1Mby1M Virtual Accelerator AI Investor Forum: With Naganand Doraswamy, Managing Partner and Founder at Ideaspring Capital (Part 2)

Sramana Mitra: Let’s double click down on your current portfolio. Let’s discuss the AI company that you have invested in and let’s also discuss the five to seven AI SaaS companies that are bringing in AI co-pilots. So, let’s do some case studies and understand these trends a little bit from the ground so that you can talk with real experience.

Naganand Doraswamy: We have, along with Accel Partners India, invested in a company called Mason. They were enabling e-commerce companies to send very personalized marketing messages to the consumers to get them to come to the website. When they come to the website, they dynamically determine when to give discounts and when to offer a sale. They were building a product that was basically an online shop floor manager.

In a lot of e-commerce companies, the owners must run their own website, but they don’t have the capability. This company basically provides a platform for them to run whatever is required for the shop and for the customers who are coming online to shop, including marketing.

Right now, with the gen AI capability for understanding user patterns, you can quickly create user specific marketing messages dynamically and increase the probability of the e-commerce sites getting customers on board and converting them. They’re including that in the whole process.

Another company called Karomi’s Manage Artworks works with various large FMCG or Pharma enterprises to manage all their artwork and label management. In this company, we are looking at how can you help these companies design their own labels, design their own marketing material, or the boxes. So, you must learn all the various artworks they have had in their company and use that to intelligently and creatively suggest what could be a potential packaging material or labeling information for that company. That’s another use case we are looking at.

A third company is a company called B2Brain, which basically uses AI techniques for sales. When a sales guy wants to talk to a company, it matches what your customer products are and what your product is, and then it gives real-time indicators and meaningful information about their prospects to have a better sales conversion. Again, generating the text for this is another use case.

These are all companies in our portfolio that we are looking at how best to imbibe GenAI capabilities. Most of the SaaS startups must look at this. So that’s one set of things we are doing.

Sramana Mitra: So, on that topic, it is not so easy for SaaS startups to incorporate AI as a retrofit, right? Because the stats are different.

If you are trying to build a product that is really AI native, the stack to build that on is not necessarily the same stack that you had already built your previous capabilities on. People are having real trouble with that retrofitting from what I’m hearing out there.

That’s one issue. An agent architecture is a possibility where you have your core SaaS product, and then there are these enhancers with agents. That’s like a co-pilot. It is one way to kind of circumvent that problem. If it’s possible within the workflow, that’s a possibility.

In terms of companies that are coming in AI native built in the AI era,

and are competing with the existing SaaS companies in the same domain,

that’s tricky. Maybe these SaaS companies have $10-$20 million of revenues. But if a fresh AI company comes in and disrupts that market, that’s a tricky situation. Right?

Naganand Doraswamy: Even if the AI company were to come in it many times largely depends on the interfaces you have had with the enterprise and the enterprise ecosystem. One part is generated by AI that you can look at and then build and get answers to the questions. Then, there’s the whole situation of how the user interacts with that inside the enterprise and get meaningful outcomes.

For example, there’s another company called Mihup, which is building contact center software. When a customer speaks, the contact center software identifies any taboo words because it is translating voice into text and catching the taboo words. So, you can say you are AI first and build that LLM-based code. But then, there are so many other systems in the contact center that you have to interface with.

So, part of AI is a code, but then interfacing with all the other software in the enterprise is another part. That’s where it becomes harder. If it is just a pure independent system where you don’t have to interface and interact with any other enterprise system, then you’re OK. However, there’s a lot of time and energy that is spent in interfacing with so much information an enterprise has.

AI can solve one part of it, but there’re a lot of other things that still need to happen, and you have to be very careful about it. If your software or product is such that it’s just for AI, somebody can come in and build something new and get rid of you. That’s a dangerous situation to be in. In selling to enterprises, you also have to understand which other enterprise software you’re interacting with.

Sramana Mitra: Absolutely. So, vertical AI means the entire vertical workflow, APIs, integrations, everything, right? In any kind of vertical AI application, AI is only part of the puzzle. All the rest of it has to work out.

Naganand Doraswamy: That’s why we have to build the mode, right? In our company, we have taken 3-4 years to build all these interfaces. We had a voice LLM. You can replace this voice LLM with something new because we are on a proprietary voice LLM. You can say there’s an industry standard voice LLM that you can plug in. But that’s just not it, right? You just have to build the entire application and that’s where I think you can build your mode.

This segment is part 2 in the series : 1Mby1M Virtual Accelerator AI Investor Forum: With Naganand Doraswamy, Managing Partner and Founder at Ideaspring Capital
1 2 3 4