Author transcript
Sean Nolan: Hi Shin, thank you very much for joining me on CX Today. So for those who aren't aware, can you introduce Mindvalley and tell us about your role there?
Shin: Sure. Mindvalley is a personal growth and transformation platform, and we have over 20 million members across 195 countries. All members are learning from some of the world's leading teachers and authors through structured programs, which we call Quests, covering everything from mindset, health, relationships, and career.
For me personally, I run the product operations and customer support teams, and they both sit within the product division. So it's kind of a strange combination, some would say, but it's not an accident. It's quite deliberate, because I look after the support team and at the same time I also own the systems behind it, like automations, the help desk, and data. So for us, support isn't just a function. We have a direct line into what the members are experiencing.
Members reach out at every stage, whether that's getting access, billing, finding their way through a program, or events. Our job is to resolve that for them quickly and then to understand why they needed to ask in the first place. Most of the time the answer isn't just a better support reply, but it's a fix somewhere else in the business, such as the product, checkout, content, or the way something was communicated. So a big part of my role is getting that signal to whoever owns the fix, with the data to back it up. Because we sit inside product, a lot of it we can act on directly.
Sean Nolan: Yeah, definitely. Thank you. I'm looking forward to diving into that a bit more throughout this discussion. Can you tell us, first of all, how CX operations typically work throughout the learning journey, and how customer conversations can help inform the work of your team?
Shin: Yeah, so I think everything starts with visibility into why customers are contacting us. Without that, we would just be answering the same question over and over. So we invested a lot of time into understanding the data side first. What we have today is that AI categorizes every conversation the moment a customer reaches out. We've moved away from a human agent manually tagging those tickets, which before was really time consuming, inconsistent, and prone to error. The data is only as useful as the tagging, so now it's accurate from the get-go.
Then I work very closely with Crescendo, who is our BPO partner, on the analysis, going through the tickets and transcripts to break down what is actually driving these contacts, and separately what is driving this [audio unclear]. Those are the two things that we focus on primarily at this point, because those are two very different questions. The thing that people contact you most about is mostly different from what frustrates them the most, and we need to see both.
One example I could share is that we spotted a pattern where members who had just bought programs were having trouble getting into them. Some of them. That's possibly the worst moment of friction, where someone is excited, they just paid, and then they hit a wall. The data pointed us to a gap in our checkout flow, so we fixed the flow itself. That's the loop that I care about, where conversations become patterns, and then the pattern becomes a fix. Then eventually we drive down the amount of members who need to write in, from the root.
We're also working on connecting what we know about a member, what they own, their history with us, so that the moment they contact us, every conversation starts informed, rather than from scratch.
Sean Nolan: Yeah, definitely. I can see the value in that change, and it seems like a much more seamless workflow for your team. You mentioned there shifting from human agents to AI. In your experience, what types of customer issues suit AI, and which ones require that human touch a little bit more and are best left to human agents?
Shin: Yeah, on a high level, the more well defined everyday questions can be handled by AI, such as "how do I access the program that I've bought," "where do I find it," "where's my receipt," "how does the app work." Our current monthly AI containment rate has been running at about 55 to 60% since July. So that is a meaningful share of volume that never needs to reach a person.
Where humans come in are the edge cases, like anything that needs a judgment call, more delicate situations, and anything where someone has to go through our internal system and act on an account. But I think another angle to look at this is sequencing, rather than AI versus human. Where AI can also take a lot of the load off the human is the gathering of information. AI does the intake and gets us the details that the person is going to need. Then when the person steps into the ticket, they already have all the details in front of them.
That handoff is where we put a lot of design effort, because the customer should be able to reach a person when they need one. It should never get stuck in a loop, being bounced between bot and team.
Sean Nolan: Yeah, thank you. I see the value in that. As you say, collecting this information and then having that handoff to the human agent enables them to make a smarter decision in those edge cases and use their judgment more effectively, because AI has collected this information and they can see the full context. I really agree. Another thing about the role of AI is where it could be used in other parts of the customer journey, for example onboarding, consistent engagement, and retention. What's your perspective on that? Where can AI add value in those cases, do you think?
Shin: Yeah, thanks for bringing that up. I think that's where the bigger opportunity is, or one of the bigger opportunities. But it starts with a shift in mindset. Support isn't only there to resolve issues, it is the front line. Every conversation is a chance to save a member who's about to leave, or to teach someone how to get more out of a program through education, or point them to what they need next.
The same conversations are also some of the most honest signals you can get of where members are getting stuck. A stuck member is essentially, for the most part, a member that potentially could leave the platform. Our analysis with Crescendo has helped us see why members cancel and where they need help, so we can act earlier instead of finding out at the exit.
The next step is reaching members before they need to ask for help. For example, someone who has bought a program but hasn't started it yet. That's why we have also discussed using AI for onboarding and retention. But the condition I put on it is trust, because someone who is coming to you for help shouldn't walk away feeling like they are being sold to. The line between helping and upselling has to be drawn on purpose, and I'd rather move slowly and keep that trust than move fast and close.
Sean Nolan: Yeah, definitely. As you say, when someone goes in through a customer service issue and that's resolved in a really effective way, where they feel like they are being listened to and the experience is really consistent and smooth for them, I think that's a really strong sign for continued engagement and retention. So as you say, that front line is really important for a consistent customer journey and keeping that going for as long as possible. Thank you.
I wanted to take a step back and get your perspective and your advice for other CX leaders who may be in a similar process to you. What do you think is a common mistake that CX leaders should avoid when it comes to adopting new technologies?
Shin: Yeah, I think one is treating AI as something you deploy rather than something you operate. Launching is the easy part, and what matters is everything after. There are three things that make that concrete for us.
First is knowledge. Things move really fast, especially in tech companies. With us, there are new programs, new pricing, new features, and the bot is only as good as what it knows. So a single source of truth that's kept current is the priority. Otherwise, the AI is confidently answering something from last quarter's reality. Anything it doesn't have context for gets tagged, and we work with Crescendo to keep its knowledge maintained. That's an ongoing discipline that we have in place.
The second is member context. If the AI cannot see what someone owns, or doesn't have information about the user who they're speaking to, it can answer a question but it cannot really resolve the ones that matter, like access.
And the third and last one is which number you trust. Containment tells you what the bot didn't hand off, but it doesn't tell you whether the member left happy. That's why we read it alongside the drivers of dissatisfaction, so the CSAT score, the DSAT score, and the drivers of each. These are the things that we go deep into in the transcripts, and we look at what is actually happening on the ground. Closing a ticket and solving a problem aren't really the same thing, and you only find out which one happened by going deep into the conversation.
Sean Nolan: Yeah, absolutely. That's really good advice. As you say, having that really solid foundation and consistently giving that information to the AI bot enables it to do the high quality work that you want it to do. It's not a launch-and-forget type system. Finally, I want to ask a future-looking question. Looking ahead, how do you think AI will change the role of CX teams in building [question cut off in source]
Shin: I think the job shifts a lot from answering tickets to running the system behind them. That means setting up how things work, checking the quality, deciding what the AI should and shouldn't handle, and sharing what members tell us, the voice of the customer, with the rest of the business. We have also built the team's career paths around that, so our internal specialists grow into process optimization, rather than ticket resolution or ticket volume. It's a different kind of skill set, and you have to invest in it on purpose. It doesn't happen by itself when the AI arrives.
For relationships, I think the effect is twofold, in the relationship with the customer. When a person is handling a conversation, it's because that conversation needed a person, so it gets real attention. And the product itself gets better because we are listening at scale with AI. For the platform, that's what the relationship is: members tell you something, and then the next version reflects that change. So support stops being a cost center and becomes a source of intelligence for the whole company. That's how I see AI changing it.
Sean Nolan: Yeah, definitely. Thank you so much. I think it's really interesting, and we'll be keeping a close eye on how AI continues to move beyond resolution and actually manage the whole system, helping human agents along the way. Thank you so much for your time today, Shin. It's been great speaking with you.
Shin: Thank you. Thank you too, Sean.
