MasterClass OnCall: AI voice coaching from top instructors (Full Transcript)

[00:00:00] Speaker 1: When the Gen-AI voice came on the horizon, we were like, okay, we think this is the missing piece. I'm Mundar Bapai, I'm Chief Product Officer at Masterclass. Our mission at Masterclass is to democratize access to the knowledge of the world's best. We do that with classes, with video classes, and that across the entire body of spectrum, like from the artists to sportsmen, to business leaders across the spectrum. And one of the things that we consistently heard throughout our lifetime was our users really wanted to talk to our instructors about going deeper into a topic or asking follow-up questions. And that's the part where before Gen-AI, it was not scalable, all the solutions that we came to the table with. But with Gen-AI, it actually became a reality that we could bring to our users. So we have been on that journey for some time now with our new product called OnCall, and yeah, it's been fun. So with OnCall, as I mentioned, the idea is you can take a class or you can just go directly to OnCall and you can get personalized advice from the world's best. Again, like with Masterclass core product, we transformed the best of the world from practitioners into teachers. With OnCall, we are transforming them from practitioners or teachers into coaches and advisors. So with OnCall, you can get on call with Gordon Ramsey and get all your cooking questions answered, that matter to you. If you're in the kitchen, you're cooking something and you have that one nagging problem, you can call Gordon, get on call with him. If you have a business problem, you can call Cuban and talk to Cuban about things. So yeah, it's all about enriching our users' lives by giving them moment in time advice from very trusted sources. What we heard consistently was users were like, hey, I learn a lot from the classes. Our classes, people absolutely love them, right? And they are great production quality, great knowledge, like they are the best vehicle we have at our hand to really impart the knowledge. What users wanted was I get all the knowledge, but then I have certain problems that are quote unquote, like very personal to me or unique to me. And I want to engage with someone who really understands the problem. And they're like, when I engage with an AI, I don't know if I can trust this, a generic AI. When I go and engage, or if I just Google search, I'll need to find out the information that I can really genuinely trust and applicable to my use case or my problem. Whereas if I have a business problem, and I know, let's say Cuban's AI is created with Cuban, I know that I can trust the source of the information that it is coming from. So we actually tried many vendors. The stuff that I would say 11 really shined on two main vectors. One is the quality of the voice is just like night and day better, right? And that's where given that it matters so much to bring the personality out, like the tone of the voice, the pacing, like all of that is super critical because people can easily detect. And we want to make sure like, for example, with Gordon's AI, there is a little bit of humor programmed into it. And it is able to like, the voice is really able to bring it out. So I would say quality of the voice. Second one is the latency. It's brilliant. How quick the voice gets produced in real time. And third one is just the partnership with 11 has been phenomenal. Like from every step of the way for the past 18 months, you guys have helped us out. So yeah, let's say those three. So we started off with just having a chat interface, a chat bot-like interface. And that worked, but it was missing the magic a little bit. So the next step was to create chat interface with instructors, for instructors. And same thing, it worked well. People got the value, but like it felt like the magic is missing. And when the Gen AI voice came on the horizon, we were like, okay, we think this is the missing piece. So then we built it with 11 as our key partner and it's been a phenomenal journey building the product with 11. But the pipeline is like the standard speech to text LLM and then 11 at the backend. But yeah, like the experience is as good as you pick up your phone and you have a problem and you call Cuban. We thought chat will be the part that will get most adoption. Even now, almost 90% of the conversations are voice. Even though it's like higher barrier for people than chat, voice has that magic that brings the experience, the personality of the instructor really alive. And it has worked wonders for us. We are also venturing into the land of role plays. Like one of the use cases that came up organically is we saw a lot of users engaging in role plays with a Mark Cuban or with a Chris Wass about negotiation. So we have stood up a product area where we just have a lot of curated role plays, like thousands of them. Users can engage in professional role plays, personal role plays, and then get coaching from the Mark Cuban or the Chris Wass or the others of the world. Deeply we integrated 11's team with our engineering and product teams to again, train the voice models, to literally you guys helped us with like curating the data and like cleaning up the data and giving us pointers on, these are the aspects you may want to watch for when you're thinking about what data sets to create and it has been phenomenal partnerships.