Investment Thesis
It seems like everyone wants to “invest in AI” these days. Nearly every pitch deck I see these days have “AI” in the deck somewhere. “AI” has been the “next big thing” for decades and in the last couple of years has jumped the shark in investor interest. To develop an investment thesis in a complex new technology area like this, my first step is to separate the “picks and shovel” (infrastructure/core technology) companies from the “prospectors” (those using the tools to deliver value).
Prospectors can’t create meaningful value before the infrastructure has reached a critical mass. AI investing has “jumped the shark” because certain infrastructure components have reached critical mass. In AI there are many different kinds of picks and shovels for different jobs with each of these categories on its own development track toward “good enough.” The trick to investing in prospectors is to wait until the specific picks and shovels are “good enough” and then bet on an aggressive team with unique insight into a particular customer problem area where huge value can be created. Last summer with the release of GPT3, it looked like natural language processing picks and shovels had reached “good enough” so we went looking for prospectors.
A great prospector will be smart, tenacious, and have some unique insight on “where to look” for the jewels. The founders of OtherSide.AI are all this and more. They are “looking” for GPT3 applications that will improve daily writing tasks we all have to do like email, blog posts, and more (areas where I personally want solutions). As a pre-seed investor, one has to be willing to make meta bets which allow for pivoting over time into a product that at the time of the investment is as yet not fully formed. The meta bets we are making with OtherSide.AI include:
- The picks and shovels (GPT3, internet connections, etc.) are “good enough”
- The team is smart, tenacious, and is looking in the right place.
The Story
Matt Shumer, like many of us, hates email. “I hate email so very much,” he told me on our first call. “I came across GPT2 and immediately I was obsessed with it. I started using it and didn’t sleep for like three days. I built a summarization tool, action tools, writing tools, so much stuff. Because it was fun as hell. I hacked together a bunch of stuff over the next months, but it was not very elegant. Then this July (2020) we got access to GPT3 and ended up building a demo that was, I mean, miles better than what we spent months building previously. Like it wasn’t even a comparison. I posted the email writing demo on Twitter. Chris Sacca loved it and started tweeting about it. All hell broke loose.”
Matt and a handful of early users started using the email tool in late 2020. He estimated it saved him 50-70% of the time in email. The VCs started calling and the company raised a round very quickly in late 2020. Turns out lots of people hate email and Matt’s hacker approach struck a cord. In addition to email productivity, the company has developed HyperWrite.ai to help people writing fiction, blog posts, and more. The company has other GPT3 enabled applications under development as they “prospect” for the best product/market fit. In the early days of prospecting the trick is to stake multiple claims and then mine those that have the best yields.
Why I am Investing
I have been investing in and looking for email and writing improvements for a long time. Back when Spam was the #1 email problem, I was an investor and on the board of Cloudmark, which started in email and moved on to all message platforms quite successfully reducing the signal to noise ratio in messaging. There has been quite a bit of innovation on the client-side to prioritize, categorize, and auto-respond to certain standard email types (Superhuman, Hiri, SaneBox, etc.). While filtering and tagging have reduced the flow, there has been precious little innovation around the time it takes to actually respond to the emails which are important. Many people (including myself) have turned to their assistants to filter the inbox more and respond to standard mails following pre-defined If/Then templates. All stuff that AI can do much more efficiently. I once read a science fiction book about an AI program that could read and reply to emails intelligently in my own personal style. I wanted that product. OtherSide AI has built it.
Other Side AI is earlier than I usually invest. The company was only a couple of months old when we invested and this is their first round of financing. The closed Alpha was released, and they are using that to nail down the most valuable use cases and the go-to-market strategy/pricing. There are three reasons I am investing earlier than usual:
1. Passionate Hacker Founder. Matt is an experienced founder solving very personal problem in a unique and innovative way. He and his co-founders are building something to make their own lives easier, and in the process going to help the rest of us.
2. Innovative solution to a huge problem. Despite all the filtering, there are still emails that must be responded to in a relevant and personal way. This is the first product that I have found which is actually delivering on that promise. It is me, just faster. If the product can deliver 50% improvement at scale, that is an extra 6 hours a week for the average email user. This type of “auto writing” functionality can and will extend to blogs, fiction, and other forms of writing over time.
3. Smart co-Investors. Many of the best investors are actively looking for innovations that create value in very specific markets with specific technologies (like I have been for writing improvements). They have opinions about where technology innovation needs to go and are looking for the best entrepreneurs to drive those companies. This is how I found Other Side AI, as did Madrona Venture Group, Active Capital, Hustle Fund, Chapter One.
Why this deal is Incisive
Other Side AI ranks very highly in key areas of our Meta Themes.
Software Eats. In this case, software eats itself (the email client). And the AI intelligence eats the manual time we spend on writing. OtherSide.AI is mining for applications that can “eat” 30-50% of the time we spend writing today across multiple use cases. This is a significant reduction in friction which gives people back a significant chunk of time to us all.
Smart founders figure it out. When the team is solving a very personal problem and working with passionate early adopters in a very smart way, I have seen great products come out. The explosive backlog while still in Alpha shows demand. In discussions with the team I have been very impressed at their decision process taking this feedback and putting it into the product road map. The hacker mentality enables them to go fast and pivot when needed.
Americans are lazy.. Americans love anything that makes life easier, lets them do less work. But in writing and email, efforts so far (like text expander, and autoresponders) have not been personalized and lack soul. OtherSide allows me to be lazy while preserving my personal style and not coming across as dismissive. This is the holy grail.
Trends I am betting on
AI “picks and shovels” for Natural Language Processing are “good enough.” Windows 1, 2 and 3 were pale barely useful operation systems. With Windows 3.1 things tipped and took off. It became “good enough.” GPT 1 and 2 were very academic and produced some comical results. GPT3 could be the tipping point and I am betting it is. I could be wrong, it could be GPT4, but that is the bet at this time.
Personalized efficiency wins. Everyone likes getting things done in less time, especially email and writing. We also know when receiving a form letter or reading something the writer “phoned in.” I would pay alot for a writing tool that would keep my style and soul while turning it out much faster. I am betting many people would.
