If you’re building just the infrastructure, just the tooling and the nuts and bolts, you look like a software company and somebody else deals with the AI/ML application. And that’s fine. But let’s say you are ingesting the data, cleaning the data, labeling the data, there’s a lot of variable cost to doing that. Every customer may have a new data set. What happens is that this impacts the margins of your business. It looks like you have lower margins because for every customer you have all this work to do. You have to make a decision early on. Do you want to be the one that’s doing that work, or is that something you can offload to the customer?
Let’s say you go to a new customer and say ‘listen, we’re going to take all your data, we’re going to clean your data, we’re going to do you models and solve your problems.’ In that case, you internalize all of that and as far as your organization, you need to know this is basically a services arm. Another option is you can say, ‘customer, we’re going to give you all these tools, but you’re going to have to bring in your own data, you’re going to have to hire people to label it, your gonna have to learn to tune your models, and we’re going to help with all of that, but you’re the one who is going to sink that cost.’
… It’s hard to do one of these companies right now because we are in a transitional time. A lot of the customers don’t even know what they’re asking for, and they’re looking for that help.
That was from an extremely interesting conversation on the a16z podcast with Peter Wang and Martin Casado, Reining in Complexity: Data Science & Future of AI/ML Businesses. So much more in the episode, required me to slow down to 1.0x speed and worth another listen.
They present an interesting conundrum for AI/ML. If you’re company provides infrastructure and tooling, then you can get the big margins of a SaaS company, but a lot of these things seem to be commoditizing. So it’s difficult to find a competitive advantage.
But then when you do AI/ML applications, each customer has their own use cases and data that won’t necessarily generalize or provide economies of scale. For enterprise B2B and B2G, you’d expect for AI/ML contracts to look like professional services contracts.
I think they ultimately hit on what the future will hold, which reminds me of what Lt. Sean Lavelle told me. The enterprise customer is going to have to bring a lot of data work in-house. That competency is critical for success and moving fast. There will be firms that provide cloud computing, data flows tools, algorithm libraries, and whatever else that look like software or content companies. And of course there will be data and model consultants that look like professional services for existing enterprises.
The DoD is starting to lean into AI/ML. I suspect the JAIC should work with industry to set up enterprise infrastructure and tooling. I suspect military and civilian personnel should take on education and training to do the basics of data and modeling for the widest range of use cases. The DoD will start being less about “boots on the ground” and more about development, logistics, decision making, experimentation, and so forth.
Leave a Reply