Inspired with Alexa von Tobel
00:12:45 · May Habib
[00:12:45] S0: So our our first company was focused on machine translation. And it's so funny because we we do a lot of translation now. Right? Writer is an agent platform for front office, and a lot of what people do when they go to market is, you know, modified by channel, modified by language. But, you know, back then, I mean, we were building statistical machine translation models. And even in some of our first, you know, more like, you know, just deeper machine learning models that we built on translation, we had different models for, like, one language direction than another language direction. Right? Like, this was, like, very, very primitive early days. And the big inflection point for us was the models we were building being able to generate English language. And that, you know, I remember the first all hands. We were across the street from where we are in San Francisco now. The first all hands where we presented a haiku on a slide that one of our models had had written. And it was like, what the fuck is this? And and that's a very different business, right, in terms of, you know, translation to, we called it, source language, right, creation. And so, in an ideal world, we would have sold the first company and that would have been like a very clean exit, and then we would have started our first, our second company. But it really didn't happen that way. And me and Wassim just had such high conviction that we needed to run this transformer thing. You know, these were encoder decoders. And we were solving, like, just a completely different problem. And, like, you know, we didn't know what the product would look like. We didn't know who the buyers would be. You know, we were even, like, we would take long walks and try to think about and imagine we were selling near the enterprise in our first company. Like, should this be a consumer company? Should this be a prosumer company? And after ChatGPT I thought, oh fuck. Like for a few months it was like, maybe we should have done consumer. Now I am very happy, right, that we are pure play and just so focused and so successful on delivering you know, really unmatched value to the enterprise. But that gives you an idea of just like, you know, how differently Mean Wasine thought about this new technology and how different it felt from what we were doing before.
00:15:15 · May Habib
And, you know, now I am very excited, right, that we get to serve so much of our original founding mission of Qordoba around, you know, really making language a non issue for people and for businesses. But it did feel like we were abandoning something that we had, A, worked really hard for and that we had raised money on, promised a lot of people things, and also made a lot of promises to customers. But you know, really over the course of about an eighteen month period, right, we've we've sunset all of that revenue and really built Writer from from the ground up. It definitely meant also from a people perspective, right, we rebuilt the team and started something new. And I am really grateful that we made the jump when we did because I I think, you know, now we are the only pure play full stack generative AI company in the enterprise that's not a hyperscaler or like a massively funded research lab. And so much of that is because, you know, we were able to spend a couple years before the hype cycle really focused and really the only player in town in terms of what we were able to offer the enterprise. And so we've been able to, you know, just bring, like, just build customer reference after customer reference in highly regulated industries doing, you know, pretty complex agentic implementations now. But it's a five year history of being able to serve the enterprise with AI. And it was a tough transition. It was definitely a tough first year or two in terms of like just the eating glass pane of transitioning customers and humans, right, out of the business. But you know, by early twenty one, right, we had really found our way as Writer and raised that first A, that first kind of big round in the fall of twenty twenty one, just as generative AI was becoming a thing. So yeah, I'm grateful for the whole journey, but it's been ten years in natural language processing and machine learning to get here. And now I get to talk about our research road map when I meet customers and potential partners. And, you know, there's just so much depth having really been a team that's been in so many innings, right, of this revolution and having the depth of being the only research lab that, you know, for years has focused on problems that real companies have. And, yeah, that started a decade ago. I love that. I'm
00:17:45 · Alexa von Tobel
[00:17:45] S1: on it in plain English for everybody out there, and obviously, Writer is involved. What is... What problem is Writer solving? And give us a sense of who your customers are, and what does that look like?
00:17:55 · May Habib
[00:17:55] S0: Yeah. So Writer wins in AI when there are two or three requirements that a company has. When a use case requires high precision, high quality, high consistency and reliability. Right? People choose Writer. A lot of these use cases tend to be front office, tend to be sales and marketing. Right? If you are now really semi automating the way a wealth advisor communicates to a client after they have a sit down meeting, right? Everybody records calls now. We've got a great CRM. It's captured everything. But really being able to understand what does Alexa need? What are Alexa's financial goals? How do I bring the best of what I know about Alexa with really what my organized knows about Alexa. Right? A a lot of what we do is make explicit what the organization knows implicitly, but doesn't apply consistently. And the thing with other AI tools is you can go ask Copilot or ChatGPT, you know, the same question asked 50 different ways by 50 different people, and you will get, you know, 48 or 49 slightly different answers. And that's just not... That doesn't pass the bar, right, in these regulated industries. And so in healthcare, in financial services, in CPG, there are millions of use cases like this, right? Micro use cases. But they all really ladder up to better high end customer support, customer service. They all ladder up into making relationship selling much more productive, much more consistent, better user outcomes, better company outcomes. And so much of the front office is really calcified into these systems of record that don't talk to each other and produce, like, pretty rigid business rules. But so much of this type of selling is dependent on unstructured information, unstructured know how. Right? And the beauty of AgenTik AI is it's really good at the unstructured. And so, you know, we really shine in these kinds of use cases. Most of our customers are in highly regulated industries. But we've got a lot of technology customers too. Microsoft, Salesforce, Nvidia, HubSpot, Qualcomm. Lots of folks who build their own AI use Writer in front office because a much more scalable solution. Because if you are now going to automate something, if you're gonna gentify something, you've gotta be pretty certain that the guardrails are there, right, for you to get the kind of outcome that you need. One of
00:20:58 · Alexa von Tobel, May Habib
[00:20:58] S1: the things I wanna give you a compliment on, which is Writer has built an incredible reputation on responsible autonomy, creating agents that act independently but remain really aligned to a company's values and compliance. I was Chief Digital Officer at Northwestern Mutual of Mass and Financial Firm, so when Yeah, you talk about [00:21:15] S0: they're also a customer. [00:21:16] S1: Yes. So when you talk about, you know, how much it matters that you can trust these autonomous agents to stay on brand, stay on brand values. Talk a little bit about how you've maintained trust at scale, because that's something... Obviously, you're ten years in, so you've had a lot of time, but that's a really hard thing to get, right, and to thread that needle, May. How have you done that?
00:21:42 · May Habib
[00:21:42] S0: Yeah. You know, the the way that we... I I kinda wanna answer your question in a meta way, because we're having, like, a philosophical Yeah. Love it. Let's do it. You know, like, a lot of the the trust we build is by not trying to do anything else other than make these customers successful. Right? If you look at the hyperscalers, you know, they're really trying to sell compute. And the, you know, labs have become full stack, but they're really still LLM API companies. You know, our ability to go in and say, we are the only full stack player that is not trying to do anything else means we can say, look, this 10 step process that you have to do this thing doesn't make sense anymore. We're not gonna just shift and lift it to AI. It really should look more like three steps. Right? And a lot of that is really questioning the assumptions around business process and org design and the calcification of legacy systems. Right? And so think of the question you just asked around brand and compliance. We can bake that in from the very first draft, but your compliance officer has to be Gabe. Right? Your brand team that may actually report to a different leader than digital, we see that all the time, right, has to be able to sit at the same. And so, you know, our technology forces a conversation because it is built in a way that agentically all the systems are talking to each other now, right? Like these silos don't exist in Writer. In natural language, you can get the guardrails to talk to the best practices, to talk to the customer data, to talk to the brand brain. Which means people need to now make decisions on how to operate more effectively and really get on the same page as to their goals. And I think the way we've built trust over time is to really never be selling anything beyond that, and to challenge the customer the way that we challenge ourselves. Sometimes our account teams are like, May, did you just kind of yell at a customer? I'm like, no, it's not yelling. I promise you they appreciate it. And the customer does come back and say, thank you for challenging me. Like, I kinda needed that kick in the pants, you know? And we do, like, that's, we we ask people to really question their assumptions. Like, what assumptions about your industry do you take for law that actually aren't? And, you know, we are looking for radical partners. Right?