Sramana Mitra: What have you invested in that has passed all these diligence tests and emerged as something that looks like it could stand on its own without getting blindsided.
Piyush Kharbanda: I’ll just give you our framework and our thinking around how we invest in AI and AI application side.
If you look at the basic rudimentary way of thinking about this, there’s the model layer, there’s an infrastructure layer and the application layer, right? I think we are not carrying the bazooka to solve at the model layer. I think, and this is my personal opinion, the infrastructure layer is in a flux and will remain in a flux for a long time. The way the model layer changes will have a very lasting impact on the infrastructure layer.
Very little of what we see at the infrastructure layer today will be the standard in the near future. So, we are not actively looking at that space right now unless we find something that is extraordinary and that’s not an active area of thinking.
Where we are focusing a lot on is applications and specifically vertical applications. Till very recently, a large part of what we’d been doing was where vertical applications were augmenting humans or a human-in-the-loop led application.
I’ll give you an example. One of the very recent investments we made was in a company called Hakimo.ai. They take physical security as a service to small business owners across the US. A classic example of what they do is, they go to a car dealership, and they say, “We will put technology on any camera that you have, and we will augment it with a human in the loop. If there is any break-in or any kind of an incident, we will run this through a stack of escalations culminating in a human who will call the cops.”
So that’s a human in the loop system, which not only solves a key problem, which had not been solved till now but also gives you an outcome to a segment of users who are not ready to trust technology to do it end to end.
We’ve done something very similar in the construction industry where there’s an entity called an estimator when it comes to construction, plumbing, HVAC, et cetera. This company called Attentive has built a product called Beam that augments the workflow of this estimator by providing them estimates based off of drawings. They would’ve spent hours just to get an estimate. These guys give a very highly accurate estimate using machine learning and ai.
It’s not just necessarily just generative ai. There’s a lot of traditional machine learning also at play in some of these applications.
These are a couple of examples of what we’ve done, but to round it all off, vertical AI as boring a sector, as much of human augmentation and replacement as possible, workflows that are not being touched by incumbents, is a core thesis for us.
This segment is part 5 in the series : 1Mby1M AI Investor Forum: Piyush Kharbanda, General Partner at Vertex Ventures
1 2 3 4 5 6