10 AI Trends in SaaS Right Now with SaaStr CEO Jason Lemkin on LIVE SaaStr Workshop Wednesday

back with me, with me, Jason M. Lemkin, Workshop Wednesday. Um, and, uh, they'll actually probably be pretty tight and focused until SaaStr Annual, cuz we're heads down on that. May 13th to 15th, please come. We'll have 10,000 of the best in AI, which we're gonna talk about today. SaaS, B2B will be about SaaStr Annual this year will be about half AI for B2B and SaaS, practical stuff, practical stuff, what's working, what isn't, and about half classic go-to-market scaling content. So come, Um, and well, in addition, one thing that's pretty cool this year is we have these VIP summits. We have a couple hundred CROs and VPs of sales, a couple, almost 300, uh, chief customer officers and VP of, of customer success, over 200, uh, CMOs and VPs of marketing. And at the end of these, they are closed, they're, they're VIP only, but at the end of them we're gonna have cocktail parties and meetups. So if you wanna meet a great VP of sales, marketing, customer success, or CFO, These, these, these behind-door summits will now have an open area where founders can come. Only, only founders, only CEOs and founders can come. But you are invited to come network with up to 1,000 of the best leaders in B2B. So honestly, it might be just coming for that for SaaStr Annual, May 13th, 15th. It's probably worth it. So today, listen, I talk a lot about AI, but I am a student like you. I am not an expert. I cannot exactly tell you the specifics of the last Gemini model that just pushed out, and I cannot compare it to Claude. And honestly, I, I, I, I'm not exactly sure which one is better, Cursor or Windsurf. I don't know. Okay. But I am a student and I've made 30 pretty good investments as part of Saastr Fund. I watch what's really happening at the portfolio. We are now running our own Saastr AI, which you guys should try. It's on our homepage. And we've done 4,000, uh, interactions with our AI in just a week and a half. And I've learned a lot from that. And I know everything that's going on in venture and otherwise and where the energy is. And so what I wanted to go through was sort of 10 trends that I'm seeing and not fall behind on. And it, listen, if you're super deep into the field and you're going to meetups in SF every single night or 2 or 3 a night, some of this may be obvious, but even all of it won't be. But mostly I want this, I wanted to highlight things that some folks are still curmudgeons about. They, they think folks, listen, I think A year ago, you could be skeptical of AI in SaaS and B2B. I think even 6 months ago, you could make fun of hallucinations. You could, you could mock some things. Not anymore, guys. Not anymore. And I want to walk you through especially the mistakes that I think curmudgeons or folks that are slow to think about AI is because honestly, I think even if you're in the most regulated industry, even if you're in the most conservative industry, software, I really hope your product and engineering teams are obsessing about AI because it's just getting going. But I wanted to share a few things. This first slide, it's super interesting, especially for folks that are not in engineering, dev, or aren't adopting AI. What this is, is I want to talk about how AI is getting so much better. And what is this chart for merged pull requests? This is from— I have permission. This is from one of my favorite investments, a company called RevenueCat. It's at pretty big scale. It's, it's, it's— the team's pretty tight, but they power mobile subscriptions for 40% of all North America apps, mobile apps that have mobile subscriptions, everyone from powering ChatGPT to Arnold Schwarzenegger weightlifting to thousands and thousands of mobile apps. They have tons of customers. They're a quiet monster. And the CEO shared with me this— the impact here on bringing in the latest tools like Cursor, Windsurf, and others. And there's some subtleties here I want to share. But look what happened to their productivity and engineering on the right of this chart. Okay. It had grown over since July of 2022. And then watch that massive inflection, massive inflection when they brought AI development in. And there's a lot of reasons I'm pretty skeptical about a one person running a billion dollar company. I'm pretty skeptical that tiny engineering teams can scale forever, if for no other reason than bug requests, refactoring, you know, so many of these tools, they are better at building 1.0s and new code rather than fixing issues. At the moment, at the moment. But look at this productivity boom. Look at it. Look at it. So if you're scratching your head and you're like, my God, how did, how did Cursor go from like 0 to 100 million in 2 months? And why is Windsurf hire 75 sales reps in 60 days? And what's happening with Bolt and Lovable and all of these? And you can say, listen, they're fatty and people switch. And look, folks switch from these, these new AI software. They do switch. They go back and forth. And for Cursor and Windsurf, for a while, Windsurf in some ways was behind because Codium started off as a plugin, but people love Windsurf and they switched from Cursor in a week. So there's lots of stickiness issues. There's lots of questions. Maybe to some extent, a lot of these development apps are just wrappers on top of Claude. That's not my point. My point is looking at real-world productivity gains from these apps. It's stunning. It's stunning. And if you see this level of productivity in your dev team, is it worth $20 a month for an engineer? This is the thing to reflect on. Now, listen, you can pay a lot more for apps like Cursor and Windsurf. If you use a ton of it, you can't— you're limited at what you get for $20 or $40 a month. But— and this will ratchet— this will go through all this presentation. Think for a minute, for between $20 and $40 a month, you can add the level of productivity gains on your engineering team you're seeing on the right of this chart. Just think about it. Your product's probably a ripoff compared to that. I mean, listen, we pay like $15,000 a year for Salesforce. We use 3 seats. We don't get this level of productivity outta Salesforce, do we, Amelia? No, we have not seen almost a doubling of our productivity on our sales team by using a, a, you know, a CRM. This is jaw-dropping. This is jaw-dropping. Now, one last point, we could spend almost a whole session on this to bear in mind. This is a meta-learning. Is RevenueCat taking these productivity gains and keeping status quo. No, actually they are, even though they're, they're kind of a monster, they actually are cash flow breakeven, mainly because they have 80% of their customers are PLG and self-serve, but actually they're plowing that back into hiring top-tier developers. And so one reminder from this is that not only is AI making the best developers more efficient, but here's the thing to really think about, guys. It's amping up the competition. It's amping up the competition. So if you take someone like RevenueCat that kind of owns its market, but it's also continuing to pull away from its competitors, not only is productivity up almost 100%, but they're taking those gains and hiring even more developers to make them more productive. So you, you're gonna fall behind if you don't keep up here. You're gonna fall behind because the best are gonna get better and ship better and better software. It's not just about, um, vibe coding and in an hour creating a Snake clone. I mean, that looks cool on Twitter, but so what? Like the market for Snake clones fairly limited. You're not gonna get to a billion AR with a Snake clone or a, or a, or a Flight Simulator clone that's really just pulling open source code and rewriting it. It's not that as impressive as it looks on Twitter. But this means that your most, aggressive, your hardest working, your competitors that push out the most features, the most code, may be able to push out twice as many features and code using these tools. It's powerful use of AI. And you better keep up. So everything's gonna get better and better. But this is a pretty good example. There's a lot of Sony baloney, a lot of, fake bravado on social media about how one engineer can do everything, can build, again, a Workday or Salesforce. Not gonna happen. But this, we haven't seen this step function in productivity. It's being poured back into even more software. And here's a fun example, but if you guys haven't tried it, go to our homepage. It's right on our homepage or chat in our little, like, window at the bottom, our chat window, and look at our AI. We've done, it's been up for less than 2 weeks. We've already had 4,000 AI chats on it. 4,000. And I've learned a lot. I've learned a lot. In fact, I just made an investment out of what my learnings here. I invested in a company that— I'll talk about it more later— that automates part of how these AI bots work. But what's pretty cool if you try it is how it's constantly getting better. So literally our AI is trained on every single piece of Saastr content since 2012. It has almost 20 million words of content. I don't know how many pieces of content, 20 million words, and it's pretty darn good. It's not perfect. It's pretty darn good. And this is crazy. I pulled up mine and Brian Halligan, who's the founder and chairman of HubSpot, he's trained it on a lot of his content. Not— he doesn't have as much as, as, as we do at Saastr. So I'd argue our AI is better, but his is pretty good. He invests a lot of time. And they're talking to each other and sharing learnings about SaaS. Voice, our voices. And, um, it was already pretty good. Um, but then I kind of put, put this point on Twitter that, that the problem is they, they, they can't quite hear each other very well in the browser, so they interrupt each other too much. Um, and here's the founder of the app, Delphi Intelligent. Um, uh, you know, interruptions are coming soon. So everything's getting better at such a quick rate. I, I, it would not shock me if in 90 days the conversations that Digital Me and Digital Brian have are pretty epic. They're pretty epic. And I don't even know what it's going to look like at the end of this year. So whatever you're judging, I just challenge you. Are you dated? Are you— are you grouchy because you're making fun of untrained hallucinations in ChatGPT, man? Go look at Best of Breed stuff. Go log into our AI and don't ask a random question like, what's your favorite flavor of ice cream, Jason? Ask a classic Saastr question, right? I've got 2 sales reps, but I've hired my third. The third one is not performing as well as the first 2. What's happening? What do I do? Right. My VP of sales came in and hired 3 reps that are underperforming. Performing, what do I do? Ask classic SaaS questions. I bet you're going to be pretty impressed by how this AI is doing. And why isn't your app doing this? Why doesn't your app have as good a user interface discussion as the SaaS AI does? So try it. Um, another data point I want to share with you about getting better. This is pretty helpful for the cynics out here. So, um, I'm on the board and the largest investor of a company called Gorgeous, which is the largest support and contact center app in e-commerce. Okay. They've got 20,000 SMBs. And if you don't study contact center and support, it was what— that and coding. We talked about Christian Windsor. This support has gone the most AI of almost any classic B2B space. Mark Benioff's talking about Agentforce. There's a lot of things you can do with Agentforce, but he's talking about resolving tickets, right? All of this stuff. Gorgeous. Everyone here is early. Gorgeous is early. Intercom, Zendesk, they're all running Talkdesk. They're all AI-first. Dialpad reaccelerated. Dialpad is growing faster at $300 million ARR than $200 million because of the benefits of AI in the space for resolving tickets. And you can make fun of these things. You can make fun of AI SDRs and say they don't set the perfect appointment. But if you look at what's really happening in contact center, that's not the goal. The goal is instead of waiting 5 minutes or an hour or 2 days to talk to a human in support, um, I get an instant response. It, in the case of Gorgias, it integrates with Shopify data. It knows everything about my shopping cart, what I've bought, the issues. Um, it handles as many of the responses as it can, and then it goes and grabs a human. It just goes and grabs a human if there's an issue. And a couple learnings, but then I wanna talk about this chart at Gorgias. The, the AI is always in the top 10% of CSAT. It's never number one. There's always humans that are better than the AI and it can't handle the most complicated issues. It has to go grab Jason or Amelia, but it's in the top 10% of CSAT. Something to think about. But because it has tons of data, remember, this is not an untrained model. Gorgeous AI is able to integrate with all of your e-commerce data that you can possibly get out of itself, out of Shopify, out of others. And what they did here, and I blocked off the competitors. Here's a chart of how Gorgeous compares to all the other leaders in the contact center. I'm a little colorblind. The one on the left, is that pink or red? What is that one, Amelia, with the, with the one that says 60%?

It's a, it's a light red. Yeah.

A salmon. Could we call it salmon?

Salmon.

Yeah. Would it be fair to call it salmon? Colorblind. Okay. So what you can see is everyone in the industry, and you can see this on social media, claims 60% of support can be resolved with automation. In fact, I don't even know. Marc Benioff claims it's over 90% at Salesforce. Okay, more power to him. I mean, he's the best of the best. But, and then these other charts, whether they're gray or pink, I don't know what they are. This is the actual automation rates for real, right, that they tested across thousands of it. And what you can see is this is both the, the, the, the, the issue and the, and the interesting reality in AI. Are folks claiming it does better than it does? Yes. Everyone's claiming 60%. No one's achieving 60%. But no one's got a zero. No one's got a zero. So are we saying that AI SDRs and AI sales and AI support and AI marketing does more than it can do today? Yes. Right. But it is improving at a rapid rate and it really is working and it's constantly getting better. So before you poop on this stuff and before you get eclipsed and lapped by the competition, understand this is working and it's getting better and better and better. Um, even our clone, we got the badge of legendary after 2 weeks. We're legendary with our 4,000, uh, conversations and issues that we've resolved. Um, okay, one other point I wanted to make, uh, on this chart to understand with the— this data of this AI claimed automation rate and real automation rate from Gorgias. And, and, and I, and I know this is like maybe somewhat obvious to folks, but I want to, to make this clear if you're not seeing it. Not everywhere, not everywhere today, but in so many parts of tech and even other industries, customer expectations are skyrocketing. Skyrocketing. Okay. Um, listen, there was, you know, this, this AI SDR agent 11x got flamed a bit this week. I defended them. I think these products work pretty darn well as long as you know, as long as you know what you're getting. Okay. I don't think an AI SDR can do perfect ABM, can schedule the call, can close the deal, can do customer success and do the renewal on its own.

Okay?

We're expecting too much from some of these tools today. But, but, but no matter what they can or cannot do, the fact that they can do some of this is creating crazy expectations from customers. And what I can tell you, because I know so much about this post-sale space, this support space, I was the first investor in Talkdesk and Gorgeous and Maestro QA, which is very enterprise. I'm a small investor in Intercom and Front and many others, maybe too much, too much in support. Is one issue here, and it's true in the, in the AI sales space too, is everyone out there is making big promises. I mean, Marc Benioff is the OG, but he's claiming AgentForce can resolve over 90% of issues. This is the OG of Salesforce. When that happens, what are the customers going to expect? They're going to expect this. And maybe you cannot deliver it today, but they're expecting you to get as close as possible. And I'll say one thing I'm seeing, especially in the enterprise, because in many ways the enterprise unusually is earlier on B2B AI than SMBs. They're earlier because a lot of stuff does require training. It requires onboarding. It requires work to tune these models. Even today, it still requires some work. And enterprises are better equipped to spend 30 days optimizing onboard and optimizing how AI works than an SMB that might have 5 minutes. Okay, there are exceptions. I could talk through those later if people want. But, but this is one of the rare cases where in many ways enterprise is ahead of SMB. And so what's happened in the enterprise is a lot of vendors, especially in post-sales for over a year, have been claiming miracles from AI, not just today, but 12 months ago. And we're seeing on renewals that folks are boomeranging. They are leaving AI vendors they bought a year ago in the, in the support space, in the post-sales because the promises could not be met. Okay, they're leaving it. And so these, these excessive promises, they're getting burned by it. But it's not like the customers are expecting no AI. They're just expecting a more balanced approach, like on the right. So, you know, you can fight this, you can mock some of this, you can say it doesn't work in our industry. You can say our customers don't really want it. But every week the customer expectations are skyrocketing. Every time Marc Benioff goes on CNBC or Fox or wherever he goes, or Jim, Jim, whatever, Cramer, and says AgentForce is resolving 95% of issues. You know, the world's gonna listen and they're gonna be expecting that from you. They're gonna, and they're expecting that at least you, you come as close as you can. Okay. Another point, just 'cause I see a lot of founders especially struggle with this, is just, you know, AI is, a rebirth of freemium and PLG. PLG and freemium have come back anyway. Folks are looking for scalable ways to grow. PagerDuty and HubSpot, for example, have kind of refocused on the low end of the market. PagerDuty has refocused on free, which it did not do for many years. PagerDuty is really just like almost all enterprise and free now. If you look at their numbers on SaaS, but almost all the, the AI apps we use, at least at the prosumer level, have a free version. Look at ChatGPT, look at Claude, look at whatever you use. Look at Opus, where I'm an investor for clipping. I'll talk about that in a minute. For doing clips from social media. Like, the free versions are really, really, really good. They're really good. And, there was an AI app that I love, and, I, I kind of got a little, I mean, not literally upset, but I got pretty frustrated with the founder because I knew this app was great. I knew it was great, but because I'd seen it in production, but they immediately wanted $200 a month on my credit card without even letting me try it once. Once. And then when I tried it, The onboarding was really rough and didn't work well. It was like, you can't do that. And the founder at first was like, and this is a great company and a great founder. The founder was like, well, I can't afford to do freemium because of I'm using so much, so many GPUs, so many cycles. I'm like, I don't care. And we'll talk about this. The leaders have figured this out one way or another. Some of them are figuring out by burning billions like OpenAI, others like Opus and others have figured it out by being efficient, right? But the markets, especially the SMB markets, are gonna be expecting to use your AI tool for free, at least initially, better, better, better, better embrace that. Because if you fight it, if every argument is I can't afford it because of AI costs, you're going to be at odds with how the market wants to use these products. And this is kind of the next point. And then a couple points that are related to think about. I don't have the answers to this, to this issue. But I think we need to think about whether we can really charge for AI. In SaaS? Can we really charge? A couple of weeks ago, I put up on Saastr, I asked this question to the CEOs of Procore, public, public vertical SaaS leader in construction, over $1 billion in revenue, the CEO of Monday, and Brian Halligan, the real Brian Halligan, co-founder, chairman of HubSpot. And I put them all there and no one really had the answers to if they can. Now, Salesforce is charging $2 for each agent force, not even per resolution, just to use it. But even if you listen to the latest earnings release, Mark, even though agent force is off to a strong start, it sounds like they're really packaged into other deals. Deals they've sold for H&R AI are flat rate deals. They're bundles. They're not really, they're not really raising prices and really charging per resolution. And I don't, I would just be cautious. I would just be cautious. What HubSpot is doing is not charging for a lot of AI today. They're learning. What Monday is doing is the same thing that Asana is doing. A lot of usage is included, and then they're charging incrementally for more usage. But I don't even know if that's gonna stick as costs come down. I don't even know if that's gonna stick. I think it is likely that we will expect more and more in our per-seat pricing, our base pricing. And no matter what the VCs say in social media, and I just compare this again, here's Cursor, um, $0 to $100 million in 2 months, um, for building better software. Now you can't— there are caps on these plans. There are caps. Okay. But look at what you get. Um, you can start for free, 2,000 completions. It's not much, but you can start for free. And then $20 a month for unlimited slow and 50 fast, and then $40 a month, like, This, this is what people are going to expect. Not another $2 a month for your lame AI agent to maybe finish something. Okay. The expirations are going up. So I think that overall AI may well lead to deflation. Everyone in the last year has been trying to increase pricing. Can I charge more for this? Can I— Zendesk now charges $50 more a month for its AI module. Is that going to stick? I am skeptical because if you look at the leaders in the space, the ones that are growing at a crazy rate, in many cases, that's not the way they're charging. What they're doing is providing more value than maybe we've seen in our lifetimes as B2B executives. The amount of value we got on that revenue chart for productivity is jaw-dropping. It's jaw-dropping. And so we may have to deliver much more software, much more value, and much more AI just to earn whatever we're charging today. I, I, that's what, that's my, what I'm seeing and where I think we'll end up. Point 5, which we talked about before, and I think we'll have time to the left. So happy to open to questions on this or really anything that people want to ask. Just, just we can ask questions at the end in the, in the chat or otherwise. Another thing I hear from a lot of B2B founders, even now I hear it at board meetings, is we can't afford it. We can't afford it. We can't afford it. Jason M. Lemkin, it's just too expensive. Like we've done little experiments, but AI is too expensive. It's not, or it is, or who cares? First of all, the customers don't care. Second of all, if your competition has figured it out, you don't get— no one cares that, that it's expensive. Okay. And I can tell you just a couple things. The customers expect it and the best folks figure it out. I invest. I did invest to learn. I'm a small investor in this company called Opus Clip. We used to make our clips and they rocketed from, I think, 1 to 15 or maybe 20 in 12 months, maybe some, some variant of that hypergrowth. And when I talked with the founders, I'm like, listen, I already use this product. I'm already a customer, I love it, but you've got to be burning millions a month making these clips. He's like, nope, we've optimized everything to the nth degree in how we do these clips, and we do big pre-purchases. Now, I don't know what it's exactly like today, but when they started to scale, they would do million-dollar pre-buys of credits of AI. Okay, now the bars are lower today, but that would bring down the cost 90% by doing a million-dollar pre-buy. Gorgeous. The, the, the, um, the AI contact center for e-commerce. We talked about it. They do something similar. They do prebuys. Their costs for both of their costs for AI have gone down an order of magnitude. They've gone down 90%. Okay. Just by being thoughtful about their spend. And then people figure out how to do things better. Like if you try to throw up one of your YouTubes on Opus Clip, when I first invested, they like did all the clips upfront. Now they, they, they intelligently render them like they know that not that many people are going to use the 10th or 11th or 12th clip. So they show you a little thumbnail, but they actually don't render it and do all the work. There's a little bit of a pause. But that saves them a ton of money. So I could share stories from Gorgeous, which I know well, or a few from Opus and some others, but people have figured— the smart engineering teams just figure this out in SaaS. And if your team is saying, I can't do this, it's too expensive, woe is me, you're going to get destroyed by the competition. Figure it out. Figure it out. And costs are falling. And yes, as costs are falling, and Dharmesh Shah and Aaron Levy and others make this point, as costs are falling in AI, we're asking it to do more. Right? For sure. So, so there is that issue. But a lot of like core functionality, it just isn't that expensive at scale and you're gonna have to do it. Even Aaron Levy made this— look at Box, just because Aaron is pretty fun on Twitter. And again, he will be at SaaStr Annual in May talking about AI. So come hear Aaron live and bring your questions. Already at Box, the stuff that they started to charge for, a lot of it they can give away for free because of the cost changes at Box, right? Um, so you're gonna have to— this is not an acceptable excuse, uh, in going forward cost. Okay, point 6, this is my biggest challenge maybe of the— of all. Um, maybe there's others in here, but this is one I really— I— it— that I really want to challenge you guys to think about to make your products better, is that a lot of social media, a lot of articles, a lot of Gartner reports are all about AI and efficiency. We saw that in the first slide with the RevenueCat, um, pull requests and productivity for software. We can see it in AI support where, you know, the support, the 40% of support tickets that are handled for by AI are obviously cheaper, right? The honest truth in a lot of contact centers, a lot of support centers, they're laying off 30, 40% of their teams and replacing them with AI. It's truth. So efficiency, but there's a bigger win in AI. There's a bigger win. If you go back and watch my interview with, with Tui, the CEO of Procore, if if you talk to folks in vertical SaaS in particular, what you'll hear is there's no one to do the work. We're, and I ask Tui at, at, at, in construction, in PropTech, in construction software, what's the number one application for AI? And he is like, well, I've got DeepSeek running on my local PC and I'm playing with this and I'm doing prompts and I'm doing all this. And watch the interview and ask him, but what's the biggest practical issue? He said, the biggest practical issue, you know, got to these big construction sites where they're running our Procore software. And they're building these big buildings and other things. And there's that, there's that big, um, trailer where they're doing all the support, the, the, the paperwork, the finance, that everything. They can't find anybody to work there. You can't find anybody to write your code. You can't find anyone to sell your product. You can't find anyone to send that marketing email. Even we can't. Um, but they can't either. There is no one to work in that trailer. Um, I, I invest in this company called Mango Mint. It's almost at $20 million that does, uh, software first, spas and salons, a tough space really. You can watch the, the great interview we do with the VP of sales on that. The real reason Manglement took off is because in these spas and salons, there's no one to work the front office and there's no one to work the back office. They can't find anybody. So it's not about saving money, which is always great. Any business, if you can really save the money, will, will give you some of it. But I believe the biggest benefits in AI in B2B are coming from automated areas where no one will do the work. No one will do the work. And that's why, unlike some folks, I'm not very critical of these AI SDR tools because we have a small sales team at SaaStr and I watch my portfolio companies and no one can find enough good SDRs. No one can find— you can find people that are mediocre or that will just like randomly assemble a list of a, of a crappy cadence and send it out with, with no work and no effort and no set up, no appointments. But I think an AI SDR one can do a better job than that. And two, you might not be able to find anyone willing to do that work, at least in the US. So many knowledge workers in IT just don't want to do the work anymore. It, I don't, it's a reality. And AI is what's gonna bail us out. So watch the, it's not just in the tech, it's not just watch this interview with Tui, founder of Procore for over 20 years. So excited about AI because there aren't people in construction to do the work and he's gonna solve that problem. Okay, that's my big takeaway. If you think about that, you might build something really, really awesome in your product, even if you don't sell directly in some of these industries or tech. Okay, just a couple other points and then we can wrap it up and open it for questions. I said this before, but it— this is my number one kick in the ass for anyone watching, listening, participating, anyone in B2B. I've said this before, but I want to take a moment to say it again because it's so important. This is the reality, folks. Winners are hyperfunctional and the pace is intense. The best leaders, the next generation leaders in B2B, one, they're pushing out code twice as fast because of AI. We talked about it before. Two, and many of you are not gonna like this, they are in SF working 7 days a week. It is a vibe. It is a reality. They are doing nothing in, in, uh, you know, during lockdown everyone was about quality of life. Everyone, uh, because frankly everyone, the average public SaaS company in 2021 grew 70%. Which is stunning at a rate we've never seen before. 70, the public, old public companies grew 70% on average. And if you think about in 2020, it was basically the same software as 2019. So of course in 2021 when demand was through the roof and you could sell 4-year-old software, everyone just was like, kumbaya, you know, work, work a good 20, take care of yourself, have spa days, work from home 3 days a week, uh, take Thursdays off. Um, Friday go play golf because products did sell themselves and demand was incredible. Today, the, the folks that many, many leaders are falling behind and the next generation is working 7 days a week in San Francisco, pushing out software at a pace we haven't seen before. And I know that's frustrating, but you got to recognize if you don't want to fall behind, you got to recognize that it's brutal. But, you know, Jack from Block Square just announced that he laid off 1,000 people from Block Square. And he was very direct. Again, he's very direct on why. He's like, we have to move faster, right? He's like, 400% of them did not move fast enough. He's like, it's performance reasons. 300 of them, I think he said, need to change what they're doing to move faster. And 200 of them are managers we don't need. And they're— now your job is an IC or leave. 200 of them used to be managers. You're coding, you're selling, or you're out. Everyone is moving faster. And this chart on the right is, uh, uh, a Q1 roadmap for us, for a super fast growing AI company I invested in that has 2 people, 2 people, and they're gonna push out all this code. What about your 50 engineers, your 100 engineers? It's tough. It's tough. So look, I know it's tough, but you just can't work in sales, in marketing or engineering at this 2020, 2022 pace anymore. It doesn't work. You've gotta find a way to go faster. And honestly, this is tough. You can't hire people that wanna work at that pace. You can't hire them or you will fall behind because of who you hire. Okay, couple more points. This one's just a learning from, from my AI, but a couple ones and then, and then we'll wrap up learnings from ours. There is— I don't, I don't know for sure what exactly. I'm only so smart in AI. I don't know what the UI/UX of 2028 will look like. I'm not sure what 2020— 7 will look like. But everyone needs some sort of S-tier or at least A-tier Copilot now. This is what maybe, maybe middle America isn't expecting this. Maybe traditional industries aren't expecting this yet. And we know a couple months ago people started to say everyone's gonna be a prompt engineer. I didn't even know what, I gotta be honest, I didn't even know what that meant. I mean, I sort of knew what it meant, but I'm like, I, I don't have time. Like, if you ask, when I started playing with ChatGPT, I'm like, this is a tax on my time. I got— I'm a busy guy. I got 60 seconds. I want to type something into Google or type something into an app we use, and I got to move on. I don't want to figure out how to craft a prompt. It just seemed crazy. And listen, I don't want to go back in time to what some of these apps looked like last year, but today, like, this is the current UI and UX, and they're getting better and better. And you can ask pretty good questions and get incredible answers. And what— and is yours great? Is yours great or is it not? You should have not a lame chatbot on your site. Not— I mean, literally not one that when you go into it, it says someone will get back to you in 2 days. You should have an incredible copilot. And this copilot should do a lot of things in one copilot. It should solve problems. Okay. Because it's got to do support. It's got to do it well. 2, it should start selling the product if they have sales questions. What's the minimum number of seats? What if I want to share this? What if I want to do this? Your AI should be good at sales and it should be able to do some remarketing because it's talking to you. It should capture your email address and send you offers and remarket to you. You need this S-tier Copilot. Um, and what's yours? Be honest. Is yours— go try again. Try our SaaS one. Yours should be better. If yours isn't better, go beat yourself up and yell at your team because you're going to run out of time. You're going to run out of time, right? And again, this is my question. Maybe I comprised 8, 9 to 1. Yours has got to be great. And try it and try our voice too. It's really, really good. And if one other interesting thing I'll tell you that we do, and maybe these last points are really one about how your new Copilot needs to be great on a bunch of meetings. We do this thing and it's kind of cool and kind of creepy at the same time. You can try it. We bring up an iMac in the office and we bring up Digital Jason using this voice. Click on the right, me and someone else like me, Amelia and Digital Jason. Digital Jason listens to our meeting and I can ask it questions if I forget. Hey, Digital Jason, I forgot when we interviewed Tui from Procore, what did— did he think domain expertise was important in proptech or not? Digital Jason will know. Digital Jason would know. Imagine your sales executive went to a meeting with a customer and didn't quite know how an integration worked or didn't quite know who similar customers were, didn't quite know what the competitive landscape was like, and the digital one knew. It's pretty great when you add this person to a meeting. It's something to learn and something to think about. And again, I'll, I'll just do two examples and then we'll break. But I, and I, and I know I said this, but I want to end on this. Fake AI is not working anymore, guys. It might have worked a little bit for, for dumb buyers. It might have worked, but everything's getting really good. And, and when I compared— here's a great tweet, tweet, uh, Twitter interaction on the right about how good our AI is, because it's got the CEOs and CMOs of HubSpot, ToastWiz, Databricks. He said, I build AI agents and this one's so good, right? So that's a great one. Here I went, I'm gonna write up a 5 interesting learnings on SaaS or on ZipRecruiter. ZipRecruiter is, is a, is an old, you know, you probably use ZipRecruiter. It is a top web recruiting platform. Its revenues are down 30% the last few years. Down, down. So I went to its homepage and its homepage is their AI bot. Okay, Phil, he's our career advisor. It's terrible, guys. Go try it if you want to try a terrible bot. Try ZipRecruiter. You know what this is? This is fake AI. This is an onboarding wizard. Okay. That looks like an AI, but it's not an AI. Okay? It's just trying to get me to sign up for their product. It's a C-. This isn't going to cut it today. It needs to be S-tier. So be honest. To be honest, a lot of founders I talk to even today that are deep in B2B, they don't really use these products. They don't really understand the value. They're there. They think it's still checking the box. It's not. It's got to be great. This— try the ZipRecruiter one. It's terrible. Try ours. It's pretty good. Try them all. Try every one you can and be one of the best. So these are my practical learnings from AI. I think we have a few minutes left. So if we've, if we've got some, I can take questions or we can just have a conversation for, for about 15 minutes.

Yeah, we need to get Visual Jason in the Zoom somehow.

We do. Well, I'll tell you what's interesting. The product we use is called Delphi AI. Try it. It's really great. You'll try ours.

And he's coming to annual.

He's coming to annual. He's gonna share, he's gonna come to Saastr and actually share how we do our AI, Brian Halligan's AI, and Lenny's AI. He's gonna share all the secrets. So that session alone is probably worth coming to Saastr annual and asking questions. We'll do an AMA after. But for example, they have, not only can Brian Halligan and I talk to each other, they have a feature where Digital Jason can come to Zoom. Okay. Honestly, it doesn't really work that well today. It doesn't really work that well, but I will bet you in 90 days we do another one of these and Digital Jason's pretty awesome. On this. So everything is just getting better. So where you're skeptical, I guess the point is where you're skeptical in AI, don't be skeptical. Start thinking where can it be in a couple iterations? Because if you go deep on these platforms, you're getting a win because they're evolving so quickly. So, all right. I know we've got a lot of note takers and stuff in the chats, but if we have any questions, I'll look for them.

There's a few.

You can bring anybody up that we have.

Yeah, I know where there were a few. A couple of people asked, Gurush in particular, or if he wants to come up to ask any other questions. I don't know if you know the answers actually, because it is on Delphi, but how you trained your DigitalJSON on the model.

Yeah. And that I am a super fan of Delphi. It's really interesting. And maybe I can even bring it up and show you the backend. Literally, you just attach the feeds. Okay. You could— we have an RSS feed from Saastr that helps with new content. So it's great because literally I was just talking with our AI this morning and it knew about a Saastr post of this week. Okay, so like if you think about, you could be critical a little while ago when you go to use ChatGPT or Claude, you'd be like, oh, it's so annoying when they tell you I only have a database through 1872. Used to get those messages. It doesn't do that anymore. It's up to, it's up to date with yesterday. It's up to date with yesterday. So it has an RSS feed. It, it has the ability to, and I don't know exactly how this part works, it's some ability to scrape. So it scraped every single piece of content in SaaS going back to 2012, right? It ingested it. It ingested it. It has a YouTube feed, which for some might be minor, but for us, if you think about our YouTube, it's got every SaaStr annual session going back to 2015. It's got Stewart Butterfield, it's got Brian Halligan, it's got, it's got the CEO of Databricks, it's got everyone's thoughts. And so the AI knows 1,000 SaaStr sessions we've done. So it pulls YouTube, it pulls RSS, it pulls our blog. You can give it individual web pages to do. And then what you can do— and I only did a little bit of this, but Brian Halligan did a lot. And it's especially easy on mobile. You can just take a note and train it. So you can, if you go to, you don't even need to use the mobile app. If you just go through the mobile version of Delphi and you say, hey, I wanna train you on what I'm thinking about AI today, you can upload the content that way. But we have so much content. Honestly, I was skeptical it would work as well as it did. I'd watched Brian's and Lenny's, so I knew I wanted to use it. But you know, it did take, I think it took a full day or so, or maybe longer to ingest all our content. It couldn't do it instantly. But it has almost all of it. It's pretty easy to do. So it's super impressive.

No worries.

Okay.

I'm going to bring Andre up because he has a— I don't think he has a question, but he has a comment. So I'll bring him up.

Yeah, Jason, first of all, thanks for doing that.

Can you hear me?

We can hear you.

Okay. Yeah. You know, just a quick supporting comment. I do want to say the following. 6 months ago, 8 months ago, I was in that cautiously optimistic camp. Yeah. Uh, and right now I'm in the, you know, leap scared camp. It's so good.

Should be scared. Leap scared. Leap scared.

Yeah, yeah, yeah. So, you know, just to share my experience, you know, I, I took this online class at Stanford which was NLM for Business taught by some really cool folks, ex-Google folks, and we, you know, we coded our own projects. And, you know, I called my own GPT, and it felt like a superpower. I, you know, I'm not— I'm not programmed for 25 years. I never programmed in Python. And my one observation is Python skill is going to become like PowerPoint and Excel. Like, if you— almost everyone will need to have some kind of skill like that to string together, you know, a bunch of apps. Second thing I found out was it's absolutely true.

I mean, there is even my son has a resurgence of Python. Everyone that's— everyone that's pretty deep into AI is a Python. Yeah, but I don't think the I, you can make your second point, but let's, let's try to keep it to one. But I, I do think it's all interesting, but I don't think the average founder CEO that isn't a CTO needs to care. I don't think anymore you need to care what model it is. I don't think you need to care what's new and what new, what OpenAI just rolled out, whatever for images. You don't have to care. Like, and in fact, if you use some of these products like, uh, Assistant UI that I, which is a framework I invested in, if you use Delphi, they're gonna pick the models for you. You don't even have to care. And they're gonna combine the models together, like for, for different, like if you use these frameworks, yeah, will handle it all for you. So you, great if you wanna like buy the latest Mac M3 and run DeepSync locally on it, be cool. But I actually think the power, you don't need to, you, you, Twitter's fun, learn from Twitter, but you don't need to learn any, they, they're gonna be so good that at least the frameworks and the middleware is gonna handle it all for you. You don't need, Delphi can, can look at me in the camera and identify different people in the room and talk to each of us.

Yeah.

Do I know what API do I, I don't know, is this ElevenLabs or I don't care. I don't have to care.

Care.

It, it doesn't matter. It's all going to be abstracted at a level unless I— unless for SaaS, it doesn't matter. For SaaS, the differences between these models is not going to matter for 95% of— and where it is, your engineering team will figure it out. So, yeah, and I don't want to tip that all the time. These are good points, and we can talk about Python and this, and it's cool, but I also don't think— I don't, I don't want founders to get distracted in the noise because everything's getting so good.

Everything—

like 2 months ago, everything's getting—

I was calling a specific model in Perplexity, and now they released, you know, it automatically figures out what the right model is.

Yeah, you don't need to know. You don't need to know.

And in like 3 days, we were able to code something where we now are able to create account research and a PowerPoint slide almost as good as a McKinsey consultant walking into a meeting, um, at so much lower of the cost because we were able to just integrate with complexity, deep search, things like that.

Pros and cons.

Mind-blowing. The kind of work people can do with this.

I think you're right. I do think— let's take some other questions. These are good topics. I do think on the niche top, because we do create a lot of content at Saastr, I do think AI makes content creation an order of magnitude simpler, right? But it also— there's already such an explosion, it's going to debase content as well. It's a nuanced topic. But yeah, there's no excuse for a crappy presentation anymore. There's no excuse for typos. There's— and frankly, But more interesting than making a nice presentation, you know what? There's no— here's the business point. There's no excuse to not be prepared. There's no excuse to have a dumb sales call. There's no excuse for the stupid, uh, SDR email. It's, it's unacceptable with AI, right? I wrote this one up that I got from Slack on SaaS, or if people wanna read it, it was literally one of the worst outbounds I got from Slack, right? Like any AI could do better. There's no, there's no, it didn't know, it didn't know any, it misspelled Slack. It didn't know we were a customer. There's just no— going to your point, there's no— everyone may have a great PowerPoint presentation now, but there's just no excuses. So, um, but, but good comments. Thank you.

Cool. Yeah, we got a few more questions. I'm gonna go to Girish next. There's a question about SDRs and AI SDRs. Let me find you here.

Thanks guys. Thanks Jason M. Lemkin and everyone for your, uh, time. So, uh, Revify is a small data management startup. So if we have built out our SDR team and experimenting with AI SDRs. Jason, I've read your recent thing about getting blocked, AI SDRs being blocked, right? My question is very specific. I'll give you a couple of contexts. Sixth Sense, Latane said, hey, they kind of reduced SDRs, human SDRs last year. And then later in the year, they actually ramped that staffing back on because the—

Yeah, that's the boomerang we're seeing. There is a boomerang effect.

So, so we're a smaller startup. So what do you— what's currently working from your perspective? We don't quite have the scale or volume. Like if I'm— we are a startup, we are trying to focus.

Well, look, let's step back. First of all, obviously all the, all the, all the news about AI SDRs has blown up. I am— Artisan is a big sponsor. They're one of the vendors. And honestly, I'm not an expert in the vendor, but we're going to do more with them at SaaS Journal May because I want them to show us all the data. I want them to do more. I'm going to push them to be like, to do even more and show not just, not just get customers, but actually show us everything, how it works. Do it on our AI demo stage. We're going to have like 150 demos really digging because I don't know all the answers. What I do know, though, is just a couple of things. One, every product's getting better. Every product's getting better. Two, in the SDR space, and VCs are guilty of this too because they don't know what they're talking about. Everyone overpromised. Just we saw the slides before of what happened in the contact center. Everyone said you can resolve 60 to 95% of issues in support. Really, it's 20 to 40. But you know what? 20 to 40 is a lot better than zero, isn't it? Especially if it's a top 10 SLC SAT score. So AI SDR is way overpromised. Many VCs haven't done sales. They don't even know what they're talking about. Of course, AI can do it better. So there is a misalignment. We oversold these products, we overfunded them. But it doesn't mean, one, that they aren't very useful. They already are. And two, the more important point to be redundant is like we don't have the people to do the work, right? Going to— so you're gonna have to use an AI SDR. So I don't have the answers today, but I would just be— I would think of AI SDR as like Canva.

Mm-hmm.

It's a tool just like Salesloft and Outreach are great tools too, right? It's a tool. And if you expect to, if you expect to, to fire up Canva and, and it does have, it, it, it, it does have a copilot and expect it to design your entire website perfectly, you're gonna be very frustrated, right? But we use Canva every day. We love Canva. We're gonna have the CCO Canvas session. Have expectations that you're gonna have to figure out outbound yourself. Just like in sales, you're gonna have to run it. You're gonna have to figure out a script that works. You're gonna have to figure out, you're gonna have to help figure out who the competitors are. You're gonna have to figure out how to make each email great. Outbound works well today, but every email, every out, uh, interaction has to be S-tier. And you're gonna have to figure out how to make that tool do it for you, not expect the tool to do all the work. That's the— it's not helpful, but that's the best advice I have is think of these tools as Canva, not as magic. We're thinking of them as magic. And I don't know that they'll ever be magic, but they're not magic today. And it's okay. It's okay that they're not magic. 20 to— making your outbound performance 20 to 40% better is probably what you can expect. 20 to 40%. So if your outbound is performing zero, what's zero times 1.2, Amelia? I think it's still zero, isn't it? So making a terrible outbound process 20% better, which is the problem with the promise of some of these products, is not gonna work. But if you already have an— this is like anything in startup go-to-market or sales. Like as founders, you've gotta get it working yourself. You've gotta close the first 10 customers yourself. You've gotta find a marketing channel that works a little bit. You've gotta find— get outbound going a little bit, and then you bring in the tools and systems to make it better, right? And scale it up. And we're expecting too much, but That's the best I could do there. Thank you. Cool.

All right. I'm going to go to Edward. I know you've got your hand raised. I'm going to add you as a pin.

Hello. Sorry, I was caught— I wasn't ready to go to the button. Okay. In short, the whole thing people say around the building looking forward, as in not AI models are going to get better. In reality, like, what does that mean? So for context, CoForm, We help companies who have like implementation, delivery, integration, cut go-live delays, sell a lot to solutions, presales teams. Is it just looking at everywhere we use AI and going, this will be better in 6 months, this will be better in a year. Let's just plan for that to be the case. Is that what you kind of mean with the—

Look, I don't know everything, but here's how I would approach it. I would approach it two prongs. First, like figure out Everywhere in your product AI should be. Look at the competition, look at adjacent products, look at others, and pick a framework and a system and just go do it. Go do it. And if it's not perfect, like support, pick a framework and just, and then even if you don't do much, if you do a little bit, you will benefit from the underlying improvements in AI. Okay? So get it going now. Get a, just get it going everywhere there is a place you could do it today, and then you will benefit from the improvements in the models and all the rest without having, in some cases without having to do any work, without having, I mean, literally even our AI, because it, it every day it picks up new SaaS content, it gets better every day and we don't have to do any work and they'll release new features and they'll add new models. I don't have to do anything for our SaaS or AI to get better. It literally gets a little bit better every day and I think every quarter will get much better. I don't have to do anything. I just had to pick a vendor. I just had to pick a vendor and got better. So, so that's the one thing you should do now. So stop waiting. Pick a framework, pick a platform, deploy it now anywhere, anywhere anyone in your broadly competitive set does at a bare minimum. Two, this— and this is the one where I do have fear— your team should be kids in a candy store. They should be talking every— at lunch, at dinner, they should be hacking. They should be talking about what they— they should be excited about this. And I will tell you, I think the majority of SaaS companies I talk to, their engineering product teams are not excited about AI. It's more work. It's more work. I already have to keep, I already have to launch that, that, that Salesforce integration. I already have to do that other thing. Ugh. And now you want me to do this? I don't wanna do it. There's still a lot of that. There's a lot of CEOs being excited on social media, but I see a lot of teams, uh, sighing, sighing because it seems like more work and it is more work. But I only wanna work with founders. I only wanna work with companies. I only wanna work with teams where they're so excited about what's happening that they can't help themselves. They're a kid, kid in a candy store, right? This is— and there are moments in time in the internet when it's— for folks that have been around for a while, I'm really gonna date myself, but when I was super early in my career and like everything on Ajax and Web 2.0 came out and you could run stateless things in the app, our jaws just dropped. That's why I started EchoSign. Cause I'm like, you know what, back in the day, I mean, I'm really gonna date myself, but for fun, when we started EchoSign, DocuSign needed printer drivers to work. It was a Windows-only product that needed Windows drivers. And we looked at just where the web was and we're like, you know, you can actually sign a contract in the browser now. And our product was terrible, but everyone had so much fun, right? Then there were moments when mobile came out, when all of a sudden everyone couldn't believe you could do this stuff on the phone. It was fun. There were moments when WebRTC worked and you could do voice in the browser for some folks. There were mo— then like, this is one of those kid in a candy store times for the best engineering and product leaders. If you, if you, if your team does not feel like they're a kid in a candy store, with the things they can finally do, you have the wrong team. That's why Jack had to lay off 1,000 people at Blocky. It's the wrong team. It's the wrong team. Cool. All right, we got time for one more, Amelia, if we have it.

Yeah, we got Patrick. Um, so let me bring him up and then we can wrap from there.

Can you hear me?

We can hear you.

Okay, good.

Yeah, so that should be a quick one. So you, you talk about the, the expectation from the market, right? So for instance, like in our case, we do is mass training with, with videos for skilled workforce, for blue collars, right?

Yeah, a well-established old category that's got to be much better with AI.

Yeah, so, but then my question is, we have competition, of course, and how much should we promise? Because the competition is promising a lot, like I said, a lot of promise that people cannot fulfill. So the expectation is high.

So should— this is it, I don't know, we are behind.

So any tips there?

Listen, this is another— I probably, if this was a slightly different presentation, that would have been high on the list. What do I do when my competition overpromises in AI? Okay. As I said, when I look at— because the contact center was so early, because I can see what happens from a year out, right? They're, um, yeah, like Maestro QA automates QA with AI and not QA. They lost a 7-figure customer due to overpromises from a competition that they just got back. They just got back because it didn't work. Okay. But they lost the customer because of overpro— so do you— they lost a 7-figure customer and they got it back. They took the high road. They took the ethical approach. They said, this is what our product can do with AI today. This is what it can't. And they lost customers because the competition did it. Right. My guess is 11X is a great product. I haven't used it. My guess is used right, it's a great product, but they got into some trouble in the media because they overpromised. Right. What do you— but if you don't overpromise, they're going to go to the other vendor. I wish I could tell you the answer, right? Um, I— all like, all I know, and I don't know, this is a question for the ages. I know that you have to overpromise by about 20%. You have to, you have to get your team to deliver, and you have to treat customers the way you'd want to be treated. What would— what don't— you can't lie, you can't do this. So there's no— this is a challenge for ethical founders. That are deploying AI but are not necessarily absolutely at the cutting edge themselves. How, how, how, how much do you exaggerate? Right? I don't know. But I, my guess is any more than 20%, it's going to bite you. That's all I know. And, and even if you lose deals, which is painful, I don't have the answer. But I do think if you treat customers the way you want to be treated, at the end of the day, you'll come to a better outcome than if you don't. So. All right. Thanks, everybody. I think we're at the end. Amelia, are we out of time? Yeah.

We're pretty much out of time.

Okay. Appreciate, appreciate you having me again as the guest at Workshop Wednesday. We'll do more. They're gonna be really tactical like this until Annual in May, and then we'll probably take a break. And then when we come back in June, we'll take a lot of the speakers from SaaStr Annual that were great, and then we'll bring them back for an encore appearance with AMAs and Q&As after. So we'll, the seasons will change in Workshop Wednesday, but thanks everyone for coming and we'll see you next week. And keep at it. Remember, at least be as good as the competition and no Debbie Downers on this AI stuff. The bar The bar is up. Let's, let's all meet, meet the challenge. Thanks everybody.