The company OpenAI and Google call when a user is in a crisis: Elliot Taylor, Founder of Throughline

How did you get into those rooms, interfacing with some of the largest product businesses in the world as a relatively small team working out of New Zealand? Probably like anyone stubbornness and persistence. I kind of fell into becoming a tech founder. What exactly is through Lines product? We build infrastructure that connects a company's users to the right mental health or online harm support. You're already touching tens of millions of lives on an ongoing basis which is pretty extraordinary. There is enough of that problem for us to be able to build the biggest mental health intervention the world's ever seen. I know, thanks. It's not a wonderful time to be in the social media platform PR business to give just a few examples. At the end of 2025, the Australian government became the first national jurisdiction to institute a social media band for under-16s covering host of platforms. And that set off moves in more than 15 countries worldwide. Last month, Metta was hit with an historic $17 billion lawsuit over the addictive effects of its platform, particularly for children. And it's estimated that in the US alone, there are more than 3,000 open lawsuits pending against these platforms, many focusing on the social and psychological harms brought about, by the way they operate. But at same time, something in the order of 70% of the world's population have a social media account of some kind, or close to 95% of the world's adult population. Stats on the use by children are patchier, but the UK's regulator found that over 50% of under-13s are currently on some form of social media, and in Australia around 80% of 8 to 12 year olds were using it before the band. Reality is that social media platforms are an embedded part of our social fabric. And they're an integral part of how we communicate, create community, learn and stay in touch. Together with the AI platforms, they're also where a lot of us spend our time and share information about ourselves. And it's for this reason that they're also an important front line for mental health services. It's into this crucible that throughline plays. Throughline, which was founded by Elliott Taylor, a former social worker who specialized in youth mental health, is an ingenious platform provider helping social media platforms redirect at-risk users to mental health services in the moment. For example, it's estimated that about a million people a week talk to chat GPT about self-harm. And if they do, throughline will be alerted and immediately redirect that person to a location specific service they can speak to to get help. Scale of impact this one tiny company based out of New Zealand can have is mind-boggling to me. And this episode's all about understanding where that idea came from, what the company's evolving into, and even what product development looks like when you have people's literal lives in your hands. A warning. As you'll have picked up already, this episode discusses suicide and other forms of self-harm. We're including links in the show notes for support services. Should this bring up anything for you? Over to Elliott. So first of all, Elliott, it's so wonderful to have you on wildharts. Thanks for being here. It's a pleasure. So I wanted to have you on wildharts. In fact, I've wanted to bring you into some format to have a conversation about throughline, as you know, for a long time. And that's because what it is that you do is so inspiring. But let's start with the sort of most obvious question for our audience. What exactly is throughline's product? So we build infrastructure that connects a company's users to the right mental health or online harm support. What that started as is a global verified helpline network. So if you think, say in Australia, people will be familiar with services like Lifeline or 1-800 respect. Their services like that all around the world. And we've built a network of those services. And then that powers crisis support on, you know, online platforms like chat GPT or Google search or Pinterest or Tinder or lots of others. So that's where it began. And then where it's evolved into is building a IK navigation layer over the top of that, which is helping that understand kind of what's actually going on for a person and then connecting them to the right support. Not just helplines, but other types of support as well. Amazing. I want to get into that. But maybe let's step back to what people know throughline for at the moment. Could you walk me through like a typical product experience for someone? How does throughline appear to them and what does it look like to interact with your product? Maybe chat GPT is a good example here. This is public data about a million people a week talk to chat GPT about suicide all around the world. And so when they do, we provide the support crisis support resources like in partnership with OpenAI on chat GPT. So someone's coming to chat GPT and they're saying, you know, I just don't want to live anymore through chat GPT and an integration. We help that person be able to connect with like what the relevant helpline is for them wherever they might be. That kind of theme follows in from different products, right? You've got this kind of identification piece of like, well, you know, a person's disclosing something or they're some sort of user behavior that gives an indication of risk and then you've got a response piece. So say that's on chat GPT, but on Google search, that might be, you know, someone searching about being in a domestic violence relationship. And similarly, we work with Google to make sure we surface the right information or that might be someone on Pinterest, you know, searching content relating to eating in body image or someone that might be someone on Tinder or hinge who's experienced a bad date and needs help. If you think about like all the contexts where middle health needs or online harms crop up online, where the people that help connect those people to the right support. So let's break that down. So taking us back to the chat GPT example, does it look like you see the word suicide or some variant thereof that your AI model has picked up as a signal of a kind of behavior or an intention that you're worried about. And then there's a pop up window that says from through line that suddenly looks like through line and says, actually, do you want to speak to lifeline if you're in Australia? Or is it actually like a full integration like a US through line building their AI that chats to that person and chats them through? I'm really curious about the kind of user experience layer. Yeah. So like if we do some example like chat GPT, so we don't do the detection, but when it comes to the platforms that work with whether they're like AI or social media or search, like that is often a content moderation piece and that market is significant. You know, there's a lot of players and it's growing a lot over the last decade. The problem that these companies have had that they didn't really know how to solve was like, well, we find stuff. You know, we know that this person's in need like now, what do we do about it? Yeah. And so that's the part that through line provides like where the response piece. So in the chat GPT case, what's the same on Claude? You know, we also work with their classifiers are determining whether or not there's risk. And then we're powering that response that can look different on different platforms like in chat GPT that's in Claude, that's through a banner. You know, we can also augment model responses through like a tool call and you know, pull our resources in and show them directly in the model response. Without our communication product, that's where it kind of takes a step further right where we are holding that conversation, where there's like some sort of risk or need being presented. And as opposed to just like surfacing a resource that a person can decide whether or not they want to access, we then can have a conversation with them, manage whatever that risk might be, help them unpack that, understand their options and then connect them to different types of support. Interesting. So can you give us a bit of a sense of like you mentioned earlier that a million people on chat GPT awake I think is what you said, you know, demonstrate suicidal orientations, that's probably the wrong language. What's the kind of conversion as it were from like that, you know, they've used their detection tools to identify that a person to trigger a call to through line. What does that kind of look like? Like and how much is that hand off? Is visible to you and do you measure? Yeah. So we're measuring the contact moment, you know, across our partners, of course, these are large platforms, you know, I can't kind of share numbers specifically. But I think like the theme is that this is happening at significant volumes. Yeah. And in some ways, it should be obvious, right? Like, you know, think about the moments that you've turned to Google or to an AI to talk about something that is sensitive or something that's private or that you want to keep private. That happens a lot if you're thinking about the entire population of the world. I've been thinking about this a little bit recently. I would go as far to say that like by head count through line has more kind of mental health touch points with people in crisis than like anyone in the world. Wow. But in terms of our like company size to the demand of like people that we touch, which is pretty cool. Yeah. That's an extraordinary thing to wake up every day knowing about what you do. We talk at Blackbird as you know, a lot about founders doing their life's work. I know that you came to building this platform through an unusual path to founderhood. Can you share with the audience a little bit about how this came to be a company that you were building? Yeah. I didn't come through like the tech path. You know, I actually ripped in community services and that type of thing for about 15 years. And it was in my early 20s that I had a really close friend who went through like quite serious, like mental health, self-harm, and suicide struggles. And I fell into the support role with her with some friends and I found that incredibly grounding. I felt like I was doing something that really mattered. And then I spent about the next 15 years doing that with people, you know, through organisations, like, you know, also around community homes and, you know, would have supporting people in the community and, you know, having kids kicked out of home, like, sleeping on my couch and being a self-house for people in domestic violence situations and that type of thing. And so I came to it as this experience of a person who has spent a lot of time sitting with people on the stoop in a time of need and trying to help them get through that moment. I've also been kind of – it was always intrigued by technology and product and, you know, I probably came more from the kind of a marketing communications lens and I did a lot of that and I kind of fell into it. I started kind of experimenting with different kind of ideas and some things worked a little bit and some didn't and over time, like, I guess, fell into becoming a tech founder and I've picked up everything else I needed to kind of know along the way. But I think for me that's been quite beneficial because with the type of work we do, like, you can only really make an impact on bugs like this audience if you really understand their world. Like, I don't think you could be, you know, just a person that's, like, had free product roles or, you know, was a software engineer and then go like, "Oh, I've been, like, started into health startup." You know, like, you, like, having some sort of understanding of that world, whether that's from lived experience or working with those people was like, really, really important. Yeah, let's get into that because I mean, I think one of the things I've always found so fascinating about your business and the team behind it is how you do product, actually, because of the very nature of the sensitive environments you're operating in, right? Like, you know, if I think about most product founders, there's a host of tools available to them amongst which is like, you know, simple AB testing. Like, okay, well, we don't know which of these feature sets is going to work best. Let's throw out like four or five of them and say what happens. And, you know, I imagine that some of those kinds of product approaches are just not appropriate for the kind of product that you're building because you're also dealing in, as you alluded to earlier, like some of the greatest number of touch points with people in crisis in the world. So I'd love to get a little bit under the hood on what it looks like to have taken your clinical experience and become a product leader with that in mind. How do you think that opens up opportunities to build product differently than someone who didn't have that background would? But also, where does it have challenges and you've hit up against constraints because of this environment you're operating in? Good question. I think the challenge when you're you're building and like sensitive spaces like this is that like you've got this kind of non-negotiable safety layer. You can't do anything that might create some sort of risk or has a risk of harming someone like, you know, the people that come to our products oftentimes thinking about ending their life that day. And so that's like a very different calculus to I'm going to kind of like run this experiment to see if I can list conversion rates, you know, on the checkout page. Yeah, you know, like it's just like it's very different. So you have this foundational safety layer, which you can't compromise on. But I would also say that like traditionally, when people have built product solutions like in the space, they haven't often used some of that toolkit of kind of like, you know, AB testing or, you know, experience experimentation or like more iterative development. And then what I've seen is that you then kind of like create a solution that is okay, but broadly speaking, kind of poor. And this is the like significant opportunity ahead of you or in front of you. And so for us, like we have that safety layer and we don't do any thing that might have a harm profile to it. Like do no harm is like your foundation. And we have clinical team members, you know, who are across all those things. But that doesn't mean that we don't experiment. It doesn't mean that we don't look at ways we can do things better. Like maybe to give you a particular example, like we run a network of about 1500 crisis services in almost 180 countries. And in one of our product experiences where we build and maintain the UI, the actual user experience, we show whether or not this service is open right now. But contact methods for services can actually change. Like you could have a service that is available over phone, text and online chat. And those service hours might be different in their opening times. We were just showing them all as being like the same, you know, the services open or it's not. So what we ended up doing is we ended up sourcing from services, what their actual opening times are, per specific contact method. And then we ran an experiment where we changed the UI to basically see whether or not we could reduce contacts to services that were actually closed where the user know they were closed. And so this was using like status indicators like we would, you know, throw up a little green dot like on a contact method for whether or not it's open and gray one out for whether it's closed. In doing that, we actually reduced contacts to closed services, I think by something like 30%. So these are people that were, like that's an experiment that we're running on people who are literally when they're looking at a service, they might be thinking about taking their life. But there's not really a risk profile to that particular change. You're you're looking at doing something that is, you know, quite clearly better. Yeah, because a missed a missed call might be the only time they make an effort. That's right. That's right. So there's a lot of opportunity there. I think also then that goes into kind of what we've been building without our first AI navigation product. You know, that has taken us, you know, a good while to be able to like get to a point where we have a high confidence in it in terms of like deploying it with a customer. And that's because it's kind of different to if you were doing like a, I don't know, CX chatbot play or something, you know, where it's like it just doesn't have the same risk profile, you know, you could find like early adopters that are like, are you all just give this a go and hope for the best? It's like, if there's a chance that this person, a person talking to this thing might be thinking of hurting themselves, you better make sure that by day one, you can solve for that. And so that meant for us, you know, we had to do a whole bunch of work upfront, you know, on things like risk classifiers and in guardrails, yeah, as opposed to being an afterthought. That's kind of an interesting place you play like because you take this risk layer, you are trying to ensure that the handoff is over to somebody who's actually well placed to help. And because you're sort of building out this network over time, it's not as straightforward as like, well, here are their like 10 best well-known players. And I can be pretty confident that they well set up to receive a referral as it were through three line. What did it look like kind of doing that vetting process or the kind of deciding who's who's on the list? So we like when I first started this, I like, I I reached out to kind of friends and and like pulled pulled together like a crowd work kind of moment. And I just like set everyone up on like a slick workspace and there was maybe like 50 of us or something. And I was like, we're going to map every hot plan in the world in a week. And this was like, you know, a number of years ago. And we are like still here, something like four years later, discovering services. Like it's it's a lot harder than people think. And we've got a lot better at at over time. Why do you think it's so hard? I think it's like it's particularly challenging because you're trying to do it like globally across a range of topics, across a range of languages and a whole different types of operators who have differing levels of kind of investment or public presence. So like something like Lifeline Australia, you know, we're talking about like a very prominent comparatively well-funded compared to other services the world service, you know, in a developed country. You know, that's very different to a government run service, you know, based somewhere in the global south and trying to get in touch with those services to figure out what that offer is hard. And in fact, our team is running an experiment right now because we've historically we've just, you know, reached out to these services how you normally would where you like, you know, you, you send someone an email and you hope to get a reply. But like one of one of our team members who works in the space Simone, he was a diplomat and used to work in government services. And so he basically thought about how when you're a government service, like, there's a formal procedure to reaching out and getting back to people. And so we've actually been writing formal letters and signing them and we created like a throughline stamp and we're sending out like, you have a stamp. I was going to ask if you had like a, yeah, no, we do, a wax, not wax, but it's just like it's a stamp, you know, and it says like throughline limited, it uses like office of the CEO to make it sound very essential. Wow. And we are like snail mailing out to like, you know, all these government run services all around the world because when you snail mail them, it then goes into a pipeline where they, they have to respond to you. Wow. As opposed to like an email where it's like, it just can, can be ignored. So they have some of the reason why it's so challenging. Yeah, I mean, but like I think it feels ironic that your snail mailing services in order to then say to them, can I now automate a process to bringing you people in acute moments of crisis where like we can't afford to wait for them person to write you an email with a stamp saying, I need help. No, you're exactly right. There's a certain and dissonance there, no? - Yeah, there is, but somebody has to do it. - Yeah, extraordinary. Let's go further into that 'cause you talked about 180 countries, I think your technology sort of services. I imagine, again, back to this sort of product and user experience, it's, do you have a very different handling strategy for handing off identified people who are in moments of crisis two different types of services because of cultural linguistic differences? - Yeah, so one of the benefits of working with a network is that we're referring to local services. So those local services are the experts in what support or cultural you support or service provision looks like in their particular location. So the approach we take is we have a standard across the board in terms of what we expect a service to adhere to. And so we make sure that every service is like qualifies according to that standard. And then that means we can have a singular promise to the user, which is like the laziest way to think of this is like, if I talk with you listen, it's kind of a shorthand for it. And so can we have that singular promise to a user? Then essentially be able to market place those services wherever you are in the world. You can access them and you know you're gonna be getting something that is someone that will support you but because it's a local service, you know, they're gonna do that with an understanding of the cultural context and the other services that are available to you. So like to be honest, if you try to do it any other way, like it would be impossible. It's like whack them all, right? Trying to, like if you were trying to build a product solution where you're trying to solve across any type of cultural linguistic user intent. - You'd have to make a lot of assumptions I imagine and they could be tricky. - Yeah, that's right. So by being able to kind of like build that new work, we kind of, you know, we leverage the diversity and the like nuanced support of the network as well. - Yeah, amazing. And maybe then the other question I would have coming back to the sort of product experience. What if you learned that might be surprising to the audience about how to handle that acute moment of care, you know, like when somebody uses and is identified by a given platform, let's say, as being in a moment of crisis, uses certain types of language. And I think from an earlier conversation that we'd had earlier, you told me that actually a lot of those diagnostic tools that a lot of the products use need to account for more than the obvious words because people are privy to not using those words. They use what emerge over time as sort of vernacular to refer to that. So it's not even as straightforward as like right align of code that picks up the word suicide or sexual assault or domestic violence, that's not gonna capture the majority of the cases. But once that's sort of been done, is it just a case of how quickly can I get them to a phone number or a service, a chatbot, or whatever the case may be? That is available or is there something nuanced about the handling that you've built into through line? - Yeah, I'll tell you a little bit of a story maybe as a way to kind of loop back to this. So do you know the story of Airbnb in photos? - Yes. - Right, so. - Tell the audience. - So before Airbnb became Airbnb, you know, we knew they didn't have photos on listings. And of course for Airbnb to make the business work, they had to increase bookings. One of the things they experimented with was taking professional photos for listings. And they noticed that they had a significant uptake in bookings, I can't remember what it was, but it was a multiplier I think in terms of the impact that it had. And so then they ran a whole paid project. They employed professional photographers across the US and sent them out to Airbnb bookings and had them take professional photos. - It was something like the first 50 or 100 hosts or something per city. - No, it was way more than that. - Was it, okay. - Honestly, they went to quite a crazy scale with it and spent a lot. But the point is is that when someone was trying to book an Airbnb, like they had a problem to solve, right? Which is like, I'm traveling somewhere, I'm going on holiday and I need somewhere to stay. But they did not have the confidence to book at that place because they didn't know what to expect. They didn't have visibility. Or if they did a visibility, it was like kind of a crappy visibility, it was a bad photo. They didn't give a really good impression. I tell that story because it is exactly the same when it comes to someone accessing mental health support or trying to get help when you've experienced an online harm. You have a problem, you desperately want help. The thing you're experiencing is overwhelming. And you also don't have the confidence to do something that is quite scary. Going and staying at someone's place that you have never met before, without any sort of visibility into that experience, is scary. And the crisis here, sector, they talk about this idea of picking up this 600 pound phone. That's what it can be like for a person in distress to reach out for support. And so the thing that we are fundamentally obsessed about is how do we address those fears and concerns that exist between a person in crisis and the help that is available to them. And that actually is a product question. That is at the heart of like what Thurline does. Like when I think of Thurline, ultimately we are trying to be a mental health and online harms user experience company. That's the intent. Is the best user experiences in the world that reduce barriers between people in distress and being able to get help? It puts you in a really phenomenally interesting position to have massive impact on the people that touch your product. But I imagine as you've alluded to it, it has lots of very interesting and challenging product questions associated with it. Maybe let's turn to Thurline 1.0. So you've alluded already that Thurline is already in its 2.0 phase and no doubt your product roadmap goes to 3.0. But I wanted to sort of ground us in what you have learned over the years, building out the sort of part of Thurline that people may know best, which is actually your integration with platforms like OpenAI and ThroPet Google Roblox Discord. And kind of that first piece where you're just sort of taking a person in this moment of crisis and handing them off to a valuable local service that might be available to them. I'm really interested because it's obviously deeply political this work now. It's in the media literally today, this kind of work that you're involved with. And I'm really interested in what it feels like also to be a product business, interfacing with some of the largest product businesses in the world. How did you get into those rooms? How did you even get the partnerships that allowed you to have the scale that you already have as a relatively small team working out of New Zealand? Did you write them letters with a stamp at the top? No. But I would say that the teams that we work with the kind of departments called trust and safety. And that is like a surprisingly small community. And so if you can build many for connections and build a modicum of trust, then you can leverage that in a lot of ways. And so for us, the core strategy in terms of working with platforms has been to-- I guess build some core trusted relationships. And then to be able to leverage those through warm introductions and then to be able to see if we can solve problems. I think the approach that we've taken in those conversations has been really important. In the early days, I didn't really know how to sell. And I would go into conversations. And I'd have this-- like what I thought was a really pretty deck and jump on a call and talk about what we do and try and convince people that it's a really good idea. I've learnt a lot since then. And probably the first things I-- one of the first things I learnt was how valuable it is to start a conversation by asking for their perspective or their advice. So a number of years back now, I went to the US on a three-month trip and went around the US trying to find as many people working in trust and safety as possible and having conversations with them. But I stopped myself from having sales conversations. I had a product, but I wasn't going there trying to sell. I went there and essentially a journalist. And I was just trying to learn. And so I'd have conversations with people and just trying to understand their problems. And in fact, one of them, I had a coffee. And at the back of it, I was walking to my next thing. And they said, you're not a journalist, are you? Because that's such the positioning I was taken was just really trying to understand their problems. And the core of that is you're getting to know someone. So you're building trust as a person. And they get to know me. And they go, oh wow, like Aliak really cares about these problems. And he's really intrigued. He's learning fast. And that builds a foundation that you can then work from. And so then as the product matured, we were able to reach back out to Folk and say, hey, this is what we've been working on. Do you think it could be helpful? That really, really helped a lot. And then that's also kind of informed a lot of how we sell today. I've built kind of really strong philosophies like this around what enterprise selling looks like. And at the heart of that is like, just never talk about myself. solution until I really understood someone's problem. So like my dream state is getting on a discovery call with a company and getting to the end of half an hour and looking at the time going, oh my gosh, oh, well, we're at time, I haven't even had a chance to talk to you about what we do, you know, I think we could be helpful. Should we jump on another call in a week? I want to understand as much about their world as possible because people buy things when they want to solve their problems, like they don't care, they do not care about me or my product or my company. They care about solving their problem. And so yeah, we've just become very obsessed with centering, yeah, centering the customer and centering the user. And I think that's gone a long, long way. So that's interesting. I mean, you've alluded to the, you know, the famous discovery call that all salespeople found as an otherwise need to learn of the sort of three things that can come off a discovery call. One is, you know, the typical version is that it sits firmly within a sales process and the idea is extract information from, you know, in the transactional version of this, right? Extract information from a potential customer that allows you to better tailor what your product already does, you know, change slide seven to slide one because that's the thing that they care most about. And, you know, power it back to them. I mean, there's all that sort of mirroring stuff, you know, there's loads of science in selling, right? So one version of the version of a disco call is transactional. It's still with an intent to sell. And it's just how do I do that better version of selling? Another might be that discovery is actually part of the product evolution input process, which sounds like you were using that a little bit. And then the third, which I think is quite intriguing is you've just gone after it as a trust building exercise and relationship building exercise. How would you sort of place each of those three things in terms of what it's given you and maybe on the product side, you know, in your three months? Did it radically change the way you were approaching the problem to take that approach with customers? Or was it really that it just helped you have a second version of that conversation? Yeah, I actually came to the end of that three months thinking there's not a business here. Really? What were you trying to build at the time that was different? I mean, we were doing, you know, through Line V1, you know, which was like a helpline network integrated into online platforms. But I came away feeling like, yeah, like there's a problem to solve here, but, you know, there needs to be, you know, a level of pain in terms of like how to solve that. And I just didn't feel like I came off the back of that and that the pain was acute enough for anyone to be able to pay for a solution. What's interesting about that is like I turned out to be wrong probably in quite a matter of months. But I think that kind of proves the point is that like I went in with enough openness and humility to really try and genuinely understand someone else's world. So then I can figure out like how what I'm doing might be able to, to meet it. And I think I probably have that temperament to do that because ultimately what I really care about is like making a meaningful difference in people's lives. And so like if it's just not going to work, like I just don't want to do it. I'll go do something else. Like life is really, really short. I don't want to waste my time. Yeah, I mean, I certainly think that us, a part of the insight like I, you know, gathered on that trip and continued to gather over time is just understanding like what the particular motivations and levers are, like amongst our buyers. Yeah, you just I think as a, as a company you always want to be getting more color on that, you know, it's like, it starts off as a sketch on a napkin and hopefully your company at full strength that's a, it's a beautiful money or something, you know. And so I think that that really helped, you know, I, because I, because of my background, you know, I came from this from a perspective of really wanting to help people in crisis. But, you know, recognizing that there are, there's, there's a whole range of levers. There's reputational risk, there's regulatory risk, there's duty of care, that there's all different types of roles and motivations even within companies. And figuring out, you know, who might be your relevant sponsor. Yeah, so, load a lot. So, I mean, I think probably anybody listening to this may be asking themselves the question. We're talking about some of the biggest, best well-resourced platforms in the world. They already do the identification piece as you alluded to earlier. Why is there actually still a market for through line to do anything for these big platforms? Like, why wouldn't they just build it themselves? So, it's hard to be able to build a global network that is reliable and that you can trust and then to be able to get that into product. Despite these platforms being very large in the user base and large in their teams, they're actually the amount of people that are tasked with solving that problem is quite small. So, you know, oftentimes when we're talking with a platform, you know, I'm talking with like one or two people who are trying to solve this, who have like no product in engineering resource available to them. So, trust and safety is often seen as a cost center. It can be challenging to try and get the resource to do the things that these team members want to do. And so, if we can shortcut that for them and basically say like, hey, you're not going to have to build us, you're not going to have to maintain it. And it's going to be infinitely better than anything you would do. And it's going to save you money. Like the advantage is I guess it becomes like any other type of value proposition where it's like, well, yeah, like this is better, quicker, faster, it makes sense. I want to then pick up, I guess, you alluded earlier to different platforms and even within organizations, you're dealing with different motivators. You're thinking about enterprise sales, which often involves, you know, who's the sponsor, who's a potential impediment to the yes. How much is getting through line over the line as it were is about someone championing because they deeply care about the problem? Like, does it matter if your partner is like, I'm doing this because the reputation or risk is there or the regulatory risk, it feels like a compliance easy, but it's not a passionary for them. Like, does that matter to you? How does that show up in sales processes for you and what do you make of it? So because we work with trust and safety teams, like it's littered with those types of people, you know, like trust and safety, folk are often people that have come from civil society. I know people that work in trust and safety who think the internet is net negative and that's why they work for like a large platform because they're trying to like make the whole thing better. You have that benefit. You get on a call with a trust and safety person and they all really, really care. And so you're actually on like the same side of the aisle where you're both standing there going like, hey, we want to solve this problem. Obviously, we want to solve this problem. You know, we just need to find a way to be able to do that, you know? And, and they're trying to use their levers as well because they might not be the ultimate decision maker or the budget holder. They may have recognized the duty of care. We may have been talking with them for like six to 12 months where they really want to do something. But then there becomes like a reputational risk moment for that platform. And so then they come back to you and they say, hey, this has actually happened and so, you know, finally, we've been able to get some executive sponsorship on this and so I think we can do something now. It's recognizing that the people we're talking to are like R on our team. We have to work with them on understanding like what the levers are for them. And it depends on the person. It depends on the platform about how overt or opaque, you know, that messaging might be. But, you know, the more you can really get on site and understand like, this is the problem that you're trying to solve. And here's like the motivations and the drivers behind it. The more you can enable that sponsor to have everything they need in order to get it over the line. And like our job is to make them the hero. Like we're not, we're not Skywalker, you know, they're Skywalker, we're Yoda. And so our, our job is to make them the hero. And they want like, this is a thing like when you, when you integrate through line and you get something in your product, it's like, we can help a lot of people. And it can be some of the most impactful work that you do. Yeah, well, I'd love to get into, you know, the reality is that through line is building right in the heat of some of the most interesting regulatory developments happening around the world. So obviously here in Australia, we, you know, relatively recently implemented under 16 ban on social media. There's a UK Online Safety Act in play. The EU, US are all looking at this. Some open court cases. I'm really curious. What does it feel like building product so deeply in the heart of one of the sort of most clickbait were the sort of discussions that is happening in technology? Like, is it a thing that's you're very aware of as a team or are you sort of just used to it and keep your heads down, keep doing what you do? No, I mean, we're very aware of it. Yeah, it's got benefits and challenges. It certainly provides a lever. And there's a wave of regulation that's happening globally and also reputational risk, certainly AI companies have seen what has happened with social media and have attempted to be a lot more front-footed and trust and safety. And that kind of makes them rhetorically trapped, you know, like you, you see yourself as a B-Corp or you, you have a narrative around, you know, wanting to build safe general purpose AI. And if you say that, like you then kind of have to invest in it. Yeah, that's certainly like being very useful when it's and it's created a lot of opportunity for us. Also, like regulation is generally quite a blunt instrument. The challenge that we find is that especially in the AI space is that there is not a. enough understanding of how platforms where work, how this technology works, how user psychology works in order to legislate effectively. And so what's happening, especially in like the the mental health space with AI is this kind of conflating of consumer algorithms with like all of them technology. So it's like it's like saying that the internal combustion engine is like the equivalent of Ford or the Ford motor car. Like like those things are different like like alarm is a particular technology that has like far reaching potential in use cases and chat GPT and Claude and Gemini and Grock or whatever are particular applications of that technology. And so that simplistic understanding then results in the in legislation that tries to solve for risks that might be present on like a consumer LLM standpoint. But then walls are soft from the potential of this technology to actually make the difference to the various things we hear about. And so you know for us that's one of the things that we have to wrestle with a lot. Like we're building AI for mental health. We're not building therapy. We're not building a coach. We're not building a companion. We're building something to help a person in distress navigate to the right information or support for whatever they're going through. There is an incredible opportunity to address fears and barriers and to like to smooth the path for a person to get help using this technology. We live in our stories to be able to tell us my story. Like a teenager doesn't know what sex torsion is. A teenager says I share my notes and now they're asking for money. And so there's amazing opportunity to use this technology to help a person be able to get to the right support just by talking. If the legislation just views AI as just being like consumer LLM's and then you know legislates against those products like we can end up in a situation where we can't even we can't even develop the types of innovations that we need to solve these problems. And so that's one of the things that like concerns me and we have to stay abreast of is like you know when we're launching launching products in certain jurisdictions it's like does our product break this legislation or not even though it's confundantly different to what the legislation was purpose built for. Yeah interesting. I mean it puts pressure on you to engage in a kind of regulatory conversation or at least be aware of it where other products in less sort of hotbed areas don't have to. How does that sort of affect your product strategy or the sort of go-to market strategy for you do you actively engage with policymakers before getting in or do you decide to be one of the voices trying to educate policymakers on the difference between the fundamental underlying technology that creates and makes possible in LLM that could save lives and the rapper layer of like how it's deployed yes you can you may have views on it's on how it approaches that. We're trying to do more of it. It hasn't historically been the main focus you know largely because we've been obsessed with our users and our customers and our product. So you know having the capacity to maybe like do that upstream work hasn't you know we just haven't really had it. We did have one scenario where we like we chased down the people who were tasked with implementing SB243 which is like a Senate bill in California around like source high risk on companion chapots and like for a lot of connections we ended up talking to like a department of health in California or something who are basically like given this legislation and they were told like hey you're the ones that are you know tasked with implementing it and we were asking them like okay so what like what are you going to do and they're kind of going like we don't know we've just been given this you know and it's like it's completely toothless platforms are supposed to report on what they're doing but they don't have to and if they don't then nothing bad happens like that's part of like where we've had these frustrating moments we're definitely trying to do a little bit more upstream stuff like in like one of the things that we've learned is that like platforms want to solve these things globally so regulation and one jurisdiction actually ends up having a domino effect. Yeah I mean the social media ban seems to be having a bit of that effect. There's a bit of nuance there because that's slightly different because if you if you have a social media ban in one country then that provide the precedent for another country to say hey we want to do the same thing. When it comes to AI products AI companies don't really want to have 200 versions of their product and so when you saw like the thing is when you solve that safety issue once you've solved it it's like marginal cost to then solve it for the rest of the world like all the effort is actually in the the product and policy work to solve that particular problem so if you take something like suicide risk and you go like okay cool we want to do a better job of solving this problem like once you have done it in one place like there's no reason why that platform doesn't then just roll it out to the rest of the world that then creates this useful lever where if you could have legislation in one jurisdiction that effectively creates the requirement for like a better approach to say something like source I'd risk yeah and then the companies then enact that they're then going to do that for the rest of the world like it is it's just the dominoes just for and so that's one of the things we've been thinking about you know particularly in the US because you know that's where you know the public companies or you know investors in private companies are based you know you can also have the lever of state legislation versus federal legislation there and so there's an opportunity here where you know if you could get something over the line that is well-scoped and appropriate you know and it's built for someone in Nevada you could you know have that making a meaningful impact to someone in Brazil or the Philippines or something like that yeah do you have a guess about the direction of travel of all of these pieces of regulation and legislation and what space because I guess you've got AI you've got social media and my opinion is probably diverged depending on the on the type because the police platforms are very different yeah well maybe let's start with social media because they are the ones that have been sort of first hit as it were and you know long-standing court cases that have led to regulatory pressures that some would describe as overly paternalistic and simplistic to deal with what's much more complex and other people would say well they've been waiting for the industry to step up in self-regulation and they haven't and so they've chosen to take a blunt lever like what's your take on how all that stuff is likely to play out yeah I mean if we're seeing it already right I think there's like is there about a dozen countries in the world that are considering social media bands as they're called and I think we're just going to see more and more of that like I said I think it's a little bit of a blunt instrument like from what we're seeing in Australia as well it's like we're still we've still got young people that are like they're the depth of using technology you know like they're still able to kind of get on these platforms and they they want to connect they want to be entertained and so whilst it may have some impact in terms of like lowering risk for some populations I think what it does is like it creates a simplistic answer to a very complex problem but it's also hard to create the it's hard to have the the appropriate solution or answer because these bands happen because of ground swell because of simplistic media narratives because they're easy wins for government if they have a constituency that's really asking for that what I'm what I'm hearing you not say is that it's inaction on the part of platforms that have a capacity to design technology away from the harms is I mean I imagine that comes from a knowledge that it is genuinely difficult for these large social media platforms to eliminate the risk that this sort of community thinking about it for 25 seconds believes they should be in a position to do so yeah I mean these are societies I think that's what we have to recognize is that when you're talking about like a large social media platform it's the equivalent of the people in your country you know and trust and safety team members are like the policy people they're like the social development ministry they're like the police like the other ones that are trying to figure out like how to make this thing safe but the police will never completely make your community safe you know like there's always going to be bad stuff that happens and so of course regulation is a lever and it's a really important one and so of particular harms are happening in our society we want to make sure that you know we legislate against those in order to reduce that harm the same thing that happens with platforms but I think I think that's kind of the point is that like because these are societies your instruments need to be as specific to the harms that are happening but often they're they're incredibly blunt the the equivalent of saying like just like don't go to that part of town anymore because bad things happen in that part of town it's like it just does not work like that you have to think about like what is that part of town and why it like what are the particular particular challenges there and how do we like solve those particular challenges I think the challenge here is that like someone that worked in trust and safety said to me a year or so ago like we have got a lot better at building platforms with a billion users than we have in building the like human infrastructure to support them whilst technology and online platforms have eaten our lives and we spend our lives on them the rest of our institutions are still the same as they kind of essentially were 30 40 50 years ago but the fundamental way that we move and be in the world and connect with others is inherently digitally connected right now but none of the institutions are caught up with that. And so what that means is that when we look at the problems that are happening, we can't have nuanced, graded instruments to try and solve for those specific problems. We just reach for really blunt ones. We go like, oh, well, this is a really bad problem. We should get rid of it. It's like, I think with online platforms, like the metaphor of cars is a really clean one. Because when we'd fit cars first came out, there was all kinds of crises of like crashes and harms and all these problems, and it took like 50 years for us to get seatbelts. But people recognize that like, well, cars have this an incredible inherent value. And so we need to have very specific types of legislation to make sure that they are safe. So we need to have seatbelts. We need to have airbags. We need to have licenses where people earn the right to be able to drive. We need to have speed limits. Like, you know, if you think about all the different types of legislation that exists when you're on the road, it's very nuanced compared to like the legislation that exists when we talk about doing things online, which is often like very blunt. My fear is I don't know how we come back from it because once you put those blunt instruments in-- No one wants to be the person that was like, I removed it, and then, you know, some horrific singular event happened the next day, right? The story writes itself. Yeah, you know, if we wipe the slate clean right now, we went, hey, this is what society looks like in terms of online behavior and these platforms. You know, like, the way we build our institutions and everything, like, we just look fundamentally different. But we're having to retrofit. And I understand there's a reason for that. But it means that we basically do it imperfectly all the time. And I think, I mean, this also leads to why storyline does what we do is because, like, my belief is that, well, the best way we can make a difference here then is like using market levers. You know, the government will never lead, it will always follow. And so if I really want to make many full difference, where the most people are in need, which is online, then I need to figure out how to use market forces. Elliot, we've talked about through line 1.0 and through that, you're already touching tens of millions of lives on an ongoing basis, which is pretty extraordinary. But I know you to be very ambitious about what through line can be in the future. And you don't want to stop at what you've built so far. What are you doing now and what gets you most excited? So this kind of tracks back to, like, my feces around building a startup, which is that building a startup is all about earning the right to solve the next problem. You might start working on a particular problem, but you know that's not what the ultimate vision is, but you know that, like, you have nothing, like, you have no product, you have no customers, like, you've got to earn a right to get to the next one. And so for us, doing the work that we did with online platforms and connecting people to crisis resources around the world, was like solving a very obvious problem. But the thing that I have, like, always been aware of is that it actually does quite little to address the fears and concerns that people have about accessing help. And it doesn't recognize the different types of help that might be available to a person. It's like, it's quite a one-size-fits-all answer. And also that a person in distress crops up anywhere whether that's an online platform or through a health system, or through an insurer, an advocacy organization, or an employer, or an education provider, and that person in distress, like, doesn't really care about where their entry point is, they just want it to be easy to be able to forget how to get support. The thesis was always to earn the right to be able to, like, work on a really more gnarly problem, which really is about making it easier for people to get to the right help. The metaphor I used to share this is, like, imagine you're standing at the edge of a ravine and out in front of your stretches, it doesn't bridges, and you want to get to the other side. On top of that, you have a debilitating fear of heights. That is what it is like for most people in distress to figure out how to get help. If you are struggling with your mental health, if you have experienced an online harm, like, that's the experience. Like, you just do not know what bridge to walk across and you're also just really, really scared to do any of it. What has really excited us is that the, like, advent of AI technology enables people to just come with their story and come with their lived experience and start with that. And then it's our job to try and understand that and then to try and point them in the right direction. And so that's what the future of what we're doing with our first step product is, is building a platform that enables, like, customizable instances to solve that across anywhere someone might be in distress, online platforms and anywhere else. - So what might that look like? You mentioned earlier through healthcare providers. Imagine I kind of walk into a GP. One would hope that they are trained to know how to redirect me to services, but we all know that it's not uniform across the world and access to those services they may not know even, you know, who best to send you to, could through line be involved in that interaction and how would they be? - Yeah, so we're thinking predominantly about that, like, online moment. It's that sense of, like, where someone is in distress and they want to maintain anonymity. So it's maybe looks a little bit more like you're struggling with your mental health, but you don't really know what to do, but you land on a website that you trust, whether that's like a website that is a large government funded, promoted one, you know, or whether that is a direct to consumer therapy website or an organization that you trust, or it might be that you're a student at a university or it might be through your employer that whatever that entry point is, you just have a struggle and you're trying to figure out what your next step is. And so what the step does is it basically gives you opportunity to be able to start with your story and then we unpack that story with you and then we work with our customers to plug in different types of support pathways. And so those support pathways aren't just things that we bring to the table. They're things that they might bring to the table. So like in the case of like an education provider, that might be like, well, here's the different types of counseling options or support group options or here's even maybe products that we've built that, you know, we've purchased that can support you. And so to be able to plug all of those into like a singular front door where a person can have a conversation and they don't have to unpick it themselves. It's the same case as if you were telling me about something that was really hard for you that you were struggling with. And in my brain, I know all the stuff in your area that might be able to help you. But you don't, you just have your story that's all you're bringing. - Yeah. - And so if I as a friend was to help you, then I would recognize my place, like I'm not a therapist or I'm not a psychologist, like I'm here to help never get you to help you figure out what is the next meaning for next step. And so there would be a couple of agents doing some work, right? There would be the part of me that's talking to you that's trying to hear your story that's empathizing with it. That's asking questions that's trying to get some clarity on your situation. But then I would also have some agents in my brain that are mapping what you're telling me to what might be some support options available to you. And then I'm gonna weave those in and I'm gonna talk to you about what those options might be and see what you might feel confident to do or not confident to do. You're gonna have particular constraints. You're gonna have particular needs. You're gonna have particular fears. And so my job is to make you feel confident that by the end of this conversation, you feel like you have a meaningful next step. And if you take like any organization that represents people experiencing that, they care about making that as seamless as possible. - Absolutely. But how do you think about prioritizing which customers you go after first? Like how do you know where online outside of the platforms you're already integrated into? People are more likely to share the first tidbit triggers hey, first step might be the right interface for this person at this time. - Yeah, so for us like we've, I guess what we're trying to do is like hunt down the markets that are already feeling this pain acutely and already have like a high duty of care to be able to solve for that. You know, we're seeing a lot of that momentum in like the community and public services space. You know, these are organizations or sometimes even government entities who they're like expressed a mission is that navigation. Like that's why they exist is to try and figure out how to help people get to the right support and information. - And go from a sort of static website with some information onto it. - That's right. - To how you can actually use us as a portal to get where you need to go. - Yeah, and it's generally like there's thousands upon thousands of these all around the world that governments have funded or that organizations have spun up. And it's like one of the key things they celebrate in their annual reports is like, hey, look how many people came to our website and went through this flow and they ended up reading these articles because that is what the internet used to be. Where like user behavior has fundamentally changed and AI affords an opportunity for a person to be able to like have more input into the way that they access support. And so yeah, that for us is like a really significant beach head. The other kind of one here is there's lots of service provider organization. that do struggle with having enough supply to like meet the demand. And so our technology has the ability to solve in that space as well, where it can function as a kind of intake layer to kind of keep someone warm or even to provide someone support, you know, when there's like lower risk presentations. Yeah. And so like doing that in like the mental health space, or the online health space in and of itself is like a very significant opportunity. And so like we're quite squarely focused on being the best in the world of that bit. So I mean, a overly simplified version is like the 1.0 is a sort of redirection service to a really sophisticated triaging layer in between those two parts of the market, the place where people find themselves presenting with something and the place where they can get the best help is that a fair description of the evolution. Yeah, I mean, it's the difference between like a yellow pages and a conversation. Nice. If I'm landing in a new city and I don't know where to eat, it's like back in the day, you open up your lonely planet and you go like, "Oh, okay, water the restaurants here. This is a guide. It's, you know, it's having someone that's walking around with you and talking to you about what you feel like eating." Yeah. And, you know, understands your dietary needs and then there's like, "Well, open up your wallet and figure it out whether you can actually afford it." That's right. And I've like, "I've got the place for you, you know?" And so like it's a user experience problem, right? And people are solving similarly user experience problems with conversation where I like across thousands of markets. For us, like the differentiator for throughline is that like, it takes a particular type of product experience to do that well for people who are experiencing emotional distress. Yeah. It's far less transactional and the risk management piece is like, is foundational. And so it's very, very hard to solve for. And so like, our bet is that there is enough of that problem for us to be able to basically like build the biggest mental health intervention the world's ever seen. I know, thank. And what do you, what do you make of the kind of, can you imagine yourself moving into? And then we pick up more of the clinical support on the other side. I mean, there's obviously this huge boom in AI, psychologists and various therapeutic manifestations of AI. Sounds like you're not doing that right now. You're still focused on how do we connect these people to these trusted networks of suppliers. But do you imagine that changing? plausibly. But I think there's like, there's two reasons we've taken the approach we've had. One is I actually believe like this is a significant overlooked market opportunity. The kind of the therapy AI builder, it's almost an obvious way to use this technology. But if you're trying to build like a really meaningful company, you have to have a buyer and the risk profile of people choosing to like buy AI therapy, you know, whether that's like an individual consumer right through to like a company, like you're not going to see enterprise platforms anytime soon, like stop the EIP subscriptions and just say just talk to this bot. There's just too much risk profile there. So like pragmatically, like there's a really significant market opportunity there. And so we're like, well, we want to make the biggest efforts we can in the shortest amount of time. So let's do that. The second thing is that like this, this navigation piece is a far kind of lower risk and safer play than doing like the therapy play. Like for us, like we don't maintain memory. There's no ongoing relationship, you know, that in and of itself like lowers the risk profile significantly. We're on the therapy side like it is still largely unproven and you know, we've only really got one RCT on generative AI, like applied in the therapy context that is like shown some early results. But we've got like a long, long way to go. I think before we can really, really prove that out. And I think there's also like it creates a heck of a lot of risk as well. If you are trying to bite off like I'm having an ongoing relationship with this person and I'm and this is an AI, like the duty of care is like very, very, very high. And clinicians are trained to be able to deal with that. And there's all kinds of kind of structures and accountability to to make sure that's done well. Trying to solve for the NNIR context is very, very challenging. So when I kind of look at it, you know, it's like, well actually like, why are we, why are we overlooking this significant market opportunity that is available to us right now? And for us during the in the right image again, that could be a part of the future of Fairline, but we'd have to end the right to get there. And I would have to see the market need for it in order for us to feel like it's going to be a meaningful area of growth. Awesome. Anyone listening to this is probably thinking, what on earth does it feel like dealing in every day in the sort of themes and content that you and your team deal in? Can you share a bit about like, how do you cope with that reality? It's incredibly motivating. And I don't have a single day where I think, what's all this for? But it can also be paralyzing because you know the extent of the need and you feel so far away from the destination. As a leader trying to strike that balance with my team is hard. I see it as my role to have my team feel an appropriate level of grief and empathy for the suffering of others coupled with a high sense of agency that they can do something about it. We have a number of rhythms at our work to try and do that. And I'm sure that we will learn more about how to do that better. Where did you get the inspiration for those? What are they? I think what it is is that it's like I, you know, I, because I worked in community services for a long time, building meaningful connection with other people where a human level is like bread and butter for me. And I want to be in a workplace where, you know, we know each other in a human level. And my fundamental belief is that you can't achieve anything meaningful together unless you have genuine human relationships with one another. And so like by design from day one, I've made sure that through line centers the person and our team not just their role. So some of those rhythms, you know, for us look like the way that we do team meetings and check-ins. When we come to a weekly team meeting, like we start with an individual check and like every single person. It's not, we don't go straight to the agenda. Oftentimes it's like, you know, a question about something you did on the weekend. You know, how are you feeling, you know, at this moment in time and some people are really struggling and some people are doing really really good? That's a kind of key part of our like one on one culture as well. Like it's making sure that we give oxygen to that person's life, not just to their work. And so that means that like a lot of times with team members, like I'm not talking to them about the work stuff. Like if I'm like, I might do it, I might have a one on one. And like two thirds of it is talking about what's happening at home. Like to me, that's really really important because that builds a sense of trust that like, hey, this is like really hard what we're doing. And it can be really affecting. And then I've also got like my own stuff going on. But like I believe leadership is like here in the trenches with me and has my back and can listen to me and can can empathize. And yeah. And so having the person that this enter like has been one of the biggest things for us. And I think we see the fruit of it, you know, and our team. You know, consistently people talk about through line being like, no other workplace that they've ever had. And I think that's because like when you walk in the doors of through line, you walk in as Kate or you walk in as Elliot. And we interact at that level as opposed to our roles or our functions or our hierarchy. If I've got crap going on, like I can bring that into the room and I can trust that people will hold it. And culture replicates, you know. So it's my job to be able to set that culture from the top and make sure that my leadership, you know, also sets that culture. And then like the things that's most joyous is when you see, you know, new team members just replicating that culture themselves. That's incredible. Elliot, our final question is always, what's something you've changed your mind about? I spent 15 years working in the mon prophet because my north star is to reduce suffering in our world. And I applaud anyone, whatever you're doing, if your attention is given to those that are in pain or hurting or in need. For a long time, my belief was the best way to do that was under a non-profit model. I came to realize that the incentive structures that exist not always, but in many cases and non-profits disincentivise innovation and iteration and the ability to actually solve problems at a meaningful scale. And so my journey into startups was actually because I wanted a bit of vehicle to be able to solve the types of problems that I want to solve. If you talked to me 15 years ago and you like 15 years ago me talking to me now, would be like, I don't really know that guy, but why is he doing what he's doing? But for me, it's actually like it's a deeper level of mission alignment. I feel like I'm finally at a position where I'm able to get to be obsessed about a line of all the incentives from me, right down to our users, and everyone in between to make the biggest difference in the world and because of that I'm making what impact I'm going to have. Congratulations, Elliot. Thank you so much for being on Wild Tarts. Thank you, Kate. It's been a joy. Thank you so much for joining us for another episode of Wild Tarts. If you want to learn more from other ambitious people building, designing and creating the world that we all want to live in, then please hit the subscribe and follow button. This podcast is produced by Camilla Herring from Blackbird. Our marketing genius is Laura Kohford and our editor is Andy Jones from Color and Sound Creative. Thank you all so much for listening and I'll talk to you next week.