The CUP Theory of AI Defensibility For Services As Software Business Models

I’ve written a bit about “services as software,” one of the models I like best in a world of AI. The gist of the model is that you take a services business where humans used to provide the service and you either use robotics ( if it’s a physical service) or algorithms (if it’s a digital service) to do it instead. But the rest of the value chain remains the same.

Most people, when they hear about this, think it’s just about cost savings, but it isn’t. It is so much more. And it’s that “so much more” that can provide defensibility to the business model.

My theory here, which is admittedly untested by the market, but that I am investing against, is that defensibility in services-as-software businesses will arise from 3 things: Consistency, Ubiquity, and Personalization, which is why I call it the CUP theory.

Consistency is the value people most often overlook when thinking about services-as-software models. The reason to have a robot or an algorithm provide a service is because it can do the exact same thing over and over and over. If you use the robotic barista CafeX in San Francisco and then again in Atlanta, the experience can be 100% the same in both places. You don’t realize how important this is until you realize how finicky humans are about many of the services we consume.

That brings me to Personalization next. Services-as-Software companies don’t personalize the way a Netflix recommendation engine does. It isn’t about “what people like you may like” next based on what other people like you did. It’s about learning what your actual preferences are by interacting with you. This is part of the value of machine intelligence – the machines get smarter about you, meaning they can provide you more personalized service. And, to the point above, that service can be consistent across time and geography.

But is it defensible? Can’t I go build and deploy many more machines that are like CafeX and then you will take whichever one is nearby?

That’s where Ubiquity comes in. The defensibility comes from the fact that Consistency and Personalization are valuable enough that, if there is a CafeX nearby, I don’t see the need to use a competitor. My theory then, is that defensibility in some of these services-as-software businesses will arise from the combination of Consistency, Ubiquity, and Personalization.

This raises an interesting question for companies building services-as-software models, which is – how fast must effective ubiquity be achieved? If these companies want defensibility, are they in a race for market share? And if so, how does that impact financing milestones?

My personal belief is that early companies in this space should figure out what “ubiquity” means from the user perspective for their service. If it means worldwide geography, then you probably have to raise more money faster than if it is purely online, delivered at home, or only applies to local geos. I also believe that there will be some market adoption time needed by consumers to shift to accept these models, and while it will be short, it does exist. I don’t think companies can scale until that transition happens. So balance your fundraising against that backdrop.

I used CafeX here as an example because I am not an investor, so I don’t know if CafeX thinks this way about their business, and if it is indeed defensible long term, but I am invested in a handful of early stage AI services-as-software companies that are thinking about it this way, and once they launch I will write more publicly about whether this thesis pans out. If you are building a services-as-software company, I’d love to hear from you. At PJC we think a lot about this emerging business model and it is one of our favorites.

Leave a comment

Your email address will not be published.