548th 1Mby1M Entrepreneurship Podcast with Lloyed Lobo, Co-Founder of Boast.ai (Part 3)

Sramana Mitra: How about getting your first product out? Were you bootstrapping to get to the first product?

Lloyed Lobo: We bootstrapped to about $10 million ARR. We just raised a round of funding around Thanksgiving last year.

Sramana Mitra: Let’s talk through the early part. You bootstrapped and you had validation because of your consulting work. How much of that consulting pool became customers of the product?

Lloyed Lobo: I have this thesis from failed companies. Customers want an outcome; they don’t want software. I genuinely believe that. They’re buying software for outcomes. They don’t want fancy dashboards or what not. Effectively what’s happening is, they’re giving us their data; we’re giving them a check. We continue to build more automation. The other key learning is, you can’t build an AI company on day one. You need to get really good at data collection. You need to collects lots and lots of data.

Sramana Mitra: That was my next question. How did you get the data to train your model?

Lloyed Lobo: The first thing is, you need to get the data really well. A lot of times, that data collection may be manual. Then you get into workflow automation. You’re automating the collection of that data still informed by humans. Then you get to a point where you have sufficient data where machine learning starts to kick in.

I’ll give you an example from Automatically. It was AI for customer service. This was 2013. We did all our customer development with large enterprises. Everyone was like, “If we had a magic wand, we want something that would respond to our customers automatically. Our customer support team is not proactive. They’re inundated.” When we tried to integrate with Oracle and Salesforce, you got to go through a one-year security review.

We panicked and ran to Zendesk. Zendesk was all open. We built the app there. We were the first chatbot that Zendesk had. When we launched on Zendesk, we had thousands of people signing up. I suddenly started getting messages saying, “Make this thing stop.” I was noticing that the chatbots were responding with gibberish. We then made it editor-approved. Still the feedback was, “I still have to edit a lot.”

Zendesk had a lot of SMB customers. They didn’t have historical datasets. We couldn’t accurately do natural language generation. We shuttered that. If I knew then what I know today, I would have gone to all those customers and asked what their most common questions are, do an if-else lookup table. 

Sramana Mitra: It’s a lot easier actually. You just do a rule-based engine. 

Lloyed Lobo: Exactly. When you’re a small startup, do things that don’t scale. We had so many people sign up. It was just a mess. Boast was done very meticulously. Do the work manually. Figure out the manual touchpoints. How do you get the data and then AI? When you’re a bootstrapped company, you don’t think of technology. You think in terms of how to keep your lights on. You have to keep the customer by any means possible. When you’re venture-backed, your mindset immediately goes towards experimentation. It doesn’t have to be that way. Early on, you got to do things that don’t scale.

This segment is part 3 in the series : 548th 1Mby1M Entrepreneurship Podcast with Lloyed Lobo, Co-Founder of Boast.ai
1 2 3 4 5 6