Latest / Unstuck Pod / Slow Down To Win (Courtland Goldengate and Jamie Schneiderman)

Transcript

Slow Down To Win (Courtland Goldengate and Jamie Schneiderman) · Unstuck Pod

• Jamie: Hi everyone, welcome to the Unstuck Pod. I am joined today with Cortland Goldengate. And Cortland, welcome first of all, glad to have you here. ⁓ And love have you share with everybody your journey, ⁓ what you're now, everything. I think there's a lot of interesting things everyone would like to hear. Courtland: Thank you. Sure. OK. Well, I started in kind the B2B SaaS startup world back in 2007 at an ad tech company. That company ended up selling for $300 million. So very early in my career, got the experience of early stage startup to then going through the acquisition process and the integration process that ⁓ happens that. I actually started out in finance, but ⁓ I did a kind of a side project for the CMO. at that company and that side project was gathering data from the company's ad server. And we found that looked like it was indicating that there was going to be a real estate market recovery based on the price that advertisers were willing to pay for website ⁓ inventory that was real estate related. so that thought leadership report. It actually got picked up in the New York Times, like the physical newspaper of the New York Times. like as a junior entry level kind of marketing guy, that was my first experience with marketing and it was very exciting. I kind of through that really got the understanding of how much reach you can have with marketing and how much influence you can have on a company and its trajectory and its awareness and all of that at a very visceral level. So, Did that I worked at many other companies. Open DNS was in. So I had experience taking a kind of a consumer grade web filtering app, repositioning it. and kind of rebranding it as the enterprise network security solution for kind of predictive security. And so got a real kind of view on that world, the cybersecurity world. And then my next job after that, I was poached by Dave McJannet to join GitHub as like to do something similar, to go to GitHub, which was at the time ⁓ predominantly was predominantly, most of its revenue was individual developers and their individual accounts less B2B. And so was still a series A company at that point in time. he me over to try and reposition it more for the enterprise and do that sort of marketing for GitHub. That was an interesting experience. GitHub was a tough place to work, a tough place to be a salesperson or a marketer or anyone that wasn't a developer, I would say. And I had a culture that was ⁓ I think, due to a scandal from the previous CEO, ⁓ Preston Warner. All stuff is public. I'm not sharing anything I'm not supposed to, I don't think. But it was a pretty frigid culture when I joined, which was bizarre to come into seeing, you know, walking in, seeing a replica of the West Wing. as the entryway into the office or not that West Wing, sorry, the oval, the oval office. And then seeing a physical statue, like a stone cast, like statue of the Mona Lisa octocat three ⁓ bars at the, the office. ⁓ ⁓ it was just a ⁓ different world. Anyway, so ⁓ I, when, when I got into developer tools and marketing developers. Jamie: office. Courtland: tools and I came in thinking I knew about marketing developers because I had been marketing to security engineers and network engineers and sysadmins and I figured out how different could this be. Realized it was very different, very, different. And so yeah, learned that the hard way, I think, but ⁓ those learnings and joined GitLab when was fairly early on and helped get them ⁓ up off the ground. Jamie: Really different, yeah. Courtland: That was a wonderful experience, great company, incredibly well run. had an exceptional CEO. I can't speak more highly of Sid Sibrangi and management style and what that enabled for the business. It was a really interesting case study to see GitHub on the one hand with every advantage in the world, brand, money, ⁓ talent, de home of open source software, and GitLab with none of those advantages. Kind of a me too company and just through effective executive leadership and strict kind of, inflexible leadership, not popularity contest style, CEO leadership, seeing what a difference that made. And know, as a result you know, GitLab ended up going public where ⁓ GitHub not get its act together to do so. And not only that, but GitLab went public for twice the value that Jamie: Mm-hmm. Courtland: Microsoft bought GitHub for, which is just insane if you think about it, if you really like look into the details of those companies. In my view, GitHub should have been a you know, $100 billion plus public company and it ⁓ ⁓ didn't execute in some really key areas where it needed to. And that's, think that was my first kind of exposure to AI because when I came into GitHub, I realized that Jamie: This is insane. Yeah. Courtland: I was seen as something that would of course be automated at some point in the near future. And marketers, business people were seeing us like, okay, yeah, machines are gonna do this better than humans anyway. Like, what's a big deal? He's got a title or something he's supposed to be important at this company. But ⁓ in reality, he's just a couple of years away from being automated out of existence. So at the time, Jamie: Mm-hmm. Courtland: GitHub was working on what they called autonomous coding, what is now GitHub Copilot. so that was back in 2015 that they were working on that and thinking about that. yeah, they knew where the world was going. And they for a long time. Again, like some of the most talented individuals at the company that I've ever worked with, just cultural and organizational dysfunction holding it back, unfortunately. Jamie: Crazy, crazy when you think about that, yeah. Courtland: And then I had an experience taking a company public, Fastly, a CDN slash edge computing platform, which is now kind of getting an interesting resurgence with the prevalence of AI agents and all of the web traffic that that creates and the impact that has on usage-based companies that make more money as people spend, people or anything, spend more time on the internet, including AI agents. then I worked at a data science startup, ⁓ this company called Streamlit that was doing something important at the time, was taking, we're giving ⁓ data scientists ⁓ ability to, without writing a line of code that wasn't Python, create a full data app. so communicating data insights through apps that were very easy to build and a familiar programming language for them. And that literally so easy to build. I could build web apps and did build some web apps using that service. That company got acquired by Snowflake. And so, you know, I had joined Streamlit because I really wanted to be back at a small startup. then I ended up at even, yeah, so, you 1200 people at Fastly as a public company, that's too big for me. Jamie: And then all of sudden you're in snowflake. Courtland: joined 50 % Streamlit and then ended up at 7,000 person snowflakes. Okay, I mean, no complaints. Great experience, great people that I got to work with and meet and all of that. But it was a pretty big departure from what I was looking for. But at that point in time, this was, I wanna say, 2022, LLMs were understood at this point and their importance were... Jamie: Yeah. Courtland: understood ⁓ at Streamlit. mean, the three founders, I think, are all like PhD data scientists that met at Google kind of when they all worked together. And so they were way, way, way ahead of the curve and ⁓ helped ⁓ of the other team members get ahead of the curve as well. And we actually saw the most popular Streamlit app that was made ⁓ this thing called GPT-0, which was identifying whether or not text was AI generated or not, you know, intended for like teachers to make sure that students weren't cheating. And this was very early on, very, very early on. And so they got huge press coverage for that. as a result, that app got a ton of traffic. And that's how kind of ⁓ I saw that, okay, this is gonna be a big deal. I started experimenting with what could be done in ⁓ LLMs for fun. Jamie: Yeah, sure. They were fired. Courtland: At work, I introduced a new way to get derivative marketing content to Snowflakes product marketing org, essentially defining what the different product marketing assets You know, what is a competitive battle sheet? What is an industry one-pager, whatever, and then feeding it a bunch of case studies and a bunch of information and then having it then output something that was maybe focusing on a different industry where there was no industry one-pager or a different competitor where there wasn't one and seeing how quickly you could ⁓ create content was like 80 % of the way there in moments. that was exciting. And then on the personal side, I was trying to basically play Dungeons and Dragons, ⁓ the role playing with an LLM and was blown away how interesting and fun it could be to just be taken through a story with the LLM acting as the dungeon master. And so, so I left Snowflake start my own company, Tall Tavern, which was ⁓ to be a AI game master. But I found that that community didn't really want that. ⁓ were much more inclined to like, you know, have a pencil and paper and do things in their old school way. And community is so full of like ⁓ these, know, vibrant artists and writers, and oftentimes people that are selling their artwork and selling their writing custom content for these games. so know, AI was an existential threat to them ⁓ and identity and what they liked and all of that. And I think as a result, the sentiment against AI became very negative, like hostile even like AI folks are not welcome here, you know, like type it was like the like the Jim Crow south, but like for people that didn't think that AI was the devil. Jamie: down. Courtland: ⁓ Anyway, that, you know, totally valid perspective for people to have, but it took all the fun out of that company for me and out of building it. Jamie: Can I, sorry, just a question for you. Like, so you say, totally valid in view though, like, ⁓ you it as picking your head in the sand a little bit, also around AI, like you've sort of looked at it and said it's coming and on the marketing guy, right? So the, so the guys at GitHub said, this guy's got a short life in terms of We don't need him for much longer. Nobody will. You went. I'm sure that that's true, but I'm going to lean into AI and try to figure out what things I can do rather than ban AI. But the guys that you're talking about basically said, you know, I, this AI thing is terrible. I'm going to stay away from it. So the right approach, right? Is it, ⁓ coming, it's coming, whether we like, whether like it or not, no matter what it's coming. So doesn't it make more sense to kind of adopt it and kind of lean in where you can. Courtland: Yeah. Well, I think that's that's a something that each individual has to determine for themselves. I can understand and appreciate like the high minded abstract view of like, this is actually not a good thing for humanity. The risks are too high. I the benefits are even higher. But like, can we really ⁓ that kind of a gamble with human civilization? Like there are good arguments to be made there. And I do not begrudge anyone that has that point of view. Jamie: Mm-hmm, sure. Yeah. Courtland: My point of view is, okay, yeah, in theory, can ⁓ where someone like that is coming from, but in reality, Pandora's box is open. I mean, there's too much money to be made. This is not something that's gonna ⁓ And so, you know, you can either participate in it or sit back and ⁓ the world change and be totally powerless ⁓ do anything about it. So, mean, I choose to participate. I think that's the rational choice to make. unless you feel like you can like the entire ⁓ and be like become the Greta Thunberg ⁓ like, you know, AI antagonism or something. But the bottom line is there are huge benefits. I mean, this this is what's going to solve, you know, cancer, this is what's going to solve like big human problems. And there giant risk. I mean, ⁓ a neural is essentially ⁓ an alien that we do not understand. And we're just kind of like poking around at it and seeing if that alien brain produces things that are more pleasing to us. There are some ⁓ like dire like disaster scenarios with that that people should be ⁓ very of. And AI I think, and governance is incredibly important. You can't, I mean, it would be so, irresponsible to just let AI do everything and be totally out of the loop and just trust that it's going to be right and trust that it's going to like, create the thing that you want and follow your your intent. So I that the have got to be in the loop, they've got to be a part of the part of the process. And it needs to be like thoughtful and intentionally designed. Let machines do the things that machines are best at let humans do the things that humans are best at. ⁓ And put humans in the loop deliberately to avoid a situation where you're a company and you don't even understand your own code base anymore, and then you have an outage and you can't fix it. ⁓ not there yet. And it ⁓ tempting because you can get so much done so quickly with AI to just ⁓ it for everything. But it irresponsible. Jamie: Yeah, that's problem. sure. Yeah. Courtland: You should use it for a lot, but you should be very involved and you should always remain in control and have the knowledge of what you're doing and what you're producing and not just completely outsource all of that human discretion to machines, in my view at least. Jamie: Yeah. No, no, which I think is fair. That's a fair reasonable ⁓ perspective. Okay. So sorry, I took you also off and we can go way deep into that world, but I'm, I wanted to talk more practically about, what you're doing now and also how, excuse me, how, utilizing AI tools, new thinking from a marketing perspective, because I think that's directly relevant everybody who's listening. Courtland: Yeah, well, something that I haven't heard a lot of people say or write about, but that seems pretty obvious to me is that if you're a founder and you're looking to like build a durable company, then you should be looking for people that have had experience way, way beyond the stage that you're at. So one of the things that I can do for my company case, which is a series a startup, is I can build the capabilities of a public company's marketing department, a marketing department of 100 people, but do that with four people and just a shitload of AI agents doing a lot of work. And so now the lowest kind of marketing that makes sense to me is a director and have a director that instead of managing people, manages a bunch of AI agents. And so that I think is gonna be like a pretty efficient model and it's pretty like readily transferable to people that have operated at scale and operated at startups before. So I think people that have both ⁓ experience in startups and at larger companies, if they're not in hot demand, they should be. think that a lot of people's attention is more on like, ⁓ wow, I can just vibe code an entire company and that kind of thing. And a lot of people think that, a lot of people think that. And I do worry that that's gonna lead to just this giant influx startups that aren't gonna clear the bar, aren't gonna become durable companies. ⁓ just gonna be competing against everyone else that can really quickly spin up an app. So I may a bit biased, but I think go to market now is a real distinguishing ⁓ factor. Jamie: Mm-hmm. Courtland: And I think experience and taste, design thinking, systems thinking, these are the new skill sets that need to be even more emphasized at companies if they want to actually become viable companies. I don't think the, vibe coded this in a weekend and now I want it to go viral approach is going to work for ⁓ most but we'll see. Jamie: Well, and I will tell you that in my experience, I talking to all these companies that have something good, so they've built something good. But, you know, ⁓ the world gets noisier, as more things are out there, as it's easier to your point to vibe code a whole application and, you know, a weekend or push, get new stuff out, the reality is differentiating between something that's just there and something that's really, really good. And you refer to it as durable. becomes harder and you need to be that much more focused in terms of what you're doing. And obviously have, like, as you said, the systems and structure and, ⁓ and approach that is continual, and that can scale with you ready to go. Cause other, it's too hard to compete otherwise. So yeah, for it makes a lot of sense. So, okay. So where, where let's talk little bit about, cause, cause people hear, AI and marketing and building systems and stuff and it feels very overwhelming and you know, obviously you have a lot of experience doing it so you can jump in and do stuff. Where would you start? So if somebody was starting and they wanted to build out some marketing, they're doing some things that are working, some things that aren't, they're probably missing a ton of stuff. Where do you start when you're looking at this? Courtland: Right now, the place to start is Claude by Tentropic. the reason is that they have gotten very good at ⁓ writing code and building ⁓ apps. So I can now spin up a dashboard or a interactive ⁓ app replace ⁓ a presentation in less time than it would have taken me to write an email before. And so... ⁓ Jamie: Mm-hmm. Courtland: It's incredible power. can it can sequence a number of different tasks. You give it kind of that one prompt like I want, you know, a website that this color palette and you just attach that color palette and it uses this brand voice, this style guide. You know, add that attachment that, know, takes insights from these customer case studies. Add those attachments. that ⁓ you know, XYZ information architecture, and it can just build the website and it won't get it right the first time then, but then you can kind of rip on it. And so I would be, I would be thinking about like, what does it mean if I can create a data app faster than I can create, like create an email ⁓ probably the fastest place to start for a marketer would be use, use Claude to create a website. Jamie: Modify, yeah, sure. Courtland: Or, you know, use Claude to try and tackle your content backlog. If you've got a content backlog, see how far along it can get you. can you pretty far along. ⁓ if it's just prototyping in a way to like share ideas. Because historically treated software as this ⁓ very, precious thing and it's so hard to create and all the rest of that. I think now it's more like a medium of communication. It's so easy to produce that you can just, like, if you have an idea that's not really thought out, you can go, you know what, I wanna just, you know, vibe code a data app in a couple of minutes to share this idea. And that's its sole use is to communicate an idea. can be ephemeral, you know, like an email, because the that goes it is so low. Jamie: So I actually want it because actually it's interesting. The point you just made is the one thing I wanted to ask you about ⁓ the barrier to entry or as the ability to generate stuff, spit it out becomes easier and easier and easier. How much a strain does that put ⁓ on the thinking that or that differentiates quality from lack of quality? on the market, certainly on the marketing front, I guess is the Courtland: Yeah, well, internally, case, we've got this saying that AI can't QA itself, which is a bit hyperbolic. And obviously, people use AI to write code as well as test code. But you probably don't want the fox garden, the hen house. You don't want ⁓ someone to grade their own homework, so to speak, because if you use an AI tool to write code and then use Jamie: Yeah. Courtland: that AI tool to see if it's good. It's kind of got an incentive to tell you, yes, it's good and be very, very good at persuading you that it is good and that it does what you want it to do, even if it doesn't. So first superpower that LLMs developed was the superpower persuasion. So if you use LLMs a lot you know, in a matter of weeks, you'll think you're a genius because respond back to you in ways that are optimized for continuing the conversation. And the LLM the user off or challenges them or you know, in a way that they feel threatened by or whatever they're not going to use it. And so you've got all these ⁓ LLMs now are overtuned to being like super polite and quick to agree. And if you push back a little bit, they instantly cave and agree with whatever kind of point of view you're taking. It's kind of comical if you have had the experience of doing that. Jamie: Yep. No, I listen, I've seen it. And I think it's very hard to, uh, I, often, I'll start often with, uh, if I, like, if I'm doing a prompt, which is, okay, I need an honest assessment of what, you know, the following XYZ, will this work for blah, blah, blah. Again, I, I think it, I think again, picking and choosing what you're asking for, where that fits in is obviously important too. Um, I'd like to think that I have enough experience that I have a pretty good sense of stuff, but I can see, I see people who are a lot younger early in their careers who don't know that are asking questions and going, this is what I was told. It's great. It's like, that's scary. so from an output perspective. Courtland: It is scary. I mean, junior employees at companies are a giant risk vector now in companies that have ⁓ adopted AI and haven't thought through AI governance and guardrails because you're have a lot of young, ambitious, you know, people in their twenties looking to like really get their career up and off the ground. And they're going to be convinced that their way of thinking and doing things is right based on interacting with LLMs. And if they don't have the experience and the wisdom gained through tough experience, then they're going to have blind spots, major blind spots. And I think that's a real risk and a growing one. To mitigate it to your point, can do little things like ask include your prompt, like link to your sources and then Jamie: Yeah. Courtland: actually check their sources and like double check and verify because LLMs can sometimes hallucinate sources and links that actually won't go anywhere. You know, it's like, ⁓ this looks like a, you know, a list of sources, even though it's utter nonsense, but you got to take the trust, but verify ⁓ approach, especially when it comes to big decisions, one day, like one way doors, like decisions that are hard to undo. got to slow down really verify that what you're doing is matching your intent, ⁓ the output of the LLM is matching your intent. Jamie: No, and by the way, you said the magic word or the magic term, which is slow down, which of course is counter to everything. Everybody's thinking about as it relates to AI faster, faster, faster, more faster. And you're like, Whoa, hold on. Not necessarily the smartest approach to this. as you go for. Courtland: Yeah, to people that think like that, I might frame it as sometimes you gotta slow down to speed up, or you don't wanna be moving too fast towards a cliff. You don't wanna be moving too fast in the wrong direction, because that'll slow you down to course correct. Jamie: Yeah, well, so. no, I ⁓ So, OK, so I mean, you and I could chat all day. This is great. I love it. But but at some point we have to wrap up. let me let me let me let me throw something your way. One question that want you to sort of share with you like the audience, which is if you were going to give ⁓ folks advice about on your experience, what seen in terms of the evolution of AI, how it's impacting go-to-market marketing, et cetera. What do think folks should be thinking about over the next year or two? Recognizing technology is going to change get all that. But how should they be thinking about what they do as they're trying to incorporate this into their businesses to have a significant impact on the future? Courtland: Yeah. Number one embrace double loop learning. Do not operate from the standpoint of your past kind of assumptions just being true. You're ⁓ have to continually have hypotheses and test them and validate them. And if doesn't work out, you gotta revisit the assumptions because it is a different world. ⁓ And you were to come into a new company, and apply a playbook from five years ago, it would be nonsense. would be like terrible leadership and ⁓ decision-making because has so fundamentally altered the landscape. And it might of course mean AI. So I think people need to ⁓ have a kind of a healthy amount of skepticism and kind of almost like assume the AI is wrong until you prove that it's not, until they're really more comfortable with it and understand some of the quirks that LLMs have in terms of hallucination, ⁓ terms of over indexing, on trying to please the user rather than get to the correct answer. But yeah, I mean, you got to learn it. It like 10x. or more individuals productivity if they learn how to use it well and learn how to use it right. And so if you don't, you're gonna be out of a job like in the B2B sector. If you don't understand this, if you're not setting up multiple agents to do work for you, if you don't have a executive assistant as an AI agent that you've configured to make sure you remember everything that is important, you don't drop any balls, all of that. Jamie: You mean like not send out, like not send an invite for a, for a podcast like on the right time. Sorry. I just, I'm just kidding. Courtland: Cool. Yeah. Well, I thought that was like you were trying to relate to me. You're like, ah, this guy's a scumbag and like this organized guy. Let's pretend like I'm like him. Yeah. mean, human brains are certainly fallible. As are LLMs. LLMs have a finite memory. So if you're using an LLM, you you got to store long-term memory in a thoughtful way that's also going to be legible to you and other humans down the road, if you wanna figure out why it's doing what it is and how it's coming to the conclusions that it does, wanna have that full kind of traceability if using AI, because when things do go wrong, you're just gonna be thankful that you have it. Jamie: I think, by the way, I think that's fantastic advice. Everybody talks about go, go, go. And you certainly are recommending go, go, go. You have to go there. But I think the, but when you go, go slow to go fast is like the, and, keep being thoughtful about you're getting, what it is, what if like a lot of those things. I don't enough people are asking those questions or thinking that way. So that's really terrific. So, thank you. This is great. So really appreciate it. For the sake of how can people find you, reach you, if they want to. Is there some way that they can? Courtland: Uh, I mean, probably the best way would be on LinkedIn. I'm pretty sure I'm the only courtland Golden gate in the world. So I should be pretty, pretty easy to, to, find. Um, yeah, that would be the best way. I'm not like an influencer. I'm not looking to build an audience or do any, anything like that. But if you want to have like a one-on-one chat about a cool topic, you know, drop me a line. I'm a real human. like to, I like to chat. Jamie: Okay. Love it. Cortland, thank you so much. Really appreciate it. It was really fun. And everybody else who's listening, watching, until next time, thanks. Courtland: Thank you.