1Mby1M Virtual Accelerator AI Investor Forum: With Sailesh Ramakrishnan, Managing Partner at Rocketship.vc (Part 5)

Sramana Mitra: Interesting. Let’s just follow through on the same case study of Dili. What is the adoption and what is the experience of the company in gaining adoption around this particular use case?

Sailesh Ramakrishnan: Dili has been moderately successful in gaining adoption. The challenge is in making sure to come to market not as a company that is going to solve all of the diligence problem, but more so as acting as a co-pilot or an assistant to the humans in the loop. I think that strategy is one of the key things that we have seen in differentiating companies that are able to get adoption versus ones that are not.

There are many companies that seem to oversell the promise of AI by saying, “Hey, we will replace all of these people.” As I mentioned, that might have been true in some of the lower technology tasks, like first level customer support, but for the most significant AI use cases, right now, the best solution still happens to be a hybrid approach.

So the first thing to note is that Dily approached the problem, not to replace the human analysts or the human legal expert, but as a co-pilot or an assistant to them.

The second challenge was in the initial stages, it’s very hard to become as good or better than somebody who’s been doing this for 15-20 years, right? They have a nose for where the issues and the problems in a given company might be. I don’t know if an automated system can learn as much. So, architecting the system in such a way that the humans can provide that kind of large experiential domain knowledge, which can then guide the system has also been helpful in gaining adoption.

But as of today, they are still being used more as a value addition rather than as the standard. That, I think, is still in the future, but the adoption has been more towards the direction of adding value because, as you might imagine, these folks are very busy and anything in terms of software that doesn’t add value will only be a distraction and will get dropped. At least, Dili has crossed that chasm where they’re able to show value. Now, the question of how much value they will deliver and how integral they become a part of the future of diligence process is the mountain they’re currently climbing.

Sramana Mitra: What is the attitude of the humans who were doing this due diligence? Let’s say they were taking a year to do this due diligence, and Dily’s value proposition is to maybe cut that short by a lot. Are they willing to train Dily to automate some of their manual functions or are they defensive?

Sailesh Ramakrishnan: It depends on the customer and their mindset. We are investors in Dili, but are not part of their day-to-day conversations with customers. From what I’ve heard from them, people are actually very open, because poring through boxes and boxes of paperwork is tedious.

There has been a lot of interest from a customer perspective, but the biggest concern that they have expressed is whether the answers are going to be accurate or truthful and dependable. It is not just about hallucination and about the system making stuff up, but also whether reasoning is right, whether the system is not putting two uncorrelated things together that a human would never do. For a system that doesn’t have deep understanding, it might. So, Dili has been solving that from day one with an approach of explainability.

Especially in these situations where you’re consuming proprietary knowledge and set up these large sets of deal documents, there are ways in which you can point to a source document and say, “This is where I found this number, or this is where I discovered this graph, or these are the three facts that I chained together to come to this conclusion.” That is one of the key aspects of what Dili is doing in order to gain confidence from human users. It can at least to a first order approximation and explain the answers.

It’s not perfect yet and there is a lot of work to be done, but I would say most users so far have fallen into two categories. One is cautious optimism that this is really going to work and save time for them, versus the second camp that says, “Well, this is interesting, but I don’t know whether it’s actually going to save me time or, is it going to take me down the rabbit holes that I prefer not to go down.” At the high level for a startup, both camps are still interested in at least trying it out.

This segment is part 5 in the series : 1Mby1M Virtual Accelerator AI Investor Forum: With Sailesh Ramakrishnan, Managing Partner at Rocketship.vc
1 2 3 4 5 6