Ep # 160: Rethinking GTM: Aligning Every Team Around the Customer ft. Paul Staelin (EX CCO, Vercel)

Paul Staelin 00:02 The goal of the product ultimately is not to sell it. And I know as a startup that's what it feels like it is for the customers to succeed. 'cause if the customers succeed and stick, you're gonna have a, you're gonna have a vibrant business over time. Those customers will be your greatest advocates. They're gonna be your success stories that help you sell to other customers, and they're gonna help you improve your product. Taylor Kennerson 00:22 Welcome to the Hyperengage Podcast. We are so happy to have you. Along our journey here, we uncover bits of knowledge from some of the greatest minds in tech. We unearthed the hows, whys, and whats that drive the tech of today. Welcome to the movement. Adil Saleh 00:43 Hey, greetings, everybody, this is Adil, The Hyperengage Podcast. This is a 160 episode, and this has been the most special episode for our entire team, entire crew, the community. We already communicated about this episode, before time. There's so much that we have explored at scale with companies. We talk about customer success. We talk about go to market. How is sharing customer acquisition has seen, raising fund has been the big challenge in the recent times. And there's so many of different opinions and notion that came up from different leaders. We explored more in kind of a marketing tool, go to market tooling side of things like sales, support, customer success. Today we are trying to do a different approach, like how these developer tools that are doing it scale when it comes to customer success. How they're helping customers perceive value at scale has been the biggest concern from us because at the end of the day, it's core technologies like these that are actually shaping the entire categories across different industries, sectors, and how people are building apps on top of them. Like a lot of these platforms are being used as an open source model, as a top-bottom layer model that people build on top. Today we're talking about Vercel and we are talking about Paul, who's the CCO at Vercel close to nine billion in valuation. We are gonna be talking about Paul, the guy who just started as an EV startup 25 years back, and now being a CCO at Vercel. Actually changing the way and making impact the way customer success operates at scale, to masses, is big. Thank you very much, Paul, for taking the time for this episode. Paul Staelin 2:32 Yeah I'm very happy to. Thank you for having me back. Adil Saleh 2:36 Love that. I remember that during last time we had you it was just 25 episodes and I was just searching up a four-year role, the company and all of that. I was like, Hey, boom. Like this is the guy. I'm gonna learn a lot in this episode. I had that, even then, and you are such a big educator that two years down with 160 episodes, I still — it's my honor to say the same, that I'm here to learn from you and all the teachings that you have and the kind of impact that you have. So whatever you can share today in this episode would be a pleasure to see here for all of our audiences. Paul Staelin 3:11 You're very kind to say so and I hope I don't blow it. Adil Saleh 3:16 No, not a problem. So now thinking about you starting up, I know that things have changed in terms of technological change at a massive level. Companies pivoting and even the vision has been pivoted. A lot of big companies, big names, that we've seen in the recent time. How do you feel as a customer success category? I know that Vercel is pretty much very strong in the roots when it comes to one of the first movers. So how do you see it as a category of CCO expanding the top enterprise customer segment sitting here, like what kind of notions you have sitting here two and a half, three years down. And what is the future that you're seeing in terms of potential in the segment? Paul Staelin 4:03 Broadly defined the developer tool segment. Like anything, people buy a technology to enable them to do things they couldn't before or to enable them to do things better. And even if they know everything about your product on the day they sign and start — which they don’t — your roadmap ensures that won’t last long. So the primary role for the customer success function specifically, and we can talk about support and other pieces as well, I believe is to keep tabs on the goals that the customer has. 'Cause they also evolve — businesses change, grow, leadership changes, priorities change, economic climate changes. Keeping your view of their goals up to date, making sure they know what their goals are. Part of this is just asking a question. If they don't have an answer, then great — hopefully they can go get it both for their benefit and yours. And making sure that you can map what they're trying to accomplish to your ever-evolving roadmap so they can continue to make progress. They've gotta evolve and change as they adopt different portions. And the next step will be different than what the last step was. So being that kind of continual guide and advisor on the journey to make sure that the steps go well each step along the way, and to make sure that ultimately their ever-evolving goals are still being met, is a way to make sure that your customers continue to succeed. And a way to make sure that your relationship stays solid and hopefully over time grows and expands as they start to use you in more places. Adil Saleh 5:38 Absolutely. And it is, like you mentioned, super important to evolve with the customer goals for the longest periods of times. And at the same time having the product sticky enough for them. Sticky meaning like consistently delivering value. Vercel has been one — when you came last on, that was the first time somebody on my team actually showed me the dashboard that they have, like they were using the platform and how, like I wanted to understand how this actually operates internally for front-end developers. So now, thinking about huge running. I know that there are different initiatives you take at the global level. Thinking about customer success — what kind of AI initiatives have you taken ever since this new evolution that came 18 to 20 months back, and a lot of these customer success engineers thinking that, hey, we may be getting automated a lot, but at the same time we can personalize. How do you find the right balance across the entire organization of customer success as a CCO? Now you have a bigger ladder to follow. Like of course the sales and the marketing and support, they have their own challenges. Paul Staelin 6:45 Yeah, so AI is interesting 'cause Vercel not only provides a developer platform, but an AI development platform as well. AI has probably transformed our customer success journey most explicitly by ensuring that we have to talk to our customers about how they're using AI or how they want to use AI as a business, and how they can use our platform to help them do so more effectively. With the AI SDK, the AI gateway, there are a bunch of things that we've put in place to make it easy to build AI solutions within a company. To have you built an MCP server so that you can actually service other agents who wanna get data from your app or your website. What have you done there? Do you want your own agents? So when people visit your site, do you want them to be able to guide your customers — the agents, that is — through your experience and make sure that it goes well? So AI has transformed what we do rather explicitly in terms of accelerating those tasks, which is what AI fundamentally does — it's a way to lever up your resources. We have been working pretty heavily, and we're not as far along as I would like on this particular talk. We've been working pretty heavily on a health score. We've got some evidence that the health score is in fact predictive of future churn. We're starting to track the CSM activities. When this happens, we do this and it works or doesn't. And then we can start to have the AI look at, okay, if this is starting to trend or if this has happened, what is it? Do we have the right flags? What are the real risk flags? And of the actions that we have taken, which one do we think is the most likely to work in this case? So we're still in the process of building out the dataset. The AI is only ever as good as the data it has to train on. So we're unfortunately early days in generating the dataset that the AI will train on. We are using it to automate tasks — QBR decks, EBR decks. There's a big difference between the two. Being able to kind of auto-generate more quickly an account summary and other things. Adil Saleh 8:50 Customer research and all. Paul Staelin 8:53 Yeah, there are routine tasks that we're starting to automate with specialized agents in our case to help make some of the heavy lifting a little lighter. Adil Saleh 9:07 Interesting. I'm glad you brought this — the specialized factor of AI capabilities. I know that it's basically all possible, you can still do it, but when you talk about customer health score, when you talk about different journeys across different segments, within those segments different products, of course customer success could be different for one customer and another. So now, how you are training — you mentioned about fine-tuning. Are you making sure that they're within those segments? And it is pretty much specific to that kind of segment when it comes to data, information patterns, all of that. You mentioned that you're going to forecast and all kinds of predictive analysis that are actually hitting on the customer score in a good or bad way. Paul Staelin 9:51 Yeah, so there's a whole bunch of interesting things there in terms of segmentation. There are a whole ton of different ways you can do this. The way that the data changes most internally for us is by use case, right? If you're using us to host a marketing site, your data looks one way. If you're using us to host an app, you're a startup and you've got all your users running through a Next.js-fueled application — awesome. But that looks very different. If you're running an e-commerce site, that looks very different. So for us it's far more — use case is the biggest vector. Then maybe industry and other things. But use case really drives the type of data we're looking for. And at many customers — lots of customers — we have multiple use cases. So even within, we're trying to figure out how do we do this to help score a customer if they've got two use cases and one is wildly green and the other is not. So I think we're moving towards the use case health score. Adil Saleh 11:03 Yes. Yes. Paul Staelin 11:05 But again, this is all — we're still learning as we go. This AI is early for everybody. No one's got it licked. And what AI teaches you about your own business will be very different than what AI is teaching us about our business. But this is one of the things that I think is a key learning. Adil Saleh 11:23 Okay. Because a lot of these actions, if you had around scoring — predictive scoring, health scoring — a lot of these leaders, they come up and say, Hey, it is good enough, but it's not the actual score, it's a myth. Like you talk about health score — it has more to do with the qualitative data analysis. People go meet via Zoom, like qualitative data points. So how are you incorporating all those factors that are externally just quantitative data points, like product usage and consumption and events and all of those? Paul Staelin 11:58 Yeah. I'm a big believer of a slightly modified DEAR framework, which was popularized by Gainsight — with Deployment, Engagement, Adoption, and ROI. We added Sentiment, which makes it DEARS (plural). And sentiment is just a plus or minus score 'cause you can be unsuccessful with your deployment and wildly happy as a customer — I've seen it happen. That is an unstable state. Now, it generally means they're willing to work with you so they're healthier than someone who is both not that successful in their deployment and unhappy with you. So it's just an additive or subtractive factor, and that is based on the read of the conversations. We actually have AI now reading through the transcripts of Zoom calls, so we can get summarized sentiment and other things. A CSM can override or overtake it — 'cause AI is not always right. And the more data you give it when it's right and when it's wrong, the better it'll get over time. But that's fundamentally how we just incorporate it — it's just one of the factors. It is not the factor. It is not ignored. It is one of the many things, one of the many ingredients in the soup. Adil Saleh 13:21 Interesting. Love it. I know that when once we had you last time, two years back, you were just starting out, so we didn't have a lot of conversation around Vercel and the CS org you have there. Now, we are in a better place. Thinking about your team — what kind of formation do you have across? I know there might be solution engineers, implementation managers, account managers for accounts with large lifetime value. So what is the formation around the customer success org and what kind of operating principles do you have? I know playbooks are pretty mainstream. You already explained that you're using AI a lot for basic tasks. Now, building up on fine-tuning your own agents that we use internally. Like many companies, they're all like a long launch group of agents, a crew of agents, in a few months. So now thinking about your organization, how you call it — you call it more towards a hybrid touch, all of your customers feel equally supported and engaged, like communication engagements. So how do you rank yourself and what kind of initiatives have you taken in recent times? Paul Staelin 14:29 Yeah, so it is interesting — not having been in a company that had a really vibrant PLG motion before. So this is one of the big things that was different for me. 'Cause when the PLG motion works, they could have played with our $20/month Pro version, right? So a couple hundred bucks a year, they could have played with it as part of a POC. They had one developer go play with this thing. It could be a startup. And it is their thing. And they're already live and they're already in production, and they already know what you do and how to use you. In the early days before I arrived, so many people were PLG motion. The CSM motion was exceedingly light 'cause they were startups in general who were growing into the enterprise-sized pants. And they just needed a welcome call and a couple of things like, Here's what's new in enterprise that you didn’t have before in Pro. It was a light introduction and then they would wait for the questions to come in 'cause the startups would proceed at jagged rates depending on what they were focused on. Startups go day to day, right? They were starting to see bigger companies come in who had used Pro as a POC, and now they were gonna move over to the real thing once they bought enterprise. And this is now a migration. They aren't live in your product and only one person has played with it, and they've got big business goals. They've got a working whatever that has to make the transition seamlessly. And it's a very different motion. As the company’s been moving upmarket, that was one of the major challenges. When I arrived, the GDR for the enterprise customers was below 80%. It was bad because we were using the light guided tour for everyone. And the migration customers were really suffering. So the first thing we had to do was start to acknowledge: okay, how did they get here? And they may not have even played with Pro, by the way. They could have just showed up blind. Then the guided tour welcome call did no good whatsoever other than confuse them. Building out a motion where we went into the welcome motion, could determine which path the customer was on, and then make sure we're having the right engagement until they can figure out what their go-live date is, if they have one, what the deliverable is, exactly who’s working on it, who can we reach out to. How do we make sure that this is going well? Setting up regular touchpoints so that we can make sure if any problems arise, they don't spin for a week before we talk. Setting up Slack channels to communicate, making it easier and helping guide them to the first win. 'Cause the first win is always the most important. If the first win goes poorly, you're DOA for the rest of the relationship. So investing in a more varied onboarding experience depending on the customer's needs. And that meant — and we had at the time — CSMs all had equal books. There was one coverage model: if you bought Enterprise, you got the one CSM experience, which was again, very light and very technical. Once we created those different journeys and started to learn how to actually do this, and then actually maintain contact once they got through that first go-live, so you have intensive coverage and engagement through the first go-live, then you need some regular engagement. Because those developers are being paid every day to build stuff. They're not being paid to just sit on their hands once the first thing is shipped. They're building new stuff, figuring out what the new goals are, what the new deliverables are, making sure that they understand, oh, if you're doing that, this part of the product that you didn't use in the last one — this would be awesome for you, and here's why, and here’s where you can go learn more about it. Those types of conversations. So setting up that kind of project management for the initial go-live, setting up program management so you can help a customer manage multiple projects, and starting to establish more of the account management — where you're actually talking to the champion, but also periodically getting to the exec sponsor so you can really make sure that you're aware if the tiller on that boat is gonna shift or change. So building out all those motions — again, one coverage model. So this was a pretty busy 18 months building this out. GDR went from low seventies up to high eighties. We actually put in some differentiation. Actually, it ended up getting up to 90 for the last three quarters. But in the timeframe I'm talking, it gets to the high eighties. We’re like, okay, now we're good enough, we want to differentiate. Because we can invest more in the bigger customers and we need to be a little more fiscally responsible about how we're investing in the smaller customers. We love all the customers equally, but we can only invest in a responsible way in each of these relationships. So we brought in someone to lead the digital effort and have been building out a pretty real digital motion. The first digital CSMs came on board about a year ago. And that motion is getting better. When we looked at the smaller customers — where the CSMs who had to cover everyone equally, of course, devoted the least love and attention to the littler accounts — the GDR of that slice wasn’t great. By dedicating the digital team for that slice, that GDR has actually started to climb as well. But we’ve had about a year, and that motion for the digital echoes that of the higher touch model. It rhymes, but it’s a little lighter, a little more reactive. You’re spending more time looking at health scores. This is why we developed the health score in the way that we did — to figure out which customers you should be talking to. There are still a couple of touch points a year. It’s light, but it’s not nothing. And in fact, for many of the customers in this segment, if they weren’t reaching out to their in-theory, uniform CSM motion, they were getting none. Because the customers that were reaching out were consuming the CSM entirely. So we’ve split into two levels of contact. I would expect we’re ready now, I think, to split into three. We’ve gotten enough very large customers that having a real high-touch motion where you’ve got 10 to 15 accounts, a mid-touch motion where you’ve got 30 to 40, and then our digital motion, where you’ve got 100–125, is the next hurdle for us on this. And ideally, we keep GDR at that 90%. Adil Saleh 20:51 Yes. Perfect. Listen, Paul, it was such a treat to listen to all of this because I was just reading a Bloomberg article this last week. You guys raised $9 billion of valuation — that is tripled since the last time, almost tripled. So one of the factors could be this, right? Doing all the transformation in the CS org because it’s more about expanding the revenue. Retaining is good, it’s been there for years, but it’s more about the customer satisfaction, having a forecast, having all the visibility into these accounts that are low-hanging fruit and they’re growing. Like in your use case, Vercel — if I’m a startup today, 2025, I’m using 30–40 different projects. Four years back? Who knows. So it’s all about growing with the customers, like a Notion approach — the way they built it from the beginning. So what’s your viewpoint? Is it fair to say that it’s been one of the initiatives taken in the customer success org, targeting towards the net revenue retention and expanding the revenue? Paul Staelin 21:53 Yes. So we do have sales reps who are focused on expanding existing accounts. So the CSM’s job — and I’m gonna use a soccer analogy here, or football depending on where you sit in the world — we want the AEs in these accounts to function as the strikers. Which is: when there’s a real opportunity to expand the account, you want someone to come in who is not explicitly the customer’s buddy that they talk to all the time, but is someone new and different. They can negotiate a little bit better and spin and sell a little bit better. Adil Saleh 22:33 Yes. Paul Staelin 22:33 That person, that role— Adil Saleh 22:38 The forward door. Paul Staelin 22:39 Striker, winger, one of those roles. Adil Saleh 22:40 Yes. Paul Staelin 22:42 The CSM really spends a lot of time — a lot of time, and depending on the account, a lot of time — in the midfield trying to build the motion to feed the striker. But you only feed the striker a few times a game. There’s a lot of work in the midfield to develop the opportunities, to build the trust, to do the things. And there is sometimes time spent on defense — if customers go awry, or the champion leaves, or someone gets unhappy, or they have an outage for whatever stupid reason. And you’ve gotta play some defense too. So the CSM is far more midfield and defense, and the sales reps are far more the strikers for NDR in the enterprise customers. That was the explicit charter of the CSM org when I landed. When I got to Vercel, NDR was over 100 but not by a lot. And we peaked right before we cut our prices at just under 150 — so 149% NDR. It drifted down because we did lower our prices so the usage revenue dropped. But hopefully, as people start to buy more of it, now that it’s cheaper to do things on Vercel, those numbers will recover. But yes — 45% gain in NDR and 18% gain on GDR. There are more customers who will buy more, and by the way, they’re happier, so they’re more likely to buy more. So you win. Adil Saleh 23:59 Yes. That shows that there’s always gonna be a bigger addressable market. So you just need to make sure that you have the systems in place to serve them at scale and deliver value consistently. Perfect. So now, thinking about a lot of these founders — they might be looking at this episode once it’s out, and they’re gonna listen because a lot of them are using Vercel. And customer success at Vercel is just being offered this longer article, like: How did they kill this? How can they claim this valuation and raise funds like this in such a quick succession? And compared to times when it’s not easy to raise funds. I know that starting off it was pretty much peaked, and now investors are scratching their heads most of the time, thinking: Hey, we spent on this AI-powered tool, and now there are tens of them alongside in the market. So now, thinking about you as a leader: what do you suggest to somebody building — from a business standpoint, from an entrepreneurial standpoint — building a startup, thinking about customer success in the first two years? A lot of these folks think that getting more logos and more customers is good, but how do you see enabling your teams — even if there are three or four people in the team, a small startup trying to deliver value and build a customer success function — what kind of advice do you have for them? Paul Staelin 25:27 Yeah, there’s an interesting approach fundamentally — and I’m gonna go back to the birthdays. Because when I landed at Vercel, we already had 12 to 15 folks-ish in CSM. Your first CSM — the very first one you should hire — is the second your first customer is actually using the software. Unless you, as the founder, want to be the CSM, which is a perfectly fine solution. I know some people have three people in their company and you can’t afford a CSM. Fine — you are the CSM. Having those regular touch points with the customers, having the time in your calendar where you can devote and look to see how the customer is using the product. And by the way, if you can’t see — that’s another indication you should probably be investing in that part of the roadmap too. The goal of the product ultimately is not to sell it. I know as a startup that’s what it feels like. It is for the customers to succeed. Because if the customers succeed and stick, you’re gonna have a vibrant business over time. Those customers will be your greatest advocates. They’re gonna be your success stories that help you sell to other customers. They’re gonna help you improve your product. Everything flows from understanding how and why your customers are using your product and how and why they’re succeeding or failing or blocked. And it needs to be a very real part of someone’s job description the second a customer’s using it. When you’ve got one CSM, you have to have someone who’s a builder, who’s willing to complain. And I know that sounds like a bad thing, but it’s a great thing. And you have to be willing to listen to their feedback and actually take it into account. As one of the leaders of an organization, it’s very tempting to prioritize the needs of your sales team to sell a deal over the needs of your success team to make that customer who just bought successful. I assure you the better. Most people are 80% sales, 20% CSM. You need to be at least 50/50, if not 40/60, if you want to make this a vibrant and longstanding company. So: play the role yourself, make sure you’re taking that feedback, prioritizing it appropriately in the roadmap. Find someone who’s gonna be vocal, who can be all over those customers and is very proactive. And then expand the team from there. Once you’ve got two or three people, you need a real leader who’s done this before and can start to recognize the patterns and build out the plays and do all the things. Because if you get two or three peers, it’s all gonna be a little chaotic. They’ll all try different stuff. Maybe they try the same thing, maybe they don’t — which is great. Good to have diversity of experiments. But then someone needs to come in and start saying: Okay, this play seems to be working really well here. This one seems to be working really well here. Let’s run those two plays. We’ll let the chaos reign everywhere else, but we’ll run these two plays in these two places. And then a month later: did it work? Did it not work? How’s it going? Do we need to make it better? Are there other places we can start systematizing? And you just need that kind of constant continuous improvement. Adil Saleh 28:25 Absolutely. You need this symmetry. And folks, this is coming from a guy that’s been doing this since times when we were at high school, skipping classes. This is coming from Paul — and thank you very much for all of this. Symmetry is super important and consistency, repeatability is super important. Even for the mistakes that you do — you don’t have to repeatedly not do it. Or you learn from people that do it. You don’t live long enough to do it yourself all the time. So now, thinking about your team and culture — this notion has been growing so fast this past year. A lot of GTM leaders, even founders, are hiring CSMs, initial sales execs, and all. And they’re thinking: with this AI, with these tools, with these capabilities, we don’t need all those smarter CSMs or red-players as we would need in the past. What’s your viewpoint on this? I’ll keep it open for you. Paul Staelin 29:21 Yeah. Ultimately, and I'm gonna switch sports to baseball here. We are in the top of the first inning with AI. No one has this licked, we haven't figured out exactly how we're gonna use it. The thing that is fabulous about it, and the reason why people are so excited about it, is that it can take a lot of the really mundane, repetitive tasks. So let's say I'm a CSM and I'm gonna be preparing for an EBR, which is an executive business review, and I'm gonna have to partner with the champion on the other side 'cause the champion and I effectively are gonna present to the sponsor together. But gathering all the data, usage data, the timeline of what happened, summarizing the events of the past, summarizing their goals as we currently understand them, doing all those things so I can build that kind of starting point of a presentation and then collaborate with the champion on it. It used to be half a day and now you can be like, go — five minutes later, you've got your deck, right? So I just saved half a day, which is incredibly potent and powerful. I would just — and again, it's only gonna work if you spend the time to say, this was good, this was bad, and all the things. So you have to make sure that you're looking to close the loop for every AI agent you want to use, but it's incredibly liberating. Yes, ultimately, if each CSM can go from covering 30 accounts and doing three EBRs a quarter, they can cover 40 accounts and do five EBRs a quarter. You will need fewer of them. But it doesn't mean you don't need them, and it doesn't mean you can give one person 185 accounts. This is incremental improvements over time and it's all on the success or failure of every AI initiative, which is going to be on your organization's ability and willingness to close the loop on what the AI did — was it good or bad. Adil Saleh 31:16 Yes. Paul Staelin 31:16 So AI, continue— Adil Saleh 31:16 Knowing the outcome. Yeah, knowing the outcome is super important because you still need to be equally smart as a CSM, like back in the years, to be able to identify: Hey these are the outcomes that I need. This is the EBR. These are the QBR surveys that I need. This is the follow-up meeting notes that I have. Okay, these are intelligent. This actually lines with our vision, goals, customer goals, all of this. So now thinking about optimizing the bandwidth with this AI — a lot of businesses are only thinking about this, and it is about this. If AI is not able to help you move fast and optimize your bandwidth, like this is one of the biggest. So what kind of initiatives have you taken to optimize the bandwidth? On this, on the entire GTM or organizations like Sales, Marketing, Success in the past, I would say a year. I think you already started that, right? Paul Staelin 32:10 Yeah we are very AI-forward 'cause if we're gonna actually encourage people to use our platform to develop some of these AI solutions, we better be doing them ourselves. So we are leaning pretty heavily into this. In my organization, probably the most dramatic use case for it actually, and the very first use case to use AI in a very real way with customers, was actually putting in kind of a chatbot in front of our support process. Because we have millions of free customers, a hundred thousand pro customers and over a thousand enterprise customers. There are a lot of cases 'cause even the free and the pro customers can file cases and a lot of 'em are simple, right? New people haven't had a ton of help yet. They're learning, they ask the same question over and over again. We put an AI chat agent in front and it was starting by deflecting, preventing — people were able to get their answer right away with the AI chat, now 30% of the time, which is a massive improvement. Adil Saleh 33:24 30% is massive. Paul Staelin 33:26 Oh, it's so — not only did we save 30% of cases, now they're the easy ones. So it was like 10% of time. But 30% of cases. And then we started to see the places where it was failing and customers were opting out, and started fixing them and teaching the AI how to answer those questions and those questions. And it's a slow, painful, incremental slog. And the people in the support team didn't like closing the loop on these things, but they had to do five a week, right? Everybody. And the dataset built up and suddenly — not suddenly, by the end of the year, we look up and it's actually obviating almost 70% of inquiries. So again, it's the easiest 70% of cases, so it's not 70% of time — it's probably 30% of time. But you give an engineer 30% of time, that's awesome. And it means that the support engineers with whom I've had the pleasure to work, they love the hard ones, 'cause they get it and they get to just zone out and dig in and rip through logs and pull on threads and it's fun. You get to spend two hours delving into this thing and hopefully you figure out what's going on and you can really help the customer out. You do not like the three-minute cases where it's the same answer every time. It feels — it's not satisfying. So we had a pretty big win in terms of the customers, 'cause 70% of inquiries get answered immediately, which is awesome. That's the best outcome you can have as a customer. Saved us about 30% of workload and the workload that was left was more fun and interesting for the support engineers. So it's been a massive win. But without the five cases every week that came in that had opted out, without that five cases per support engineer per week — which there was grumbling about — that doesn't happen. It really takes that kind of commitment and just daily focus on, I'm gonna set aside this amount of time, I'm gonna make sure I'm teaching the AI how to do this right. And I'm gonna continue to do that. And you can see tremendous results. But don't just launch an AI thing and walk away. Do not be a Bond villain. Launch the AI thing and then walk away and assume it all goes to plan. This is a bad idea. Adil Saleh 35:45 Yeah. Not in even Elon Musk’s vision, AI does it at this time like this? So great. So now, just closing on the list, because since I have you here and I want to get the most out of you, like for our audiences. So now what is that one thing that makes you excited and what is it one thing that you advise to anyone listening to this? That anyone could be — someone who's afraid of getting laid off, someone who's not doing so good at a startup, thinking about a lot of companies having a lot of funds, more funds, they can beat the shit out of him. A lot of these folks that are still in school, trying to think what approach they should do — do we even need a degree, or can we just have e-learning or something, we can go and listen to OpenAI's videos and courses, or something? Somebody that is like myself, in the middle of building a startup, having a handful of customers and thinking about like where should we do a pivot, or should we get a little more customers, get more validation, or maybe expand more, or go upmarket, or something. So this is an extremely important opportunity to hear from you. Paul Staelin 36:58 Yeah, so there's a lot of questions baked in there. I'll try to pick off a few, and I know we're getting down to the end of the time, so I'll try to be quick. One: I am the son of an MIT professor. I believe very firmly in education. If you learn how to think and you learn how to communicate, it will always be valuable, always. And so number one. Number two: AI does great at patterns, which is awesome. Startups don't have patterns. There is nothing. You have to be a human to do it. You can't just have an AI startup. I wanna start, I wanna do a startup for me. That won't work. Humans are great at recognizing early patterns and having vision. I don't know how AI will ever get there. I did my master's thesis forever ago in the 1900s — when I commuted to school by horse — on speech recognition, with machine learning. You have to have the inputs and then what the output was and what the decision was, and that's how you teach them. So anything where you're venturing into the unknown, AI can't do. And humans have to do it. So as we see AI start to push into spaces, it's going to make rote, repetitive portions of every job easy, right? And it will allow humans to focus on the new, the creative, the building. Thinking about how and where we should be doing these things, thinking about how to construct these things. Maybe AI could design a better network infrastructure, but I don't think so. Not for your company in your use case. Like they don't know where you're — it's really hard. So I expect this’ll be somewhat akin to the first wave of the computer revolution. It was terrifying for people 'cause it would take rooms of accountants, rooms of them — at Ford and GM, like floors of buildings with people just adding data all day long. And it turned 'em into spreadsheets and suddenly those floors became like a pod on a floor and it freed up very bright minds to go do other jobs. As long as it doesn't happen overnight, there's time for — slow dislocations are fine. People will find ways to be productive and be creative and further humanity. But some of these jobs are gonna go away, not entirely, because the creative part I still think is gonna have to be human-led. But there are gonna be fewer people doing those tasks. I know there's a lot of concern about coding jobs. Yeah, the coding jobs where you have to make one little change to facilitate an upgrade by going through an entire codebase and spending two weeks doing it — that should go away. Frankly, I don't, I can't imagine anyone wants to do that. That's a good thing. Software continues to eat the world. There's gonna be a ton of new opportunities to build things and if AI helps you ideate and build some prototypes and do some things and get there more quickly, you can be more productive. In the long run, I think we'll be fine. As long as the AI revolution takes decades — and I think it will — we'll be fine. So that's my long rambling response to your question. Hopefully that was, yeah. Adil Saleh 40:16 I absolutely love it. So thank you very much Paul, for taking the time again and being so genuinely concrete and original and love the energy. And I would love to have this episode out in a few weeks time. Thank you very much for your time. Paul Staelin 40:34 That, that would be fabulous. Thank you for inviting me back on and congratulations on all of your success. I know when I was on, it was very early. 160 episodes — a lot of episodes. So well done. Adil Saleh 40:46 Thank you very much, Paul, it was not possible without people like you and giving the knowledge and exposure that I get to be able to get the energy to get onto the next episode. Thank you very much. Paul Staelin 40:55 All right. Thank you Adil.