Speaker unverified: belief

Transcript context

…nce is the reason why we're not a hundred percent… It strikes me that a lot of the, how you include the account context and the amount of account context you include is a big part of the secret sauce. Perhaps, but it varies customer to customer. It really is one of these areas where it's a thousand lead bullets. It's not like a single silver one. It's not, if you look at a resolutionary graph, there’s no pop, give or take one or two little tweaks. It's mostly just, “Hey, we ground out through optimizing this prompt and changing this handover, we ground out another 0.7%,” and you see the AI team celebrate that on the balcony on the Friday being like, yay, 0.7 up, whatever. It's hard to work out exactly what bits and then there's obviously multiplicative benefits. You might have a win over here, and it cost you something over here as well. How deep down the stack will you go Intercom's version of Apple Silicon? That I don't dunno for sure. We're going to chase any edge we can get. Right now I think custom models is definitely where we're going, and that's a large investment from the AI group, which is the most contested resource that we have. Every bit of work they're doing is finding a new edge in resolution rate or resolution quality. So right now it's paying out pretty well. So I think we're going to kind of place all our chips there until something changes. What does selling AI look like? It's quite difficult. It's difficult in marketing and selling because I think… … our cloud is better than their cloud. Yeah, and our performance criteria are better. It's like at times it might feel like Intel/AMD, or at times it's like it's our response times versus theirs or whatever, but you're ultimately coming into it with a battle of metrics and stuff, like our resolution rate and our CSAT versus theirs. But then people say, well, why? And then you have to then explain what's actually happening beneath the surface a little bit so that people can actually get a bit of conviction other than just trust us or please just go and try our product. Not that easy to try Fin. You have to still have to turn a lot of keys and open a lot of APIs and stuff. So the challenge genuinely becomes how do you have a sales team that's actually able to speak with a good degree of familiarity about AI? l have to turn a lot of keys and open a lot of APIs and stuff. So the challenge genuinely becomes how do you have a sales team that's actually able to speak with a good degree of familiarity about AI? It's funny you mentioned this. We have this specific problem at Stripe, which is invariably when people switch to Stripe from a legacy processor, they see a significant revenue uplift and you think businesses are in the business of finding ways to get more revenue. You’d think they'd be really interested in this. And we have this thing that sounds shockingly good, which is if you just move over to Stripe, you start immediately getting more revenue. And it basically comes from two places. One is conversion on the actual point of payment—that if your mobile app or if your web flow is kind of janky or doesn't offer the customer's preferred payment method or something, they will abandon. And if you just look at the abandonment rates, if you're seeing a only 85% conversion rate on that form, then obviously getting it up to 90%, that's a huge deal. And those would be very high numbers. Most businesses would not see anything close to a 90% conversion rate on that form. And so there's huge improvement possible there to make the customer checkout experience as smooth as possible. And obviously things like Link then where you're not asking people to reenter their payment details, that delivers a big uplift. …hard to believe. It's undifferentiated as a message. I remember even when we switched back to Stripe, either you or Patrick was saying, “Oh, well don't forget to do Link.” And I was like, really? I was like, is this really a thing that people have? Some business has forgotten its credit card or something like that and you're going to be able to renew it. It feels unlikely. It just feels implausible, but at the same time, the data is not really debatable. I think people like to be able to explain it to themselves, and I was talking to a guy who runs procurement and he was saying, “You have to understand I'm just getting bullshitted to all day. It's just all day relentless, absolute nonsense in my face. So if you think that you're like, ‘Ooh, 65% thing is going to stick,’ it's not. I divide it by 10 at this stage and you tell me, ‘Oh, you're going to save me 2 million in CS salaries,’ or whatever. I'm like, yeah, maybe in three years time we'll see 200 grand.” That's the kind of default posture for a lot of these people. I think it is just, it's like they've developed quite an adverse reaction to marketing. People do not have enough empathy for the procurement person who just has to endure nonstop nonsense… Absolute garbage. That's funny. I mean this kind of gets to a topic you and I have discussed a lot, which is product marketing. How do you effectively product-market in a world of everyone making claims? One is the guarantee your guys' million dollar guarantee—has that worked? Certainly it's worked from a point of view of, I dunno actually know how many people are in the program right now, but I could say that what has worked is… It landed the message of, “We stand behind the product.” 's worked from a point of view of, I dunno actually know how many people are in the program right now, but I could say that what has worked is… It landed the message of, “We stand behind the product.” Being able to say, here is the reason why you can bank on us. I think that's a strong message. I mean obviously being able to point to real customers or real results and you can say, “Hey, go talk to Natalie at Nuuly,” or whatever company you want. Go talk to that person and ask them, that's their name, that's their job title. They work there, they're saying this. With 6,000 customers, it gets more believable as the numbers kind of tick up. So I guess either you can make crazy guarantee claims, you can just point to a lot of successful customers. For us, it might be different depending on your domain, but for us it's not like we can show you—we can show you, hey, here's a beautiful backend product, here's fancy reporting and all that, but that doesn't speak to—courting someone's buying, when they're buying AI off you is to some degree a replacement of work that they have to do. And the two things they care about are how much work are you going to do for me and how well are you going to do that work? And you basically need to product-market both of those things. And it's very easy to say, we're going to do all the work and we're doing really well. So you have to actually really help them understand how to appraise this scenario. Sometimes we put time into actually helping people identify when they're being lied to in a sense. We will say, like, “Hey, try this type of question” or “Ask them about this help.” You almost teach 'em to be much more conscientious buyers. Yeah, I can see that. What are your other pet peeves when it comes to product marketing? Actually, do you want another pint? Yeah, sure, sure, sure. See again this one, did I not let it settle for long enough? And then you have the small head? Is this settling?…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.