link is in Korean and this link is my short thought on this written in Korean too.) Last year’s inaugural event focused on creating metabolic stability prediction models for small molecules, while this year’s theme is predicting IC50 values for IRAK4 provided by Daewoong Pharmaceutical (the data will be disclosed in early August.) Predicting IC50 for IRAK4 is a small and (relatively) simple problem, whereas developing new drug candidates inhibiting IRAK4 is a large and difficult problem (regardless of how attractive the target may be). There could be other intermediate goals set in between. The distance between participating in an open competition and achieving good results versus building a company based on such a business model and successfully running it is vast. Of course, they are entirely different endeavors and cannot be compared on the same scale.

Key takeaways:

1. Core Value Alignment: The success of an AI-driven business model, especially in complex fields like drug discovery, hinges on how closely the core value of the business aligns with the value provided by AI technology. Without this alignment, the effectiveness of the AI tools in driving the business’s main objectives can be compromised.

2. Strategic Focus: Startups should focus on either expanding the capabilities of their AI technology to deliver more value or narrowing the scope of their business model to specific tasks where AI can make a significant impact. This strategic focus helps in addressing clear problems and leveraging AI effectively without overextending resources.

3. Risk and Return: Engaging in startups involves a high-risk, high-reward approach. It’s crucial to have a realistic understanding of the challenges and value associated with solving specific problems within a broader process, such as drug discovery. Simplifying problems often leads to simpler solutions, thus balancing risk and return more effectively.