Investing in Robotics: What Drives Defensibility?

Happy Sunday and welcome to the Investing in AI newsletter. Also be sure to check out our podcast. The last two episodes are both great - Dave Girouard from Upstart and Karim Lakhani from Harvard Business School.

If you want to join an angel list syndicate focused on investing in AI, I suggest ScalingAI, which I’m a member of.

— Interesting Articles —

Microsoft Uses GPT3 To Let You Code In Natural Language. TechCrunch.

The Ghost Work Behind Artificial Intelligence. Slate.

One Study Finds Business Leaders Are Disinterested in AI. ZDnet.

TinyML: What Is It? Dataconomy.

How The Financial Industry Can Apply AI Responsibly. IEEE Spectrum.

Voice AI Politeness. Voicebot.

Startups Using AI to Tackle Climate Change. Forbes.

— Commentary —

I think of early stage investing like I think about counting cards in Blackjack. You can’t guarantee a win on any given hand, even if you have an edge, but if you do have an edge, you can win more than you lose over the long term.

I use that analogy a lot when evaluating robotics deals. I’ve invested in several robot companies over the years and while I am very bullish on the prevalence of robots and the work they do down the road, I’m less bullish on the ability of an investor to make money in the space.

Most robots are built to be direct labor replacements. There is a task that is done by a human and now a robot will do it. This is happening in many areas of manufacturing and agriculture, and will continue to accelerate through all types of physical labor. But knowing many jobs will be replaced by robots doesn’t make it easy to pick winning companies. Why? Because when robots are just direct labor replacements, they rarely have any special defensibility or competitive advantage and thus, the winning companies tend to be the ones who had the best luck or timing, or simply raised the most money the fastest.

Contrast that with other types of robotics companies that enable things that were not possible before. Think of robots that do tasks and, as they do them, scan and monitor the world they inhabit to collect data at levels not done before. Now we don’t just have automation - we have a sensor to provide real-time data and a new data set to use for other tasks.

Let’s take an example. Say Blue Apron decided that not only would it sell you a meal kit, it will also sell you a robot that opens the door, picks up the meal kit, goes to the kitchen, and prepares it. This saves you time. But is it a defensible business? To answer that I’ll ask - how difficult is it to change out the robot to a HelloFresh robot? Probably not that hard.

Now take a robot that replaces a farm laborer. Not only is the robot performing a task, but it’s inspecting the crops every day with cameras and sensors. This allows a data layer to be built on top of the robot’s work. This data layer gives the company another product to sell, and a more defensible position. Rolling up all of that historical data and correlating it to crop conditions at harvest is valuable. Those are the kinds of robotics investments I prefer to do.

This second kind of robotics company doesn’t always win. Building companies is hard and fraught with risk, and so is investing in them, but, it provides a more likely chance at a nonlinear outcome. And while the first type of robotics company will have some winners too, I just find that for me, who will win is more difficult to predict.

In investing, you have to play where you have an edge, and I don’t see an edge in direct labor replacement robots.

Thanks for reading.

@robmay