Ash Fontana, Managing Director of Zetta Venture Partners recently published the AI-First Company, a strategy book on AI for the entrepreneur and executive. Knowing Ash and the depth of his thinking on AI, I bought and read the book immediately.
Ash was also kind enough to provide some additional insight for the benefit of the newsletter readers. I hope you enjoy his thoughts and the book. Scroll down for my reflections.
Peteris: AI has been tremendously powerful in enterprise "value" products: running predictions, extracting insights, even saving time by generating text/images and other assets. Equally, AI has been helpful for consumer "engagement" products: recommending things to buy, articles to read or people to follow. Then there are engagement or workflow B2B tools like Slack, Webflow, Zapier, etc. Do you see AI making an impact there?
Ash: Yes, and now is the time it will make an impact. We've been trying to use AI to add functionality and utility to such workflow tools for years but it's mostly been in the form of search, and that hasn't utilized many self-learning methods. I think that we're only just starting to see the potential of recent developments in natural language processing, i.e. transformer-base models, manifested in workflow products. The potential there, to be clear, is to infer the workflow underway from the sequence of phrases people type into such products, e.g. questions they ask a colleague in a chat, what they name a step in a macro, what they type in a search box, etc.
Peteris: Have you read 7 Powers? I'm always looking to fuse different strategy frameworks and see how they match up. In terms of how you think about AI - do you think it fits with the 7 Powers framework (e.g., perhaps data moats being a form of network effect) or do AI-first companies have a different set of powers?
Ash: I have read the book. Generally, I wrote The AI-First Company, because I don't believe that anyone has, to date, articulated the type of competitive advantage built with AI. We tend to use concepts like a network effect, scale effect and learning effect but none of them fully describe the automatically compounding competitive advantage built with AI. That competitive advantage - which I term a data learning effect - is more powerful than anything we've seen before because it gets better with every run of the machine learning model, for example. Data learning effects involve getting a critical mass of data, using internal/learned capabilities to turn that data into information, then putting that information into a network of models that learn as calculations are done in and between different models.
Peteris: What kind of AI-first companies are you most interested in funding these days? Any technologies or market verticals that exist now vs. 2019 that create new opportunities for Founders to jump in?
Ash: I'm completely focused on funding the tools and infrastructure used by people to build AI. There are so many applications, in so many industries, but they need the utilize the expertise of people in those industries to get going; we need domain experts to define the predictive features of AIs and train them up to a sufficient degree of accuracy. So, I'm investing in tools to build, train, deploy, monitor and manage ML models.
As you can glean from the answers, Ash has contemplated existing strategy frameworks carefully in coming up with a holistic approach on how AI creates value in companies. My bias is always to attempt to fuse and integrate different frameworks together and I've long been on the hunt for the default tome to describe "data moats".
AI First does this and more. It introduces the more general term "data learning effects" (DLEs) and describes how they intersect with several traditional moats. It's immediately become the #1 reference on AI strategy for me and it is great to finally plug that (big) gap in the strategy bookshelf.