Latest / The Founder's Journey with Fexingo: First-Time Entrepreneurs, Failures, and Lessons Learned / How Rejected Chatbot Scripts Built a 55M Customer Platform
Transcript
- Lucas: So there's this story I've been wanting to dig into — it's about a founder who built a 55-million-dollar customer service platform starting from a stack of rejected chatbot scripts. Luna: Rejected by who? Users? Investors? Lucas: By banks, actually. The founder, her name is Misha — co-founder of a company called Ada Support — she was pitching a script-based chatbot to financial institutions back in 2018. And the feedback she kept hearing was, 'This feels robotic. We can't put this in front of our customers.' Luna: Ouch. That's the exact problem they were trying to solve, but it's also the thing that got them rejected. Lucas: Exactly. And look, I know that if you've been listening to this show, you know we love a good rejection turned success story. But what's interesting here is not just that she persisted — it's what she changed. She didn't abandon the chatbot idea. She abandoned the scripted approach. Luna: So she went from scripted to... what? Generative? Machine learning? Lucas: Natural language processing, specifically a model that learns from past conversations rather than following a decision tree. She essentially rebuilt the product to understand intent, not just keywords. And the same banks that rejected the scripted version became her first paying customers within a year. Luna: That is such a great example of listening to the 'why' behind the rejection. It wasn't 'we don't need a chatbot' — it was 'your chatbot feels fake.' So she made it feel real. Lucas: Right. And speaking of things that feel real — you know, this show exists because listeners like you support it. We keep it ad-free, no sponsors, no interruptions. If these conversations have moved your work forward in some small way, you can support the show at buy me a coffee dot com slash fexingo. And that's f-e-x-i-n-g-o. No pressure, just a sincere thank you to those who already do. Luna: Yeah, it really helps keep the lights on and lets us chase stories like this one without a corporate agenda. Lucas: So back to Misha. After she rebuilt the product, she didn't go broad. She focused exclusively on fintech — credit unions, digital banks, payment apps. That niche focus meant she could build really specific models for financial language — things like 'dispute a charge' or 'check my balance' — which generic chatbots struggled with. Luna: And that focus also made her sales pitch tighter. She could say, 'We understand your industry's vocabulary.' Lucas: Exactly. And it worked. By early 2020, Ada had signed 20 financial institutions. Then the pandemic hit, and customer service volume exploded. Their platform was processing 2 million conversations a month by the end of that year. Luna: Talk about timing. But I'm curious — how did she fund the rebuild? That NLP pivot sounds expensive. Lucas: It was. She raised a seed round of 1.2 million from a mix of angel investors and a fintech-focused VC. The pitch was basically, 'We were rejected by banks for being too robotic, so we're building a system that learns from real conversations.' And investors liked that she had actual rejection data to inform the pivot. Luna: So the rejection itself became part of the story that sold investors. That's clever. Lucas: It really is. And the numbers backed it up. After the pivot, customer retention went from 60 percent to 92 percent within six months. Because the chatbot actually understood people. Luna: What was the key metric that convinced the banks to give it a second try? Lucas: Resolution rate. The scripted version resolved maybe 20 percent of inquiries without human handoff. The NLP version hit 45 percent within three months. That meant real cost savings for the banks. One credit union estimated they saved 300,000 dollars in call center costs in the first year alone. Luna: That's the kind of number that gets a CFO's attention. Lucas: Absolutely. And here's the thing — Misha didn't stop there. She kept iterating based on what the chatbot got wrong. Every unresolved conversation was fed back into the model. So the system got better the more it was used. Luna: So the rejection didn't just shape the initial product — it became part of the product's DNA. Continuous learning from failure. Lucas: That's a great way to put it. And I think that's the real lesson for founders: when you get rejected, don't just ask 'should I give up?' Ask 'what specifically is wrong, and can I fix that without abandoning the core idea?' Luna: It's a subtle distinction but a powerful one. Most people either quit or ignore the feedback. She did neither. Lucas: Right. And the result is a company that's now valued at over 55 million, with customers in 15 countries. All because a few bankers said 'this feels robotic' and Misha actually listened. Luna: I love that. So if you're a founder facing rejection today, maybe the question isn't 'is my idea bad?' Maybe it's 'is my execution good enough?' Lucas: Exactly. That's the thought I want to leave you with. Thanks for listening, and we'll catch you next time on The Founder's Journey.