Thought Leaders in Artificial Intelligence: Francisco Webber, CEO of Cortical.io (Part 5)

Sramana Mitra: Everything that I have listened to so far is pointing to the fact that this company needs to be scaled as a Platform-as-a-Service company more than a pure solution company. 

Francisco Webber: Interestingly, investors prefer you to be focused, especially in the early round. 

Sramana Mitra: Absolutely.

Francisco Webber: It’s a bit of a problematic situation. 

Sramana Mitra: It’s not problematic. I think it’s how a Platform-as-a-Service (PaaS) company typically gets built with one type of solution built on the platform, and then later, you open it up for other use cases. It has become the way to go to market for a PaaS company. I think you have much bigger potential than within a specific vertical or specific use case.

Francisco Webber: Yes. As I said, as we tackle the problem at the fundamental level, we also have the advantage that it fits all the dependent use cases. It’s extracting specific information from a document. It was an obvious case with contract documents, but it can also be any other documentation. It can be scientific documentation or medical reports.

Currently, most of the interest comes from insurance companies and banks, and to me that makes sense. All of their products and business interactions are based on language. The product is basically a description of a statistical insurance model that they have. 

Sramana Mitra: Yes, and it’s huge piles and piles of documents and contracts. 

Francisco Webber: The same is true with very large companies. We just got a customer in Europe, Siemens. Companies of that size have 300,000 employees, so can you imagine the documentation behind the business engine with so many people. It is so large. They have millions of documents.

The problem in such a situation is that all of the documents talk about and use the language that is relevant in their business domain. It will all be about engineering and technical projects. The problem that you get with this is that a word like transducer becomes a stop word in their world because every hundred documents contain that stop word. The problem is that discriminability – dividing a part of two documents on a statistical level – is very hard.

There are even some scientists saying that it is impossible at that scale. The traditional full tech indexing works well as long as you stay below 2 to 3 million documents. Once you are in the 50 million documents space, there might be something like 20% of the documents that are not selectable by any query that you could post. The only way out of this is that you not only take into account the presence of certain words but you also take into account the patterns of meanings behind them. That is exactly how humans would differentiate the documents. 

Sramana Mitra: I understand all of these very well, and I understand its impact. Let me ask you a few slightly different kinds of questions. First, how did you build the business? You started in 2011, started to develop the technology around 2016, and got your first bank customer. What’s the trajectory? Between 2011 and 2016, how did you finance the company?

Francisco Webber: We mostly financed it with an angel investor. We also had an untypical investment path. We started with an angel investor who went on to get more fascinated by the potential, and in that process, they became an investor. The other end of the revenue scale was selling our services by using our library to solve use cases. Not all of them were production-grade. Some of the companies just wanted to do some exploration on a small scale and see how well it would work.

From there, we got a mix of 70% of the revenue from service and 30% from development licenses that we gave to AI teams. We then had three or four systems that were built for production. That allowed us to slowly grow to the size of 33 people, which is still tiny. On the other hand, it also shows that we can handle four or five customers concurrently with that team thanks to the efficiency of our approach. 

This segment is part 5 in the series : Thought Leaders in Artificial Intelligence: Francisco Webber, CEO of Cortical.io
1 2 3 4 5 6 7