Des Traynor (Intercom): Design, PMF, and What Founders Get Wrong Early

28:41And then obviously Android and all, and then making other versions, can you skin the messenger, all that sort of stuff. They're all things that we would do once we had kind of been like, yep, it's definitely a thing that people want to see and talk to their users. And that's why it was called Intercom. Eyes and mouth, see and talk to users. The whole thing makes sense. Right. Everyone's buying this. Let's go. You know.

29:01Des Traynor:I mean, it sounds like you actually sort of ended up finding product market fit quite early. A lot of startups, you know, they have to pivot. They have to kind of wander around and try a bunch of different things that don't work. And sometimes it is the product, sometimes it's the position, etc. But it sounds like there was a real market kind of pull, which is great. Well, with the caveat that, as we said earlier, we weren't trying to build an intercom. We were trying to build an exceptional, and an intercom was the byproduct, if you know what I mean. But yeah, you are right. I think ultimately, look, SaaS was a richer, more easy to play in space back then.

29:34There were a lot of ideas. We had loads of other ideas that we thought were going to be really important too. We had ideas for things that ended up looking like Amplitude, ended up looking like any of the revenue analytics products you see today. obviously then like you know I'm trying to think a few others like we like ideas for like product research and product discovery but that's because SaaS was probably realistically only like four years old like their sort of heyday of Basecamp and Rails and all that sort of stuff so like there was still the idea of building products for these people who were inspired by that wave to build their own products was pretty lucrative like most things hadn't been done.

30:12I think SaaS today is drudgery largely because it's been tapped. And I think what you're seeing now is all the people who are in AI startups and solving internal AI-shaped problems, like, oh, we need to version our prompts or whatever, they're now going off and starting their own prompt version or AI product or whatever because they realize, I'm going to build stuff for the AI guys, which is the same thing. So it's like every sort of technological wave has this, like you can sell the pickaxes or you can sell the gold or whatever, or you can do a chase for gold or whatever. And I think what we're seeing here is that again.

30:45But I do think, yeah, we got product market fit early. I think a lot of the companies in that era would have because it was a very, if you were willing to take the overall market risk of, I think SaaS will be really big. There were a lot of opportunities you could have chased. And I think we chased a really good one, a really valuable one. I'm one that kind of suited our skills as well.

31:03Des Traynor:I mean, again, I think, as you sort of mentioned, timing is super important. You were sort of in a bit of a greenfield time where there was lots of opportunity to build new platforms from the ground up. Back in the day, I think it's still true to some extent today, like the vast majority of SaaS companies seem to be like an interface on a spreadsheet. And it's amazing how many companies still see today like that. But you were doing something quite fundamentally different. And there was really opportunity there. and it sounds like you found some really good backers in the early days and you really leaned into that.

31:34Des Traynor:Reaching product market fit is often the first challenge and a lot of companies don't even get there but it's easy to drop the ball even once you've got there when you're going from like we found product market fit to scale. I'm just kind of wondering like what were those sort of you know those scaling days like? Did you have any particular rituals or processes that allowed you to scale your team from five to 50 to 200 and beyond like you know what what advice would you give to a founder that's going through that same kind of scaling journey today the advice i'd give is probably slightly different to what we did what we did was um we ultimately like the thing that allowed us scale was understand we did like a lot of research specifically actually hired bob mesta and the rewired group and we we looked at how everyone was actually using intercom and we found three or four very different usage patterns.

32:22It was basically a customer service, there was product onboarding, there was product research, and then it was sales, basically. That's the different ways in which people were actually using Intercom. And we decided to make each of those things individually purchasable. Because a lot of people would say, I really wanted to use you guys for customer onboarding, but we're a Zendesk house. And I'm like, why the hell does that matter? Because the answer is, the way we were marketing ourselves at the time said that it was all in, or you had to do it or not do it. So we ended up saying, well, how do we increase our market share?

32:53Well, we could spread the product, sorry, split the product into individual components and that would work. And that basically worked. We went from selling Intercom to selling products called acquire, engage, resolve and learn.

33:10Des Traynor:And these are sort of like jobs to be done for everything to some extent. Very much so. So I guess it creates a bit of a routine. Like if someone's happy with their current product, but they want just a small piece of what you offer, I guess they can buy one of the five tools and then you have the opportunity to slowly sell. Cross sell, share wallet, all that sort of stuff, land and expand or whatever. And generally, like honestly, that worked. What we got, in my opinion and in Owen's opinion, I think wrong was like trying to push all of that upmarket got really hard because it was basically four different products.

33:44and you like to think, well, they all have shared functionality. The challenging stuff is people using the onboarding thing are all about marketing, attribution, analytics. People using the sales things need really deep Salesforce integrations. The weight of each of the products is pretty heavy when you go upmarket. If we had stayed perfectly in mid-market, I think we could have stayed where we were. We ultimately went upmarket and then we're skipping ahead here, but that was what ultimately led to us zooming in and support and led to Finn and all that other stuff. We can get into that if you want.

34:16But I think that's the, that was what we did do. What I would advise founders, if you have, if you've, if you feel like you've really, you've caught onto something, I would say like, trust your intuition because clearly it led you to that point. Don't lose the muscle of actually talking to customers. Don't let somebody tell you something about upmarket or at scale, like as if it's this sort of predicate that lets them rant anywhere they want. Because you'll get a lot of this, well, I'm sure that this works fine, but at scale, you're going to need a team of blah, blah, blah. Don't believe a word of it until you've actually seen it.

34:55So I think basically avoid the force-fed advice that you will get. It's almost always out of date and sometimes just was always wrong. uh and and and like so that's the first thing i'd say secondly is just like don't move too fast and sometimes the fundraising can trigger this where you you're after going and raising 30 million on a valuation of 120 and now you but i'm sure need to grow into it so you better get going like i think that can actually trip you up a little bit as well like it's really really important like if you're onto something good the chances of you uh making a mistake that screws it up is actually quite high.

35:33So don't move too fast without understanding exactly what it is you're doing. And that often means not outsourcing understanding of it to your newly hired COO, to your board, to your most recent investor, but actually means you go and you understand it properly yourself. And don't do it until you actually understand what you're doing. And if that means talking to 15 people who've done it before, make sure you're talking to the right people. Make sure you're talking to the guy who scaled go to market at DocuSign, cool, his version of events makes him look like a goddamn hero. Go talk to the CEO of DocuSign in that period and ask them how do they scale?

36:10Because that's a different story almost always. So anyway, there's a lot to that. But I think in general, if you feel like you've backed a winner, be careful about screwing it up.

36:21Des Traynor:Yeah, absolutely. I mean, this is definitely what I wanted to lead to. And just to your point, we will definitely talk about ai i think i'd lose my official podcast license if i didn't talk about ai so we'll get on to that in a second but um i think you're absolutely right i mean again on a smaller scale like i always sort of advise founders before they start going out and hiring their cmo or their head of sales or whatever you want to kind of do the job first because you actually don't know like if you've not got your playbook you don't know how to hire someone to fill that playbook and if you hire someone that's never built that playbook before and expecting them to kind of do the hard yards there's a really good chance that they they won't that there'll be someone that operates from a playbook rather than builds and so I think that's true around all these things I think hiring a you know a you know a head of growth or you know a head of HR or COO like your best route is by feeling the pain in doing it yourself and being able to kind of bring someone in and I think you're absolutely right everyone in a job interview will tell them how they're the hero of the story and I think asking around and trying to get different people's opinions I think would be super helpful um like was there a point where you suddenly kind of like the team was growing so quickly that you'd come back after weekend and there'd suddenly be 30 new faces in the company that you didn't recognize and like if you had that kind of mega growth like if there was a point where you went from 200 to 500 in a blink of an eye like how did you as a founder deal with the the changing culture that must have brought yeah I mean I guess uh me personally the thing that I always noticed was I was living in San Francisco for bits whenever I'd come back to Dublin the office just felt like it was like like I'd regularly come back to find my desk was gone I'm like you know why not get a desk but like we were just growing so fast we were like you know just like meetings in in like every single possible room in you'd literally be walking in a stairwell and there'd be three people with their laptops out trying to discuss the design or whatever it was just like that type of thing was always crazy to see and then more recently i think um post covid and post owens return we had an event um just to kind of kick off the sort of the next wave of the strategy which is about ai and all that sort of stuff and it was the first time we had basically the whole company uh well not not even the whole company we'd half of the company to all of the north american employees were in san francisco all of the dublin team all the european team were in dublin that was the first time i was like jesus this is a you know the last time we'd been all together would have been like 2019 or 18 or something I thought it was the first time I'd seen that many people, which was crazy as well.

38:45On culture, I would say...

38:47Des Traynor:How many are you now? About 1 ,100 now. Okay, wow. Yeah. Good size. It's like relative revenue, all that. I feel good about the headcount. The thing I would say on culture is I think there's a few different takes, and they're going to sound inconsistent. One is like your job isn't to preserve the culture. It's to let it evolve in a way that you believe it needs to evolve. and to stop it from evolving in a way you don't want it to evolve and changing it when it's no longer fit for purpose. And I think people get that wrong. And the ways in which this can go wrong is, a very common thing you'll hear in a startup is like, well, we've just hired Billy.

39:28Billy's our head of sales. And then some 20-something who doesn't yet understand business will say, well, now that we've got a sales team, is this going to change the culture? And the founders are tempted to say, no, no, no, no, of course not. We're still the same team. And the actual answer you have to say is yes, we are deliberately changing the culture to be one that is fit for the business that we want to be. And that's an example of like, you have to reject the idea of cultural preservation. There are other times when you hire somebody who decides that your culture needs to change and you don't agree, or you don't know what they're actually saying, but they've said things like at scale and we're going to be upmarket and all sorts of shit.

40:03So you just think that means we're going to treat our customers like shit and keep them at a distance and all that. And for some reason you let that happen. I think you have to reject those changes. And then later on, it turned into more about like, look, we're competing against every high-flying startup out there. Like, you know, you've seen emergence of things like 996 and all that. Like, it's a real thing. Like, the stakes and the competitive battleground of startups is really, really high. And as such, you need to build a culture that's like fit for battle in a sense, right? Which means obsessing about the right things and not talking about the wrong things.

40:35And I think that change is the one that I think most people chicken out of. It requires a pretty brave dictate from on high. And that was one of the things that Owen did when he returned, was sort of say, we need to adapt our culture for this new ground we're heading into. And I think it's easy to talk yourself out of that one. It's extremely easy to say, no, no, some small amount of change will be fine here. We'll do something figured out, sorry, something that's like a little token gesture. And then we'll go back to the same things we were always doing. You need to go around deleting processes and all sorts of stuff to actually get real change.

41:08Des Traynor:yeah i mean i i think a lot of small startups really like to use the metaphor still to this day around a family um and it's lovely and it makes everyone feel really warm yeah but you know like you you don't let members of your family go if they underperform and you keep everyone around and you want to make everyone happy and i think the sports team mentality is is a much sort of better one because you're there to do a job you're there to deliver an outcome and if people underperforming it's not because it's a you're a mean person for letting them go you have a conversation around performance and if someone else comes along that can do a better job your goal is to kind of get to the you know the top of the league and so I think you know it requires quite a bit of a a leap particularly if you're someone that comes from a small agency kind of we're a family mode and then you'll kind of know we're a performative unit I think it can be challenging a lot of the problems I see because again I work in the early stages is I often see you know like the people that want to go work for a 20 person company often very different from the people that want to work a 2000 person company and actually if you're someone that really enjoys leading a 20 person company like suddenly you start scaling and you want to try and have this sort of feeling well I can run a 200 person company like a 20 person company and there's a real avoidance of processes a real avoidance of of kind of um kind of the things that you need to scale and often that can be quite challenging because you become a you become a you know you can become a real seagulling leader that kind of you know micromanaging because you're you're losing you know you're wanting to keep control of everything when it was a 20 person company but you can't possibly be in every single decision at 200 or 2000 and so I think often you know some of the reasons that people struggle with scaling is because they don't actually grow the revenue or the user numbers as fast but sometimes it's because the the culture kind of hits a brick wall because there's there's resistance in the founder or there's not resistance in the founder but there's resistance in the founding team and that can create some really challenging conversations so you've talked about it a little bit you've hinted at kind of not hinted but you've mentioned finn like i'd love to kind of understand the evolution of how you went from a more traditional human power tool into realizing that AI was this thing you needed to lean into heavily.

43:28Des Traynor:How did that all come about? Yeah, it was November 29th, 2022. Everyone remembers where they were when they loaded up ChatGPT for the first time, right? Obviously, the texts were flying. Have you seen this? This thing's crazy. Oh my God, this feels like a real step change. Bergle, who's our head of AI, our chief AI officer, that wasn't his title back then, as you'd guess. He pinged me and said, hey, have a look at this thing. I played with, I met Virgo the following day and then a phone call with Owen that evening. And then we spoke all through the weekend. I think the things that were obvious was like AI was going to change software because we basically had this new set of probabilistic behaviors that weren't previously possible.

44:08It wasn't obvious what that meant immediately, but it was obvious it was going to have some useful applications. We agreed on that we'd start building something immediately. And four weeks, Five weeks later, we were actually the first large software company to integrate LLM-powered tech. We were using GPT 3.5 to do things like inbox summaries and translations and a few other bits and pieces. We launched all that pretty quickly and it got a rave review from the industry because it became this sort of used as the example of like, hey, maybe AI is actually going to help the incumbents more than it might help the entrants.

44:44And I think that the pendulum was swung in many directions, including this week alone. And then when we got access to GPT-4, that was the first point where I think Fergal had said, hey, because I don't remember when a launch was all about hallucinations. Everyone wanted to talk about hallucinations and guardrails and how you couldn't stop it from making shit up and all that sort of stuff. With GPT-4 and with better prompting and better prompting skill, we were able to pretty much get ourselves to a point where we were confident it would not hallucinate. So that was when Finn came to life. And it was very obvious once we could see that Finn could answer most questions, given any source material whatsoever.

45:22Given a document that roughly explains it, and given effectively the corpus of humanity, which is the other thing that all LLMs have, can you infer the answer for how do you reset your password? Yes, you can. But also, can you infer the answer for how do I merge two projects if one of them is archived in Asana or something like that. There is an answer to that, and it might not be perfectly documented, but it is there. You can find it in the tokens. There's a sentence over here and a sentence over there and something over here and some understanding of what the word merge can mean and all that sort of stuff.

46:00You'll put it all together and you'll find the answer. It very quickly became obvious this thing is going to do all customer support in the future. We don't know how long that will take. At the time, voice wasn't even a thing or whatever, but like it was very obvious that like this thing is going to do most support in the future and it's going to do it quicker and faster and cheaper than a human and the market wasn't ready for that message at the time honestly and the reason i say that is like everyone including us were probably probably limped in a little bit too much and sort of said hey this thing can help your support team in practice that's not what's going to happen what's going to happen is it's largely going to do a lot of the work that the support team would have previously done and it'll do that because again, it's just the end of business.

46:40It's better, faster, and cheaper. And then there'll be new roles created and all that sort of stuff, and we can talk about job displacement if we want. But this is what we were kind of wrestling with in 2023. We launched FIN in March 2023. When we launched it, it resolved on average only like 25 % of the conversations it saw. Today, that figure is 65%. And today, FIN's doing over a million resolutions a week. So now it's a big at-scale beast of a thing with thousands of customers. But it's very clearly, like now no one debates it anymore now it's like within the industry people say yeah obviously obviously you know the only question is like how many winning companies will there be here

47:16Des Traynor:so you so you feel that eventually something like finn will solve 100 of the problems or do you think that it will sort of tap out uh i think you said it's 60 at the moment would it be 70 80 90 i think it'll it'll vary business way but i don't know if it'll ever get to pure 100 because you could always make something up that requires new knowledge that doesn't actually previously exist and unless people want Finn to actually write support strategy, that might not happen. But if you take a very basic internet business, like a DTC product that only sells one thing, it's reasonable to assume it could do 100 % of that because it's just one type of question.

47:52There's like 10 questions that account for 99.999 and in any given period that might be 100. So I think that's one thing. It's different when you think about, say, what does customer service for like work day look like well they probably get like 20 000 questions a month but they probably fit into 963 different categories and the chances of you having enough documentation have gone deep on each of those 963 categories is low and then also on top of that layer on new product features getting shipped all the time and all that and all of a sudden the docs will go to date and as a result stuff will overflow and then there'll always be new scenarios that people hadn't considered.

48:28I think anything around the 90s is quite reasonable. I think 100 % is achievable for certain specific industries. But I think in general, getting to 90 is going to be good enough to change the entire organization.

48:45Des Traynor:Does Finn have an agentic angle? Because I guess with customer service, a big part of customer service is just finding information that's hidden from the customers, often in FAQs, often in their account details or whatever. But often the reason people phone a customer service line when they're really stuck is because there's some series of actions that that customer service person needs to do. So if I come and I'm stuck with, I'm using Intercom, I'm using Fin, can Fin actually transact? Can Fin actually look something up? Could it hit a button? Could it change a database listing? Could it go and look at three or four different products and kind of do actions?

49:28Des Traynor:Or is it purely just kind of gathering information? So we have this feature called procedures. It's like where you basically describe how to do something. And it can include API endpoints or it can include MCP servers, custom data lookups, anything you might need. So if you need to rebook a flight, if you need to check a package, if you need to change the address on something, if you need to route an order, check on the status of an order. Anything that a customer service rep can do, generally, Fin can do. And we literally built procedures for this exact thing. So the way we categorize support is there's informational queries, which is what you're describing when you said, hey, this is a question that is answered in text or whatever.

50:05There's what we call complex queries or procedures, which is like, hey, something has to change somewhere in order for this to be resolved. And then you've got other stuff like troubleshooting, which is another type of skill that Fin has, where somebody might give you a scenario like the light on my modem is red. And okay, you might have a series of steps, and it's ultimately kind of a workflow of triage that you want to go through. Check if this, is it plugged in? Is the wire connected? Is the wire torn? So again, it's a type of a procedure where you sort of say, if somebody is reporting a problem and they're not being super specific, here's the diagnosis steps.

50:44And so you need to do all that. But like, yeah, we build everything into Finn that we think a human does to do customer service and including updating its own knowledge and all that. Like Finn will see what the humans say and then adapt and react to that as well. So like it's like the whole product is like designed to be ultimately your customer support team.

51:02Des Traynor:I mean, you know, when you sort of explain how Finn can sort of think through problems, it reminded me I was having a lot of problems with like frames dropping using Elgato Cam Link. and I went to the Elgato website, couldn't find anything, went to all these forums of people. I ended up going to ChatGPT and it did exactly that. It walked me through all the troubleshooting. I told it my camera. It decided the right settings. And it was much more patient. If you're a customer service rep and you've been told that you have to answer every call in two minutes, ChatGPT kind of sat with me for two hours trying to figure this stuff out, uploading images.

51:37Des Traynor:And so I can definitely see how in many instances it can give you sort of one-to-one kind of a real sense that you're being kind of looked after. We still in our industry and in this society at large talk about AI like it's like subhuman as in like oh they don't have to send me over to AI but when you imagine I challenge you to imagine a better experience than you had with that Elgato scenario which I've been through as well you actually want somebody who's like no pressure like as in you know after breathing down the phone have you done that yet you get stressed or whatever it's actually exactly what you want is and uh um the behavioral traits that we see is when people use fin and it gives a good answer they actually come back for loads with loads of questions which actually as a product owner is exactly what you want because you want them to get the answers to all of you're like you're probably in a better elgato customer now than you were prior to that support experience uh because you've resolved your problem and now also you probably had you probably know a little bit more about how the tech works as well, right?

52:36And you're probably going to use fancier features and maybe you might even buy a new camera. And all that stuff is downstream of a good support experience. So I think there's one thing which is just understanding that AI is often just better full stop, no asterisk, no apology, not an alternative, just superior, superhuman, whatever way you want to say it, is interesting. The other aspect which we've been working on and we just announced a couple of weeks ago at our event is this idea of Finn as a customer agent, which is not just Finn doing customer service but actually fin as a concierge inside your product so imagine put on your product hat for a second and just imagine what it feels like to say hey and by the way inside this product there's going to be an ai agent who is going to guide somebody through everything and will answer every single question about how to do anything including going and doing the things if that's what they want them to do all of a sudden like you think about all your kind of core product goals as a pm you might be designing around you're like hang on a second this thing will actually be the solution it will improve all of our onboarding it will improve all of our adoption it'll improve our social sharing whatever it might be like uh and i think this is the way in which ai is going to dramatically change software is that we're going to realize that like the ui the skills that we have spent two decades acquiring might be best off summed up in a text box titled what are you changes, we still like, even like, even I myself, knowing all of this, I still trip myself up and be like, but we need the user to click the button.

54:06And like, you don't actually need the user to click the button. Like the button can click itself. That's not the thing. What you need is the user to want to do the thing or to know that the thing can be done. And once you realize that, like all of your like, how do you say, natural ways of thinking through UI design, they all kind of go away because you realize what we actually are is like uh you know in our world we're like primarily like what are existing workflows and processes and how do we apply ai to them to basically make them go away entirely and deliver the best possible experience for the end user on the other side of it and that's just it's a different muscle versus me and you who might have grown up like opening up omni-graffle or something like that and dragging boxes and arrows around like that's like the boxes and arrows are still relevant because the journey is relevant but the wireframes aren't you know it's just a different world i mean i think when you when

54:55Des Traynor:you interact with some kind of chatbot or or ai agent or whatever you want to call it like you're breaking a complex problem down into steps and actually the ordering of those steps when you're using voice can be very very different when you're using interface as a designer we almost force people through a particular flow like most most of the the uis we work follow a flow and that flow might work for 60 or 70 percent of the people it might not work for the other 30 And I definitely think we're getting into a world in the future where interfaces will recombine based on different users' preferences for how they think through problems.

55:32Des Traynor:And I think AI is the start of that. It's much easier to do in a chat interface. But you can definitely imagine how actual shopping interfaces, the cart could be reordered or the browsing process could be reordered. And I think we're moving into that really interesting world. And so I think it's fascinating. thing i'd love to just kind of move a little bit into one of the other things you do i mean obviously you know intercom is a big part of your life but you're also doing a bit of angel investing can you tell us a bit about that and what what kind of things you're investing in and what's exciting you about the investing space at the moment yeah i mean i started angel investing after like probably like whatever the first secondary we did in intercom was and i i felt like i always have felt like i have a good eye for a product you know like i can look at something and basically determine if I think these people understand the landscape.

56:16And I would say, like, genuinely, the first five to ten checks that I wrote, they were all basically successful, right? But again, this was the heyday of SaaS where you could actually spot a looker in the bunch pretty easily, right? I think then, like most people, I probably got too overconfident and probably overinvested in 2020, 2021. But, like, I would say at the time, like, the thing I'm a sucker for, and I'll tell you about the weakness of this in a second because it sounds like a tautology. I'm a sucker for basically a real product that solves a real problem that people really care about and really pay money for.

56:51Des Traynor:Which most people would say, isn't that every single startup out there? Every single startup would pretend or lie and say that they are that. Oftentimes the product doesn't exist. Oftentimes the problem it solves doesn't really exist. Oftentimes the problem exists, but no one cares. Or they care, but they're not going to spend money on it because it's not in anyone's budget. So I think whenever somebody can show me a paying happy customer who isn't their friend, that's usually an anomaly that is worth having a look at. But then separately, you can eyeball the product and work out, is it obvious what they're trying to do here?

57:23And you can make inferences from that. The downside of all that, what I didn't say there was in a big, large, total addressable market with no competition or any of that, I have had the experience where I invested in a good-looking real product for a real solution, blah, blah, blah, blah, blah, and then found out that it was just a better one. And I just didn't find it, and I'm not a VC. so I don't have the time to do a landscape map and to actually look at all the due due diligence and all that sort of stuff. So I've been caught by that before. So these days I try not to write checks off my own bat.

57:53I try to go in alongside somebody who I know would have had the time and ultimately wrote the check at a higher value such that they definitely would have done the homework. So that is a thing. These days though, because I think that the problem with a lot of these SaaS apps is they didn't get out in time so now a lot of them have to reinvent themselves like we did as an actual AI thing. so I don't know actually how the majority of that portfolio will work out I'm optimistic in a load of cases but then to some I'm like I don't know even though you have a brilliant product I just think it's like you have the best fax machine and unfortunately printers have been invented or whatever you know it's that type of thing so these days I'm basically looking at AI with the same criteria but I've added a few extra criteria like I sort of say like the four kind of investment criteria for AI again there has to be like real product adoptions people have to actually be using the thing there has to be deep ai so it can't be a thin wrapper um has to solve a real problem like as in something the business really cares about and then lastly concern i think people aren't keeping enough of an eye on is like is the margins or is the token burn uh like a problem or not because i think a lot of people are just going to like alakazam their way to like far better margins and I don't know if that's always going to be the case.

59:11If your product works by burning the haystack to find the needle and you're paying the same amount of cost for every piece of hay as you are for the needle, there are some cases where that just might never be profitable, in my opinion. Now, it depends on, obviously, you can solve it with price, but then when you bring the price up, does the whole value go away, right? So I've backed out of a few like that where I think, gosh, you're doing a shitload of work that's expensive to do to solve a problem that's actually not that valuable. And as a result, your margins are horrendous today, but they're also horrendous if the token price comes down by 10.

59:50Because just the ratio of burn to success, if you like, a wasted token to a successful token or whatever, is so off that I don't know if this equation, I don't know if the maths ever maths, as the kids would say. right yeah that like so that that's what i'm looking at right now but i will say i'm not doing a lot right now just because finn is like it's probably probably done maybe only like five checks in the last year uh just because it's not it's entirely because i'm trying to focus and i just don't have the time to review all these things also a lot of the ai startups are quite infra which means i have to go and talk to somebody to work out what the hell's going on and at that point what am i doing i have no expertise to bring to that company i should just stick the money in Bitcoin or an S &P or whatever and walk away.

1:00:34Yeah. Yeah.

1:00:35Des Traynor:Absolutely. I mean, I think the thing about kind of token economics, I think is really interesting. I mean, I've played a lot around kind of Gen AI design tools. And the frustration that I have is you start playing with these tools and almost immediately you run out of tokens. And so you haven't hit that aha moment. You haven't got any value out of the product. You don't know if it's any good. And immediately they're saying now, you know pay us you know 50 bucks a month or whatever like i was using chat gpt for months and months and months and months of getting value before i was like okay i feel now i'm ready to actually pay and obviously they've got a huge ton of vc money they can sit on and they've said that even their pro accounts don't make them any money because i presume all of their money is going into kind of customer acquisition costs but you know if you're if you're a gen ai tool doing some kind a lovable clone and people can't see the value because you you have two prompts worth of tokens and then it you know you just run out of tokens and you're never going to get that kind of aha moment so i think that's a real problem that people are struggling with and like you say even if even if it comes down 10x you know like 10 prompts or 20 prompts might not be enough to design a really really complicated kind of landing page let alone a complicated kind of you know workable application so i think it's it's a challenging space at the moment i think people talk about pricing a lot in this and like how do you price ai and we were famous for outcome based pricing but i think i think ultimately the way you're going to solve this is with pricing but then when you apply real pricing like chat gpt's pro accounts are what 200 a month or something like that that's when you start to realize oh oh shit this money is actually real like and and sometimes that is enough to put manners on people and make them realize they're moving out of hobbyist territory or out of prosumer territory into wow this thing better really deliver value which is a clarifying moment for a lot of founders.

1:02:20Des Traynor:Obviously, you come from a design background. You're a very successful design founder that's investing in small startups at the moment. I'm just kind of wondering, some of the audience here will be founders who are interested in design because they see some value it brings. Some of them will be VCs who also want to kind of lean into design, but some of them will be designers themselves who are curious about starting a business. I'm just kind of wondering if there's any particular advice you'd give to designers who want to become founders or founders that want to lean more into design. For designers who want to become founders, I think honestly obsessing about a problem a lot and thinking about it deeply and then playing with, like it's never been cheaper to explore the solutions based on it is right now.

1:02:58So if you really think that the way such and such a business problem or a consumer problem could be solved can be massively improved and you, in your mind's eye, have a solution, there's very little stopping you building a pretty credible, at least prototype, if not production-ready version of that. And I think, so whenever I hear somebody saying that they want to work on something but they don't have anything to show me. It's always a very bad smell these days because what it means is you couldn't even be bothered prompting it to get me to a clickable prototype to give me a sense of what it is you're doing.

1:03:36You'd rather jazz hands me with slides and a pitch deck or whatever. So I think the best thing I'd say to a designer who wants to become a founder is obsess about a problem until the solution begins to emerge. I think it was Einstein who said a problem well stated as a problem half solved or whatever. I think most of the time when people can't get the solution right, it's because they don't have the problem right. I think really, really refining that is really important and then demonstrating what you're going to build clearly and cleanly through a prototype. Honestly, coding is getting easier to do too.

1:04:09Emmet, who's our VP of design, is building real serious bits of software. It's a real thing. I'll come back to that in a second. I think that the biggest thing I'd say is obsess about the problem and start to play with the solutions. That's the best way you can refine. everything else like product market fit and markets and raising and all that they're all problems that take care of themselves once you actually have a solution to a problem that people actually care about and it's a real problem. On the sort of founders who might want to lean more into design we did just publish on our ideas blog which is ideas.fin.ai about the changes we've been making in design so specifically I think one of the reasons it's more justifiable now to invest harder in design than it has been before because I think you can ask your designers to own the entire front end of the product you can ask your designers to vibe code the entire roadmap so that you can show your customers all the stuff you're going to build.

1:04:55It's more possible now to have your designers not file issues with Figma screenshots versus UI, but instead actually credibly just go and fix the thing by asking Cursor or Windsor for whatever to do it. So I think the ROI of the AI-enabled designer is substantially higher, such that it's definitely worth leaning into it. that's my main thought the only other thing I'd say is design is changing there's no way around that and that sounds like some thing that would have been true in 2007 as well but I think the UI of AI is still being evolved and be careful not to hire a create, retrieve update, delete old school designer who hasn't adapted your skill set because design ain't what it used to be understanding prompting and understanding how to design solutions using LLMs is an entire new skill that you're going to need a lot of.

1:05:53And there is an interaction between how the AI works and what the UI needs to be. I'll give you a very simple example. It would be like by playing with the AI, you work out what is the minimum amount of input you need from the customer to deliver the solution. You need to keep playing with that until you get that as small as possible. And then you need to go and update the UI to solicit that input from the user. and then that becomes itself a kind of a feedback loop where how you actually explain these things really really really matters so that you get the right stuff from the user so you can feed the prompt correctly and we call that role an AI designer but design is changing and that's just one example so I think just be careful you don't go and hire somebody who doesn't acknowledge that design has massively changed in the last two years.

1:06:37Des Traynor:I mean this is sort of slightly self-serving but one of the things I've always loved about our interactions and one of the reasons I follow a lot of your stuff is I feel that a lot of my thinking and your thinking is kind of like sort of in sync whenever you say something I'm like yes absolutely that's that's what I've been seeing as well I really love the way you're framing it and often it's counterintuitive not counterintuitive but often the stuff that you say and some of the rest of your team can sort of put the sort of the you know the ruffle the feathers of the kind of you know some of the design and product and tech community but often what I found is like six months later a year later your assumptions have been kind of played out and have found to be true.

1:07:16Des Traynor:So I think it's always worth people kind of listening to what you've got to say because they might not feel it now, but they'll probably actually come to agree in six or nine months' time. So where can people follow you if they want to find out more about what you're doing and what you're thinking at the moment? Yeah, so I'm just Des Trainers. It's just my name on X or on LinkedIn for that matter. And then all of our latest, I'll share everything on either of those accounts, but all of our latest thinking and ideas are at two different places. If you're really into AI, go to fin.ai slash research.

1:07:46That's where our AI group publish their blog. If you're more into the sort of product implications of AI, go to ideas.fin.ai and that's where we're publishing most. The Intercom blog is obviously still alive and well, but when we're going all in on AI, we're keeping it under the FIN moniker.

1:08:01Des Traynor:Brilliant. Well, look, Des, thank you so much for joining me. That was brilliant. I had a wonderful conversation, so cheers. Cool. Cheers, Andy. And thanks to everyone listening. Thanks for tuning into The Design VC. If you enjoyed this episode, please subscribe, leave a review, share with someone that cares about the intersection of design, startup and venture. And I've been your host, Andy Budd. Till next time, goodbye.