Framework to think through AI problems
Artificial Intelligence is coming of age. We as a deep tech AI Focussed fund have met close to 1000 startups in the last 3 years and we find that entrepreneurs are using AI in meaningful ways to solve deep problems. While working with entrepreneurs, we have evolved a model to understand the key value that the startups bring to the table. Our model can be understood by using a framework that can bring more clarity to your product roadmap, GTM, product marketing etc.
There are two dimensions to the framework. One dimension captures whether the application augments, scales, or transcends human capability, since at a high level, AI technologies end up mimicking human perception and ability.
The other dimension captures the impact the startup is creating as a result of their innovation. Largely we have seen that it creates impact in one or more of the three ways — either it helps bridge a wide demand and supply gap and/or it leads to unlocking massive efficiencies and/or it leads to extreme personalization. As we will see in our examples, one can impact more than one dimensions as well.
So, here is the framework
Let us map it to a couple of examples and see how it works.
SwitchOn is a startup (part of our portfolio at pi), which does industrial IoT solutions with cutting edge IoT and AI on the edge. One of their solutions is a sophisticated computer vision set up which identifies products coming out of a production line (at a very high speed) and rejects the defective ones. Sounds simple — however once you factor in the kind of defects you need to detect and at the rate at which you need to operate this (100s of product items per min), it puts an enormous load on humans. Most companies end up doing a 1-in-N testing, which means taking out one out of the many items coming out of the line and testing the item out. However, 1-in-N testing leads to poor coverage and defective products end up reaching the market. SwitchOn’s system can work at the speed and look at every single item. I should mention that they have also designed their models in a very data optimal way and hence can get operational with minimal training.
Now if we map the framework on to this. On the Human Capability side, they are clearly doing what humans can’t (more from the speed angle) and hence their solution transcends the human ability. On the impact side, they clearly unlock massive efficiencies.
And hence their framework diagram would look something like this.
Let us take another example from another portfolio company called Wysa. They are a mental health chat bot. One can download their app and have conversations with the therapy bot about a range of mental health issues right from feeling stressed to depression etc. Since it is an app at your fingertips, one can now get high quality therapy whenever you want it.
So, on the human side, in this case, they are scaling the human ability. Humans are better than the bot, however a human may not be available whenever the patient needs the therapist. That too for free! On the impact side, they are bridging a massive supply and demand gap and also do extreme personalization to make it relevant to each individual.
Mapping it to the framework….
Hope the above two examples explain how to apply the framework to the problem you are solving. In our mind, applying the framework brings additional clarity to an entrepreneur on how to position their startups, fine tune their product marketing and product roadmaps and GTM strategies among other things.
We would love to hear your thoughts on the framework and what you derived out of it. Drop in a note email@example.com