Bootstrap First, Raise Money Later and Build a World-Class AI Startup from India: Raghu Ravinutala, CEO of Yellow.ai (Part 4)

Sramana Mitra: How developed are your system-integrated channels? The kinds of things that you’re talking of doing, this is the kind of thing system integrators do well, right? You have talked about customer support use cases. You talked about IT support use cases and HR support. If you went into an enterprise, you could do millions of dollars of business in each enterprise. Are you penetrated into that channel?

Raghu Ravinutala: Yes, I talked about Roche Pharma. That’s a joint win with Accenture. It’s deployed by Accenture. We have deals with Infosys. India’s largest government implementation, which is for India’s income tax, runs automated servicing for their customers and is completely handled and deployed by Infosys on top of our platform.

TechMahindra exclusively chose us for their customer’s conversational AI implementation. We are well-entrenched in the SI-based ecosystem. We also have the ability to build a good part of that within the company as well. We have a reasonably-sized onboarding and professional services team. Some of our premium customers handle the professional services integration.

Sramana Mitra: Talk a bit about the multi-lingual capability. Technically, how complicated is it to introduce a new language?

Raghu Ravinutala: The important part of the multi-lingual capability is the architecture of the model that you use from day one. There could be 50 different models that you’re potentially using, but the complexity is that a single change in the core training is somebody needs to make those changes across the 50 of them.

The core part of how we’ve designed that multi-lingual system is having a single model across several different languages which are retained based on differences and similarities. That enables a particular customer to have a set of conversations trained in one language and have it available in different languages. This is the most complex part of handling multi-lingual conversations. At the same time, we need to make them scalable. For us to add another language, it’s prebaked. It’s the same model. There is no R&D required.

Sramana Mitra: When you started this, did you use existing technology from elsewhere? It sounds like an enormous amount of work. You must have used components from elsewhere.

Raghu Ravinutala: Our first deployment was based on a Microsoft stack. This is something that I believe as well. Building a product is like building a swimming manual. You can’t build a swimming manual without swimming yourself. For the first few implementations, we were using outside technology. This helped us figure out what needed to be built that can help us differentiate in the space.

Sramana Mitra: You moved out of the Microsoft stack?

Raghu Ravinutala: We were there only for a few months. Right now, we don’t use any company’s technology stack. We are just hosted on different platforms. We don’t use any company’s stack.

Sramana Mitra: The speech-to-text is also your own?

Raghu Ravinutala: Yes, some parts of it are our own. We also use a lot of open-source models for that. We use some of the text-to-speech which we believe is reasonably commoditized. We use some components of that from different providers as well. The core NLP is 100% our own.

Sramana Mitra: Interesting. What use cases are you seeing the maximum adoption of?

Raghu Ravinutala: 60% of where we see the demand is on customer support. COVID has led to a lot of explosion of marketing and commerce use cases, especially in emerging markets. People want to buy things on WhatsApp. Somebody can select their shoes and buy Adidas on WhatsApp. We have seen an explosion of commerce use cases.

Sephora is another large customer where it is automation plus live sales experience where the consumers use WhatsApp and talk about their needs. It’s just like what someone experiences in a store. There is always an assistant that can jump in. We have seen commerce and marketing-related use cases explode. We are also seeing HR and IT automation use cases. This is especially in super enterprise companies where they have 100,000 employees. This is exclusively for companies that are across geographies with 100,000 to 200,000 employees.

This segment is part 4 in the series : Bootstrap First, Raise Money Later and Build a World-Class AI Startup from India: Raghu Ravinutala, CEO of Yellow.ai
1 2 3 4 5