When Zapier called a “code red” on AI, it marked a turning point for the company. In this episode, co-founder and CEO Wade Foster shares what that urgency looked like in practice, how it reshaped priorities, and why changing culture is harder than shipping new features. From rewarding experimentation to building curiosity into the organization, Wade offers a candid look at leading through AI’s rapid rise and making adoption stick.
Wade Foster, co-founder and CEO of Zapier, joins Henrik and Jeremy to talk about how AI is changing the company from the inside out. He shares the moment Zapier declared a “code red” on AI and the steps they took to turn urgency into action — encouraging more experiments, removing tolerance for inaction, and celebrating wins along the way.
Wade discusses his own AI use cases, the importance of internal examples in driving adoption, and why duplication of efforts can speed up learning. He reflects on the leadership challenge of guiding a 14-year-old company through cultural transformation, balancing productivity gains with employee well-being, and preparing for a future where AI agents work with each other.
This episode offers a clear, practical look at what it takes to embed AI into an established organization, and keep it moving forward.
Key Takeaways:
Zapier: Zapier: Automate AI Workflows, Agents, and Apps
LinkedIn: Wade Foster | LinkedIn
00:00 Setting Company Culture: Rewards and Tolerances
00:43 The Rise of AI at Zapier
02:19 Wade's Social Media Presence
05:06 Challenges in AI Adoption
07:32 Personal Use of AI: Health Tracking
10:21 Business Applications of AI
13:34 Automating Repetitive Tasks
20:35 Voice of Customer Program
24:26 Customer Brief Generator
33:27 Code Red: Embracing AI
35:32 Subtle Encouragement and the Impact of GPT-4
36:38 Code Red: A Turning Point
36:51 Embracing AI: From Fear to Familiarity
38:13 The Journey to AI Adoption
39:11 Challenges in Organizational Change
40:41 Managing Resistance and Encouraging Experimentation
43:55 Building a Remote Culture with AI
46:29 The Future of Work and AI
48:33 Agent-to-Agent Communication
51:32 The Importance of Duplication in Innovation
56:43 Final Thoughts
📜 Read the transcript for this episode: Transcript of Inside Zapier’s Code Red: How CEO Wade Foster Hit Pause to Reinvent for AI
[00:00:00] Wade Foster: if we wanna be like very reductionist about how you set culture inside of a company, it's what do you reward? What do you tolerate what do you not tolerate? And so to use this as an example, like in action, wouldn't tolerate at bats. We're like, Hey, that's good. And And then what we reward is like successful attempts.
Hey, I'm Wade. I'm one of the of the co-founders, CEO at Zapier. Zapier's, the most connected AI platform, and I'm looking forward to with y'all about how you can can transform your organization to take advantage of ai.
[00:00:34] Jeremy Utley: Okay, Wade,. We've been really stoked to talk to you, really stoked. Um, What are you seeing are the common obstacles that other who are maybe earlier in their adoption curve than you are? What are you seeing are some of the challenges that folks are facing? And I'm really curious about, particularly non-technical organizations, you know, service organizations, because I think a lot of people read this stuff and they go, well, of course Wade can do that as Zapier. I mean, they're a tech company, right? you know, about like, when you're talking about phishing before we hit, , record, uh, you, you're putting lines in the water. What are you seeing from others in, in the water?
[00:01:10] Wade Foster: I think there's a couple things that we're noticing. First and foremost, most of the refrain around AI is, is the It's when is a GI gonna get here, you know, AI is replacing jobs. Uh, you, you know, if you don't get into ai, like your job might be at risk. Like there's that noise, the volume is like turned up to, you know, know, 11 there. Then the like, actual how to specific stuff is a lot harder to find. Um, and it's the, the sort of refrain is, is like, like, well just get in there, try the tools, figure it out yourself, et cetera. Mm-hmm. I can appreciate that because like, you know, I, I kind of of like to dabble and experiment tools myself. But, you know, I, I know a lot of folks in my in my life that just, they need a little tips. They need some inspiration, they need some guidance. And it feels like, you know, with AI in in particular, we're in such the early innings where there isn't like, courses on this stuff. There isn't training manuals, there isn't a book to go read. Um, you know, you can talk to the prompts and you can talk to Theis themselves, and they'll often coach you you on some of this stuff. But a lot of how I've gotten better at this stuff is just watching other people use it and going like, oh, what'd you do there? Like, how did you, like, how did that work? Why'd you do you do it that way? And then trying it myself and, uh, you know, so I, I think the, the big gap for a a lot of is just simply like, what do I do I use it for? You know? I, I've tried chat dPT and like I know that it can come up with a good poem and I I know help me summarize an email, but like, I'm having a hard time, like taking it to the next level, particularly for for those non-technical like co-generation for engineers is like, you know, people, those folks, like, they they really got it dialed in for the most part. But for for the rest of us, there is a little bit of like, well, what is is the use case for me?
[00:02:58] Jeremy Utley: Uh, what do you tell, like your family, example, non-technical friends even companies, what are the, one or go-to try this kinds of interactions you recommend?
recommend?
[00:03:09] Wade Foster: Huh. Well, uh, i'll tell you like one of my my favorite use cases that I have is I have a really long running chat with chat g bt now where I basically took all of my, like, health data, so, you know, workouts, , stuff I I eat, um, sleep patterns, you know, blood tests, like all this sort of stuff. Um, I'm, and and I'm not a tracker, you, like, I I don't, I don't like track this stuff religiously, but the describing it, you might think, oh, oh, like one of those nerds who like obsesses all this stuff. Not really.
[00:03:41] Jeremy Utley: was thinking that you you read my mind.
[00:03:42] Wade Foster: yeah, I, I'm not really like, I just have this data because like, kind of, you know, the app that I that I use for workouts like kind of keeps track of it the, you know, you get a get a blood so it's like, okay, you you got this stuff. So I just uploaded it it all to chat GPT and I I said, Hey, given what you know about me, like tell me, like with, in specific details, tell me me how I improve my overall health and and wellness. And boom, just like spits out like a whole like recipe of things to go try and, uh, you know, about every, , three months or so now I'll just come back to that chat. I'll upload, refresh it with all the, fresh data and and say like, run it again. Tell me me what to do.
[00:04:21] Henrik Werdelin: What is a concrete thing that it suggested to you that you didn't, that you didn't kind
[00:04:26] Wade Foster: well, so like I didn't have, um, I've not been like a big like supplements or person or anything like that. I just kind of, you know, I, I have a basic workout routine and you know, I mostly have, I don't really diet, but I kind of have like restricted a things. I always really well, so like those things were covered, but it was like, hey, you probably ought to, you probably ought ought to be taking some vitamin D, you probably ought ought to be some fish oil. You probably ought some creatine. You probably ought to do this. Uh, and I started doing it and it's like, oh, well, of, kind of, of helps sometimes.
[00:04:57] Jeremy Utley: Did the, did, did the impact up in your or you you know, if the loop on experiment, so speak, you like gT's recommendations were validated in your
[00:05:07] Wade Foster: Well, the creatine one is interesting 'cause like, that one I felt like I could tell right away because literally, I I do weight lifting is what I do. And like, I literally, I'm I'm like, oh, I can actually pick up heavier stuff now. Uh hmm. And so it's like, all right, like, I guess that seems to have worked. Um, and like I not crazy scientific, about this stuff. There's the, like, you know, uh, who's the the guy right now? Uh, Brian johnson, who's like, I, you he's like tracking everything. Like, I'm intrigued by that, but I I am not even close to Like, it's just so, so far and away. So do you. Okay.
[00:05:39] Jeremy Utley: But, so, so, mm-hmm. Yeah, I was gonna, I, I think he and I are gonna same place, which is like, you start with a health, uh, coach Nick Thompson, who's a big time runner, CEO of The Atlantic. He told us recently about his routine of uploading his, you know, uh, Strava data into Chad G PT, for example. Yeah. How do you go from something like that? So say you are a family member, or you know, a new business contact tries that to a business application. How do you start to close the gap between those two
[00:06:08] Wade Foster: Well, so I I think you start to do it for something this and you get like such a better experience than you would Yeah. Going to see your doctor or going to to talk to a, you you know, health coach and you start to go, huh, I I wonder what else I could try this for. Uh, and so, at work, like, you you know, we have built a strategy, GPT. and so this is when anyone comes to us like a one pager, or you know, three pager on like, hey, here's a proposal for an an area or an opportunity to go tackle. There's always like stuff that that people just are not thinking of. You You know, both in of like actually through the strategy, but then also just like in how the readers are gonna interpret it. Like common questions, like that. So we just built built a GPT like, you know, you feed it into it and it has like common recommendations. It's like, Hey, you're you're probably missing X, Y, and Like, you know, you didn't didn't present any or you're, you you didn't actually, uh, propose a to go in. You're sort of just saying, hey, here's some information. You know, you're, you're missing this, that, or the other. And so it people just just come a lot more prepared for strategy discussions. So now all these examples, by the the way, are like back and forth chats, which I think is like a really good, that's where most people start. The The big leap. The big next leap is how do you you actually turn this into structured automation? Uh, and so I'll I'll give you an of where this is really powerful, which is, um, something like generation. You know, if you wanna wanna do research on a company, you know, you might say, Hey, I'm gonna with, yeah, I, I'm meeting with Wade from today, so you you might have a chat that's going back and forth with me sort of like do hands-on research about this. But the kicker uh, you know, you're having to do that yourself every time. Now what if you're trying to record, I don't know, multiple podcasts every single day? That's a a lot of research. Um, what gets interesting is when you can say, Hey, I actually run the system on every podcast guest every single time. Right? Right. So here's what I'm gonna do. I'm just gonna upload Wade's contact information, and then that's gonna fire off an that's gonna run through the same set of prompts. It's gonna gather the same set of and and it's generate a report that now when I I wake up morning, I just have a prompt on on I have a, like an output on the three other guests I'm gonna have, and I can just read through this stuff.
[00:08:25] Henrik Werdelin: as a concrete sample, we do that with Inbound. We get quite a few inbound, uh, folks who wants to be on the podcast. And it used to take. Our Producer quite a lot of time to research all of them. And so she she went about building an optimization that did the same thing, went, tried to find some other podcast this person's been on, tried to look at what that, figure out what have worked well for us before and then come with a recommendation. But the question I was, I, two questions that I, I guess like, I think a very difficult to answer but, which I've been very excited to talk to you about. Um, the first is this kind of like self prompt of like, , what is the best thing a one can do to constantly get reminded that this is probably something that I should go and make a SAP full or like whatever optimization system they use.
[00:09:12] Wade Foster: I, I think the, For these ongoing automations, you, you mostly are just to pay attention during your day to the common tasks you yourself doing, whether it's you individually or the organization as a whole is doing. And And then those are the places where you ought to ah. You know, I, I'm like, you look at someone like me, like I do a bunch of interviewing, like I'm interviewing candidates for jobs, like, you know, probably, I I don't know, , five, 10 times a week, something like that. And every time, like I'm running through the same steps, running through the same process, just to get prepared for things. And so that's an example of of oh man, I, I should just automate the research process stuff and I shouldn't be to go through the same, you know, 30 minute research steps every single time. Why should have an automation for that? So you can can often look at your calendar, you can at your common and say, Hey, where are the places in here here that I ought be automating? So that's an way to start. the second place then. Is, and and this is, I find is folks, but it's to think through the the places you ought to to be doing something, but you aren't because you actually don't have. The capacity of their of their bandwidth to go this stuff. So again, I'll I'll kind of stay, stay on this like, uh, you know, researching people, uh, use case, which is one of the of the common I have to do is I, I'm often just meeting with customers, prospects, et cetera. And in the past, you know, I I'd go to an event and, you know, there's a hundred people at the event and I'm I'm like, Hey, which one of these people go spend time with? Or they there any people in here be, you know, good, sales candidates or, you know, know, prospects, or heck recruiting prospects. There's a whole just like, like, question that stuff. And And I have to judgment call. Like, do do I go talk to an AE someone in the sales org and and say, Hey, could you do research on these a hundred candidate, like a hundred folks for me? Or do do I go do it myself? The answer is like, if I ask an an ae one of two is gonna happen. They're They're gonna do a like a a shoddy of it. Because they're got other stuff to do. I got like real leads with real quota that are are way better get after, so so I'm just gonna get it done. Because Wade asked me to, that's like, I'm like, eh, that's not exactly a thing. The second thing that might happen is they might might go, well, Wade asked me, I'm gonna really, really good. So they're they're gonna go spend a of time on this way more than I I actually need. Uh, and that's gonna away from them actually closing quota, deals, So that's like a weird request. Then I'm like, well, I could do it myself, but I'm busy, so I'm probably gonna just do do the shoddy job that the did. And the reality most of the time I would just be like, eh, i'll just wing it and and just going whatever.
[00:11:49] Henrik Werdelin: But very, but very concrete on that one. How do you then, 'cause one of the things that I think that highlight is you then have to decode what is actually an interesting person for you, right? And mm-hmm. That might be different from you and you, me, but also might be different from you from, you know, fundraising, recruiting, uh, wanting to get the message out, want to, uh, reposition Sia, whatever kind of vi it is. So. What do you then do to figure out, like what is the prompt in the process that allow the AI algorithm to kind of pick up what you are actually looking for? If you stay with that concrete example, right? You're going to a conference. Let's say you go to Masters of Scale in October and like, uh, you, you're gonna talk, so you're gonna be in the green room. And so you have a high end, there's a lot of people there that you might wanna meet, but there's probably 35, 40, 50 , uh, people who are talking, uh, speaking. What's the, what's the process?
[00:12:43] Wade Foster: So, okay, this is is where it does take a little bit of effort to these things out. So first and foremost, like we know what our ideal profile looks like, so we out, hey, based on information, these types of people and these types of companies and these types of roles tend to get the the most value Zapier. Generally I'm gonna want to talk to people that like them. And so we have have prompts that outline, you know, hey, , you know, know, are in marketing this type of company and this type of role with with this many at this da, da, da. So it's a whole bunch of stuff that we've, you know, know, and said, Hey, this is closer to what we're looking for. So that's in a And so you can then, you know, know, take an email address or take a and say, Hey, go enrich this, this whole bunch, uh, out, and then go match it against examples and then score it, you know, how, how, how good of a match is this, or, or bad of a match is this for what we're trying to look for? And so say there's a hundred at the event, you could do that for every a hundred folks. Get a score out of it, sort it by the the score, and then you could start to see, okay, here's the 10 people that that I probably wanna try and spend amount of time with, and here's the the bottom 10 like, you know, I, I can safely ignore. And, uh, it sort of just helps you like figure out out how to your time. But it takes young, for us, it was like we had had to understand who's the customer profile, and then we to take the time to to like, write that down specific enough language so that, you know, feed in, uh, an email address name or you know, a LinkedIn profile or you, you name it, uh, it can go , enrich that lead and then then figure out how to compare and contrast.
[00:14:21] Jeremy Utley: Yeah, it's what you're getting at is um, the power of examples as well. To me, when you say, your two criteria are either look at your calendar to see where are you doing repetitive tasks that could be automated to where should you be but aren't, I think imagining where you should be doing stuff is D more difficult to imagine mm-hmm. for a lot of people. And so that's where something like a community of practice is helpful, where you're hearing examples where you see it, you go, oh, we could do something like that. And it made me kind of wonder if we shift gears a second to talk about the Zapier adoption journey. Were there internal uses that really kind of sparked momentum, you know? And the reason I ask this because we had Shiia, he who's the head of AI at Notion on recently, and she told us about how they were tracking internal adoption metrics and they weren't really getting lift off. And then one of the engineers on her team posted a two minute loom of preparing for a performance review. And she said, she after that Loom posted, they all of a sudden it was like just, uh, it was like it hit an asthma tote. Internal usage just skyrocketed, which to me speaks to the value. There's also a danger, but the value of kind of discreet use cases. Can you think of a couple of examples like that in the Zapier adoption journey that helped ZA and just kind of go, oh, I can do that.
[00:15:40] Wade Foster: It's funny you mentioned reviews. I like that was a that was one for us. I think think it's 'cause everybody has it. Nobody likes it. And it's just the type of thing that people are like, , how can I this better? Like, nobody thinks that that performance process is like a good one. So that's one that's like created like widespread adoption. Um, there's two other examples I can think of that have catalyzed ambition. Um, because performance reviews is, that's like a helper. Like it's useful but it doesn't actually really change the trajectory of your all that much. . The two that I can think of of that have really changed the ambition of what is possible is one, is uh, our voice of program. So, you know, Like many companies, like we're to take in signal from what we're we're hearing sales and support and on social and and you know, all the that we have. We're trying to to pull those all into like a repository. And Then we're to like have a a bunch of metadata with that. So, uh, you know, know, is this a enterprise customer? Is this a small business Is this a free customer? You know, know, all that sort of stuff to help us slice and these things. And we're then we're trying to extract all these from it. You know, these types of love the product. These types of customers seem to hate it, it, these types of features, uh, people are are craving and we we are not supporting well. This parts of the platform kind of sucks. So we have, uh, you know, a voice of customer program that does that pre this solution I'm about to talk to you about. This was like a labor love. It's just like, you know, manually trying to to get stuff certain areas. It's people like reading through notes, uh, trying to suss out themes. Um. and The quality of it. Like you're always, you're always unsure of like, like, how good is the quality of it and the speed at which you can actually synthesize stuff is is like pretty slow, so you don't actually use it as as you would like. Mm-hmm. Um, fast forward, we we had one who said, you you know what, I can do this a better. And effectively what done was she. Set up a database where she has a bunch of automations powered by Zapier that are pulling in gong transcripts into this database. They're They're pulling in Zendesk tickets all this database. They're pulling in all the social into this database. They're pulling in all the the user's research calls into this database. So it's all just getting sucked into one area, just pulling in context, pulling in context, pulling in context. Then, , she does a couple things with it. One, she has her own system where generate reports every month that synthesizes this for every team inside the You know, so pricing and packaging, here's everything we've heard on in the last month, and here's the common complaints, here's the common, uh, you know, things that people like the core product. Here's what are asking for, da, da, da, So she does does her own version of that, is great because it's like every month like clockwork, we're getting thematic results on what things look like, and we can track up up and down, are the trends on that? So So that's thing one that's awesome. But thing two is she's she's also built chatbots and query on top of these that now all of our of our product managers are designers. Our pmms, access to this stuff. So on demand, when they're a question about like, Hey, I wonder. You know, I I noticed this had this problem. I I wonder how much that up for us. They can can go in and query and say like, Hey, how does this happen? You You know, and sometimes you might find like, oh, that's an edge like, you know, know, that definitely stinks for that but like, I don't don't know if we wanna like drop and go solve this. Other times we go like, holy cow. Like this is like pervasive across the, the customer base. Like we really need to figure to do it. Uh, and then the, the second thing powerful about that is when they're they're doing that stuff on demand, they can now track here are 10 specifically who who have talked about this problem. And so they can go reach out to 'em and say, Hey, I'm working on this. Can you give us some feedback on like, this implementation thing? And so, you know, it's really been to just provide like a level of detail and context and analysis and follow up. Like there's just so many areas of this of the business that are just way. Way more effective than they they were at the past because someone the time to just suck all this context one place and build a a couple on top of it that made it a lot more, uh, actionable at the end of the day. Uh, so that's one. The second example is, , we have, um, a customer brief generator. This is similar to the , use cases we were talking about before. This one is just a simple form, and you come in in and you enter a domain and that domain then goes and hits a couple things. It hits, uh, a, web search, deep, deep research, and it pulls out a whole a whole bunch of just generically like, tell me about this company. Tell me me about like, what are the things they're doing, et cetera. Second thing it does is hits a glean internally. So we we use glean for company stuff, and it goes and pulls out everything internal discussion that we've had this, company internally. So are they they a customer? Are they a customer? Um, you know, are they. You know, imp in, a renewal process, Are we trying to close them? Like oh, you get sort of a whole bunch of rich, thematic information that. Uh, it it goes and hits that voice of customer so it extracts all the information outta there. So, you you know, support queries, uh, sales requests, categorize by date, like are there common issues that they're hitting? Um, it pulls in like the account team that's with this. So I know know the ae, the sales rep, the support team that are, are, dedicated to working on this. So if I have follow-up on on 'em, uh, what to do. And the last thing it it does is it pulls in usage stats. So what are they using us they, you know, doing stuff with Jira and are they doing stuff with MailChimp or Salesforce? They're doing, you know, AI related use cases, non AI use cases? And so if I'm gonna go meet somebody, I I often will just like. You know, put their domain in in and now I can show up and be ready to talk to 'em in a way that uh, a. A lot, I'm just a lot more than I was in the past. Um, you know, in the past I'd I'd show be like, I bet you're using Zapier and I and I bet you're using it for that, and the other. And just just because I've been working on the for so long, I kind of of intuitively could mostly, um, but but most of our employees can't do that. Most of our employees haven't at, you know, 14 years on the business. Uh, and so having a tool like that is really helpful. And even for me, instead of just guessing, like I actually know now, I can say like, Hey, Hey, I think you're this, that, and the and the other. Um, you know, I think think you have an to do these four other things that
[00:22:10] Jeremy Utley: It, it makes, it makes you and your teammates superheroes really. That's how I hear it totally does. It's like you show up to a call and it's like, this person's like, have they read my email? Do they know me? I mean, so to me that is a great example of exceptional AI use. I'd love to talk about your adoption metrics a second. Cause a big figure you put out in the world's 89%, and just so you know where I'm coming from, I'm hugely skeptical of potentially, you know, self-serving metrics. I remember Henrik and I run a, you know, a global kind of meeting with a professional services organization and this person said. I kid you not, for the first time in my 40 year career. I'm seeing a hundred percent across the enterprise. This is, you know, very very large enterprise. I had happened recently to have conducted a multi-week training with several hundred of this person's employees, and I would not categorize them as
[00:23:03] Wade Foster: a hundred percent, you know,
[00:23:04] Jeremy Utley: even, even a fraction of that. But to me it just, it speaks to there, a, there's a little bit of a moral hazard in reporting metrics that. So how do you think about, when you say 89% adoption, how you think about, call it quantity use, uh, versus quality use, and what does, what does it take for you to be able to boast an adoption rates? if you had to defend that metric.
[00:23:29] Wade Foster: Yeah. Uh, so, you know, look at usage across a bunch of different tools. So, you know, we've got our engineers using Cursor, we have our own using Zapier and the AI capabilities within Zapier. have, uh, CHATT GT licenses for most people the company. Uh, and then we we run internal surveys, uh, we say, Hey, like, you know, tell us, what your most impactful AI use cases are, how often are you it? Things like that. And And so, know, know, through like those, uh, I guess tools, we're able to a sense of how much AI adoption is happening and what is the level of depth that it is at. And so so when I boast an 89% metric, like I'm pretty confident that we've got like 89% of our folks using it, with some amount of regularity. Now your point, the the depth of that varies. Um, you know, there are are definitely folks are, inside a tool like cursor. Every minute of every day. Like that's, that they are living and inhaling and this stuff. And And then you've got who are like, you know, using Chacha PT Daily, but it's like, you know, a couple times a day you, I'm know, I'm going to IT for tasks. Uh, so that's kind of what it ends up looking like at the level. Then you pop it up, and this is where I where I think a lot of the power is. You got a lot of benefiting from AI that other folks have built. So like the voice of tool is a great example of one individual who has built a tool that now can empower inside of the organization.
Now, are the users of that using ai? Yeah, I guess. Um, they, like, are you you know, are they the they're definitely the benefactors of this like, work that this other individual done? Um. But, uh, you know, are they using ai? Are they not using ai? I don't You be the judge, right? So like, there's stuff like that, but those are transformative use cases nonetheless,
for sure.
[00:25:21] Henrik Werdelin: One thing that I. am so curious about how you think about is what I think normally about like the itemization of flows. And so I think because I've been a SA user forever, and other tools, it's very easy for me to look at a problem and say, Hey, this is actually how I'll kind of like, basically cut it into slices and this is how I'll have different savior or sap kind of doing different things, right?
So my brain almost like works like that. I think for a lot of people, you know, when they have a problem they go like, oh, I need a new headcount to do this. And because the way that they kind of like objectify how to solve the problem is they get another human to do it. It, it seems to me that it's a bit of a skill to atomize stuff. , And you've done this and obviously invented the whole system that does this. How do you explain to people who are new to, let's just do savior, because I think that's like such a good architecture for even thinking about how you make automated workflows or, or kind of like, uh, bots. How, how do you normally explain to people how that thought process go?
[00:26:28] Wade Foster: Yeah, the, so I mean, you're spot on. the best users of AI that I watch are really good at things down step by step. And yeah, there's this new trendy context engineering. Like, they're basically going bit by bit and saying, gather this of data from here, gather this piece of data from here, feed it to this prompt, and then feed it to the next prompt, and then feed it to the the next one. And they're chaining all things along. So spot on that, like that is I think, uh, you, you see the like most impressive results and most people don't intuitively do that. Um, you know. The way in in which I often try and explain it is to use examples is to like step back and show how this can actually work. And, um, I, the one tool that I like to is Zapier agents. So Zapier agents. the thing I about it is when you you go to prompt something you wanna build, it will rewrite your prompt for you and it to you. So the the example I like to do is, Hey, can um, I, I'd like you to to my email. And so, you know, most people when they say like, oh great, I can have an AI to to my email, they'll come in and say. Reply to to my email. Like they're, they don't sort of think through like, okay, if it's a recruiting email, I want you to reply this it's a sales email, I want you to reply this way. That, and just like, you know, start breaking it down bit by bit, by bit, by bit. So So when you prompt the zap your agent, it rewrites your prompt and it sets it up until like 1, 2, 3, 4, 5 steps. And so you start to it and it sounds pretty human, you start to go, oh. Interesting. Like it took that and it into this. And And then when you read through what it turned it into, you start to realize, well, on step three, like that's not actually what I want it to do. I I want it something more like this. And so so it starts to your brain to realize like, oh, I need to like guide in this way. Uh, and you know, know, I was just giving it it this like generic step, but because it rewrote it it for me, I realized, ah, I need to add steps 6, 7, 8. And then you can, you, you we'll often see that as folks learn the skill. It, it turns into like their their Zap titles will start to be like version version 93, et cetera. So they start to like realize like, oh wow, there's a whole optimization process that can happen here to really make this system like work great for me. Um, versus just like reply to to my email.
[00:28:55] Jeremy Utley: okay, we gotta talk code red for a minute because clearly that I think is a seminal moment in your history and in the organization kind of as an inflection point. But I think it's also, it's a real, it's really great branding because I think a lot of organizations , they hear code red, they see how you describe it and they go, we need to do that, or even maybe more pointedly do we need to do that? Could you talk for a second about, uh, how you assessed as a CEO when it was time to call Code Red and how you advise other leaders to know when the right time to Declare, code. read is,
[00:29:31] Wade Foster: Yeah. We've only done a coder red once, and so it in this moment, and it was right after the GPT-4 launch. The the reason we did it was, uh, we were, my co-founders were messing around with, uh. GPTI guess it would've would've been three before chat GPT had launched. Uh, and so folks in the in the organization that like, Hey, we we were curious about this We were were interested. But I think time it felt more like, oh, that's a weird founder thing. off kind of like just Exploring like doing what they do. Right. Um, but no one, like, it wasn't like a serious discussion. Internally, chacha PT comes out, starts to change a little bit. We're We're like, Hey, whoa, this stuff's cool. Like, and and it has more widespread applicability maybe we thought. So people are getting like more excited about this stuff. You're starting to see like a handful of individuals and teams like. Start to think through like, Hmm, could we build a feature around this? Could we automate this or the other? But I'd say still, you 90 plus percent of the org is like, you know, laser focused on my roadmap, goals, my, you you know, short term things that I gotta go hit. We have an offsite in February of, I guess it was 2024, uh, where, You You know, we ha we get on stage, some customers on stage that are doing cool stuff with ai, we're more things this way. And And so more of the starts to like get interested and about this. But we're still like, pretty organic about our push to go adopt this stuff. We're just more like, Hey, Hey, this is, this is, a trend. We should be paying attention, da da. Da.
[00:31:02] Jeremy Utley: And by organic you mean it's, you're, you're, you're encouraging bottom up experimentation, but there's nothing top down. Okay. Yeah.
[00:31:09] Wade Foster: no mandates, there's no no goals. Like nobody's getting hired or fired because of stuff. Not really like, like, but it's, you there it's like like subtle encouragement and pressure to to like go, go try these things. G GPT-4 comes out I think March of that year though, and that was like a wake up call. because the improvement between 3.5 and four enormous, and the release cycle between 3.5 and four really short. And so so we just at that and said, if we are gonna have releases happening every months. And the the improvement is rate is going to be that. Then how we're thinking about our roadmap, how we're thinking about our operations. Like we are are just, we're just totally miscalibrated on a of areas. And so we didn't have like all of the answers, we we just knew a lot of stuff's gotta change. and I I remember a late night call with my co-founder saying like, we, we, we, gotta, we like, we do something different. Like we we gotta chart a course on this stuff because this is gonna be the, this is the uh, for the next decade. And so we called the the code red. I don't think it was popular at the moment. Um, there there was definitely in the organization where we were were like alarmist and, you know, sensational and like comments like that came up. Um,
[00:32:27] Jeremy Utley: Yeah.
[00:32:28] Wade Foster: but you know, I think think we did some things One, we we ran an all hands so we said, Hey everybody, stop what you're doing. You know, we'll put out a little bit of couple loom training videos on like how to use the open ai, uh, APIs, um, how to use various tools, and we just people a week to just put their hands on the technology, stop what they were doing, and and just build some familiarity with it to make it less like, fearful, scary thing and more of a oh, here's some awesome stuff it does. Here's some weird things it does. You know, at the time, like hallucinations were like really rampant inside these tools. folks are like, you know, you can have have some fun over like, it's like it clearly it's it's not gonna job yet because it can't do this, that, or the other. So that like, familiarity with it made it a lot less scary and it made it a lot more just like, oh, here's another tool in our in our toolbox. Now granted it's different because widespread applicability. Then we we just kind of rinse, and repeated that like every, you know, six months we do a hackaton try do show and tells at all hands and. and that cadence I think just got, every single time we we do that, we just see more cases, more ambition going up, uh, more folks sort of jumping on the bandwagon. And so what started off as like the are kind of off there like doing oddball stuff, like it totally inverted now, where it's like if you're not using ai, like people are kind of like, what? Like why?
[00:33:49] Jeremy Utley: Right?
[00:33:50] Henrik Werdelin: So from zero to 10, if zero is like not using AI at all and 10 is like the most, most use of AI that you can imagine your organization could ever be capable of, where do you think you on the journey right now?
[00:34:05] Wade Foster: I mean, probably like a two. And I think we're doing better than most
[00:34:12] Henrik Werdelin: I think if you talk to Moderna, which do very well, we do Logitech that do very well bark, which I think it does very well. I think most founders of those organization are with you. We feel that we're at two. Right. And, and everybody else is like impressed like, like we are with you. Like, this is like really incredible what you're doing. Why do you think it is? Is it just because humans, uh, uh, you know, take some time to, to change? Or what do you think is like the underlying thing that makes it complicated for people to adapt this, this is back to people who were printing out their emails in their early days of the internet or, what do you think is the, the root cause of this,
[00:34:49] Wade Foster: I I think it's a lot to ship the tech than it is to change organizations. And this is one area where I, I have some for the companies that that have started chat GVT, is is that they got to just start with a clean slate
[00:35:02] Jeremy Utley: from the ground up. Yeah.
[00:35:04] Wade Foster: Yeah. And so like you often talk to founders and they're like, you know, Claude is writing, you know, 95 plus percent of my code.
[00:35:11] Jeremy Utley: Yeah. And
[00:35:11] Wade Foster: you're like, that's incredible. Uh, whereas, you know, we got a 14 year code base that like is not easily worked in in by AI yet. Now we're working to a lot of that stuff, but but that's not something we can just snap our fingers happen. So I there's this this whole, just like we got, you know, these, these, organizations older than Chate have a whole bunch of stuff that have to get reinvented and rebuilt. And that just time and effort, um, to work your way through. And, uh, and there's involved. Like it's it's not always like, Hey, if we shed this department and recalibrate it and retune it it into this way, will that actually work? Like it, it might, but it's not a certainty. And sort of like, take that leap, like that's, that takes some conviction to get there. And so I just, I think it just a lot of time to to like really get to, you know, like a 10. I, I I don't know. I, I don't know what any organization that's at a 10.
Well,
[00:36:07] Jeremy Utley: what, what, i'm too, a big part of this has changed right? I mean, totally. Part of, I the reason you're a two outta 10 is because of the . You mentioned earlier the depth of adoption varies greatly. Right? You also mentioned that when you declared code red, not everybody was happy about it. I'd love to for a second about how you think about. Organizations managing, you know, you mentioned in your memo, for example, I was just looking at last night. Yeah. I, I'm reading verbatim here. Set Q2 personal growth goal to use AI or GPT in at least one area of your work. If you're not sure how, talk with your and if your manager isn't sure, they talk with their manager and so on, until the answer becomes clear. So I love that and I, I totally agree. Find one way to use AI daily. Here's my question for you. How much patience do you give a human who needs change before you say, you know what? I think the, the phrase sam Altman has used is gonna make it. There's some people you go, they're not gonna make it, like they're not gonna make the transition. Totally. , How much kind of encouragement, incentive, motivation do you provide before you say, you know what, we actually need to hire a different kind of individual?
[00:37:16] Wade Foster: Yeah. Well, I think for us the answer is, has empirically about two years, like we called the code red two years ago. And where I'm at today is I can of tell, you know, if I I talk to somebody and see like, hey, resistance is like the biggest thing, like resistance, I'm gonna make it. Um, that, that it's like, hey, I we just can't work with that. Like, that's not gonna be acceptable now, uh, people who who are curious and, want to do successful, but maybe just haven't found the right use case or working in a part of the company that like doesn't lend itself well to these yet, and they're just kind of of banging their against the wall still that I'm like tolerant of. I'm like, okay, yeah, we'll work our way through Like, let's, let's go. This is not, you you know, these things are, you know, as impressive as as they are, they still are just tools moment in time. And so, there, there there is kind of like a line between like resistance versus like interest struggling. And it's like, as long as as you're willing to learn and put the in and try, like, , we're still willing to have you on the boat, uh, here, I
[00:38:24] Jeremy Utley: think at a, I don't know if you agree with this. I think at is a great metric. actually not , like what your batting average is right now, but if you show up to a performance review in six months and you say, I haven't tried anything, that's a problem. say, here's the 20 things I've I've and they've all failed, I go, great, what's, what's, the next 20? Right.
[00:38:42] Wade Foster: Yeah.
[00:38:42] Jeremy Utley: I know if that's the right we,
[00:38:43] Wade Foster: actually of actually like that
framing. Mm-hmm.
[00:38:46] Jeremy Utley: Okay. Okay. I think in a lot of organizations the problem is inaction is permissible.
[00:38:53] Wade Foster: Yes.
[00:38:53] Jeremy Utley: and to say I haven't done anything perfectly acceptable, so you actually have change the frame. But that is a really painful thing. I dunno dunno if you have any reactions to that.
[00:39:03] Wade Foster: I a hundred agree. I think if, you know, if we wanna be like very reductionist about how you set culture inside of a company, it's what do you reward? What do you tolerate what do you not tolerate? And so to use this as an example, like in action, wouldn't tolerate at bats. We're like, Hey, that's good. And And then what we reward is like successful attempts. Uh, so you could sort of set up a pretty simple system there for, how to enable managers to, to go manage against that.
[00:39:32] Henrik Werdelin: follow on the culture stuff is for me is you have a pretty remote organization. You did this weight test. I didn't you where you were kind of like trying to see how many people could recognize it was you or ai. Yeah. Yeah. So obviously AI has, like, you know, a lot of use of AI has this tension between kinda like close-knit culture in organization because you might just talk more to your you know, chatt BT than colleague. It has like little bit of like this dehumanization, you know, suddenly like your, your, uh, your manager suddenly replies most of the replies, uh, done with ai. And the same time you like your 12 year organization, like, you've been through like a few cycles of people now and stuff like that. Talk to me a little bit about like, how do you think about AI remote and then culture and how to kind of like do that. You mentioned that you guys do offsites, which sound like maybe be one answer to question. Are there other kind of ways that you try to kind of create like a, cohering sense of togetherness?
[00:40:34] Wade Foster: Yeah, I, I, I don't know that like AI. Actually does much for us on this at the moment. Like, I think the big way that we bring, togetherness and comradery is one, the in-person gathering. So that for the whole company once a year, and then we have smaller groups that get together. Periodically. Uh, two, we have things like show and tell and weekly demos and like, you know, know, all these sorts of things that give people a chance to you know, connect and give feedback and learn and share and, you know, uh, cross pollinate knowledge. We have, uh, you know, a bunch of like, off topic channels at work. So these are all like prefixed with fun. And so there's like fun gardening and fun home ownership and fun, whatever, where people can sort of, you know, uh, get to to know each other about, you know, interests. Uh, if there is one around ai, we do have a fun AI channel and that's where there's a lot of like sharing cool stuff. You You know, it everything from news about AI like. Wow. Did you see this demo, , that someone did internally or externally? Um, that like, just creates like a sense of like. You know, just ah, the of possible around like what, what what we could be doing these things. And so all these just like little, I don't know, like cultural rituals or habits, I do think help create, um, you know, a know, a sense of togetherness, a sense of belonging, which is pretty critical if you're, you know, have a distributed team. Which, you know, I'd I'd say most have, at least some part of our of our company are doing that these days.
[00:42:06] Jeremy Utley: When you about and togetherness and belonging it, I don't know if this is the right spectrum, but on the other end of the spectrum, I think about a phenomenon like burnout. And I think in remote you know, it's kind of, you're kinda always on, you know, in a way. how do you think about this is, and if, you know, if listened much to our pod, you know, we get existential and practical. Here's, is an exsitential question, where should the benefits accrue? The benefits of, if you think about efficiency gains, gains, if you take as a given, a lot of folks are burnt out. stands to reason they should actually, individual employees should probably accrue some of the benefits of the gains. At the same time an organization wants drive, you know, everything to the line. How do you think about where do we net out in years from now? Are people actually legitimately working less? Are early startups high, much more profitable? You know, where, where do we net out in terms of where it benefits accrue
[00:43:04] Wade Foster: I, I, I know so far capitalism feels undefeated, so, you know, Yeah, I think there's this belief in circles that, oh, margins are just gonna go through the roof because we're gonna have a lot fewer people and you you know, a lot gonna get done. But I I think the is like, as if if margins get bigger and and bigger, somebody, some competitor's gonna see that and say, Hmm, you know, know, the Jeff Bezos quote, you know, your margin opportunity. And it's, they're gonna step in and say, Hey, we're gonna wanna go compete for that stuff. And you know, I think that like, net net, , it probably doesn't change that like I think the of work that AI can for us obviously changes a a lot of how we operate, you know, how much effort we're putting into work, how how many we're working, the cultural norms around that, like, I don't know. I it. it. I, probably would bet it stays pretty similar, would just be my guess um, because I think capitalism just sort of has a way of like evening this stuff out.
[00:44:09] Jeremy Utley: Yeah.
[00:44:10] Henrik Werdelin: I got the last question for you, um, agents talking to agents, um, so obviously we've, uh, all learned how to use the autonomous agents now, like the, you ai, you Chad, your perplexity, whatever. Um. It, it seems that a lot of people believe that they'll be agents, so you'll have generalized agents and you'll specialized agents. Right. and and at one point there'll be kind of this hand over and, and know you guys done some work with MCP, which for people who don't know, is this one of these protocols that are kind of like being, or these languages that are being kind of explored of like how agents talk to agents or when one agent can kind of hand over to another resource. Um, when you talk to a venture capitalist and you ask them, when do you start to see. Like real agent to agent kind of business kind of emerged, they say six to 18 month. But when we talk to people that actually do this every day, they're like, we don't even know what language they'll talk yet. Um, where are you on the whole spectrum of like agent to agent communication, how fast it will move when we'll start to see stuff? What might be the first kind of use cases in that kind of like whole universe?
[00:45:18] Wade Foster: Yeah, I mean, mean, I think we, we, can do agent to agent communication today. The problem that we see in practice is the reliability is really poor, for any reasonably complex task. and I think the why is that you start to to just chain along. You know, if you, if a complex task has like five or 10 steps, and every step is an agent handing off to another agent,
[00:45:41] Jeremy Utley: There's a waiting there. The reliability.
[00:45:43] Wade Foster: yeah, it's like if the reliability's 90%, like, well, by by the time you get to the end of that like. It's way off in left field and this is just not, it's just not good. Um, and so like that's the real challenge that I think, you know, we're facing. And i, I, think what my sense is that the way in in which we're going to get these much better is by having agents that are like you, you basically have to find a way like scope the agent down to like like a narrow enough set of tasks, like narrow task to increase the reliability such that you can get closer to, air, like an error rate that is tolerable for the organization's, you know, risk. Uh, and so, that, that, that that takes away some of the magic the day where it's like, well, if we're really narrowing the down to be so so tiny, like why aren't we not just writing like deterministic code here? Like we we ought to just be stamping out, like this very practical thing. And so, you know, there's, there's something in and that that like, we're gonna have to go um, to, to, to get this a lot stronger, um, to these agents step these out. But, uh, I think the, I think the end result is we're gonna see like a universe of agents that are very good at like very narrow sets of things.
[00:46:57] Jeremy Utley: Wade, I have one last question, if you're willing. Um, a number of our audience we kinda float, hey, you know, upcoming guests and say, Hey, if you've got, uh, suggestions for questions, let know. One of the words from your that have come up a few in audience questions is this idea duplication. You mentioned in memo there's gonna be duplication of effort, we're gonna have duplication of experiments, et cetera, et cetera. Can you talk for a second about, I'm a huge, as like an innovation junkie. I'm a huge believer in parallel so I get it. Yeah. But talk for a second about how do you manage duplicate efforts and how, , how do you organize knowing that duplication of effort is, is going to be a part of this process?
[00:47:41] Wade Foster: At the beginning of of this, you don't manage it. You You mostly are just managing the like, of the team. Like there it, it, it, feels so wasteful to so people internally that there would be multiple efforts. Why aren't we on this? Why are like, like, why are you doing Why? Like, I'm doing this. Why are you doing Like, my thing should be the one to do it. It's like that's the thing that you is really painful for organizations. And so so mostly you're just trying to let people know okay, it's
[00:48:09] Jeremy Utley: Well, how do you manage this? When, when somebody says, why are you, why are we doing this? What? what do you say? How do you manage your psychology and to say
[00:48:16] Wade Foster: it's it's okay, Like at some point in time we are gonna, you know, we are are gonna actually have a better of of like, what path forward? And these things will consolidate. They will sort of come into one. But for right now, we actually don't wanna get on that off ramp. Like we don't wanna pour concrete around this a solution because we know yet that this is the one. Uh, and you see this like with with some of the early products out there, um, if you go go 'em, you can actually see where they an off ramp. Mm-hmm. And like their entire architecture of how the product works. It's backwards now. Um, like I can can think of a very popular product use internally where their agent architecture is wrong. Like it's just wrong. Uh, and like, like, I of get it, they sort of, of, they were fast to market. They They pushed a out. Yeah. You know, if I was in their like made, might have have made the same choice. But the is like they're gonna have to go rebuild all that now to to take advantage of the way these tools work, at this point in time. Uh, and so there's just a sense of just like letting people off the hook for like what they feel is like a failure. Um, you know, that that there there there would be so much disorganization that be like a thousand flowers burning. Blooming and saying like,
[00:49:31] Jeremy Utley: or, or, burning, burning may be the better analogy. Yeah. Analogy could be burning too. Be, but that's, no, I, I mean, what I'm hearing is you have conviction, you know, as an innovator that some duplication of effort and parallel experimentation is necessary. And I think that's actually, it's, it's important to state that because I think that there's number of people that don't know, kind call it the prior probabilities around innovation that don't know how it works. If you think about managing it, like you're managing routine work where it's deterministic, where you, know, uh, where you commission a single experiment to do a single thing. You know, I mean, one of my favorite examples is Jobs parallel commissioning experiments around what would become the iPhone. You you know, Tony Fidel, I don't know, most folks may not know this, but just for who are kind of innovation history nerds me, you go, okay, Tony Fidel was to make a click wheel version of the iPhone because they didn't know at the time whether Multitouch work. And so there were multiple teams in apple building different iPhones, right? People don't, people think there's an iPhone team, right? Uh, you know that one team with one approach. It's all say, what I'm hearing you say way it is, you as a founder and CEO know that of effort is necessary to identify the best forward. Therefore, you know, your job to manage the psychology of individuals who don't know that. Mm-hmm. first order thing for a leader listening to this show is you have to know, as a leader, duplication of effort is required. Then question becomes, how do I manage people's psychology who feel like this is wasted effort, et cetera, et cetera.
[00:51:05] Wade Foster: Totally. Totally. And then then at some point in time, you probably are have to deal with The, the the inverse pain, which is now everyone's trained on this one way of working, and now it's it's like, oh, we have some answers. We do need to make some bets, and we have to figure out a way to like, bring these approaches together, uh, and be a thing. And And that's like, that's a moment in time and it requires you to shift gears again. Uh, and so it's, you, you, you, you, talked about change earlier and it's like, God, that's like, it's like such a, you know, you know, once your organization a certain size and scale, like a lot of that is the, the, name of the game.
[00:51:40] Jeremy Utley: I Anything you wanna say to tip your hat to anything regarding future.
[00:51:44] Wade Foster: we've got, uh, we'll be be announcing, zap Connect is, uh, opening up for registration here soon, this week. And folks will, uh, it happens in the fall and , if you wanna see what a lot of stuff we've got cooking around AI is coming, I would say come to that. 'cause we'll have some fun stuff dropping.
[00:52:03] Jeremy Utley: That's super cool.
[00:52:03] Henrik Werdelin: That's awesome. Fantastic. Mic drop. Nailed. Really, really, really enjoyed the conversation. Thank you so much for doing this. Yeah, you And, and thanks for making the product. I'm a, I'm a super use of it, so, uh. Yeah. I really appreciate it. Oh, he's already.
[00:52:16] Wade Foster: I love it. Thanks for having me guys.
[00:52:18] Jeremy Utley: Thanks Wade. Take care. Bye
[00:52:20] Wade Foster: bye.
[00:52:20] Jeremy Utley: Um, Okay, so Henrik as a super fan of the product, what stood stood out to you a customer of Zapier? Zapier?
[00:52:30] Henrik Werdelin: You know what, the thing actually that mostly stood out with me is that he started his company around the same time as I started bark. And, uh, similar type size. And so it's kind of, it's always nice to just hear somebody kind of, uh, be at the same point in their organizational path or journey. , I think, you know, a few things. Let's start there. For example, I am. Always impressed by people like, uh, weight that we have on that is just pushing all this AI usage through the organization. And what stands out is that he says that he's a two out of a 10, right? Mm-hmm. And I think obviously objectively, if you compare him to any other organization, he's probably a nine out of a 10, but, right. I, I think people like him believe that there's so much opportunity for not just making. Not just making it more efficient and like reducing staff, but also just making the organization more and do more of this thing that he thinks that the organization to serve to do. And so I think the one thing is that thing, the kind of a permutation of that is something you said, which is just inaction is permissible in a lot of organizations and I think a lot of founders, as you can hear on him. It's just tired of and I don't think that that is going to stand for the next five I think leaders are just gonna go, Hey, if you don't even experiment about this, it's gonna be really tough for you to have a long-term career. And I think that's just a universal kind of statement that people who basically still wanna print their emails don't be Fred or whatever it is that, uh,
[00:54:09] Jeremy Utley: Don't be Fred.
[00:54:10] Henrik Werdelin: Yeah. Um, then I think. If you don't already all the feedback from your customers and put it into a place where an agent can access it so that that data is available throughout the organization. That just seemed to be kinda like an absolutely no brainer for any organization at any size. Have all feedback loops from the organization. In a place where agents. Chatbots and thus the organization at Launch can get hold of it. And I think it was just a friendly reminder from him that they're doing it in secret benefits from it and everybody else should do the same. And and I mean like it's something that's basically, and I was kind of thinking about asking you about it 'cause you're such a monster of that. I, I, I do think that a lot of people listen to this podcast who use AI a lot, just can't get their head around why people are not using it more because it seems so useful when you start to use it, and I think it's just very conceptual how you do a lot of these things. And so just having very concrete example, I, I saw a LinkedIn post you posted the other day where you were talking about something that. Somebody from the Forest Service was doing, uh, some of that. And so you seem to have all these forms where people are just showcase this kind of like endless amount of like use cases and you go to waits, kind of LinkedIn. And he had this kind of , graph also with like a lot of, he used it for onboarding. He used it for offboarding. He used it for these like, you know, know, like just like. 20 use cases. And so I, I, one takeaway is just like the, the simplicity of just constantly finding more avenues to show endless amount of use cases so that people who are not using as much can go like, Hey, wait a minute, can you use it for that? And then go out and do it.
[00:55:53] Jeremy Utley: Yeah. I, um, I really resonate with your comment about practicality and his his comments there's very little kind of simple, approachable how to, there's a, I think he said the refrain of hype is at a volume level 11. And I think similar to that hype, , there's a distribution hype on the positive and doom on the other end. There's a ton of hype and doom, but there's not a lot of kind of humble. Here's something really cool. Here's something really fun. And I I think what differentiates Wade from from many many that I'm observing online is he's just being super helpful and super practical. We should link to several of the resources that he's provided in the show show notes because he's got a got a great, I you know, play by play, breakdown of his hackathon week. He's got a great breakdown of the code red red that they announced. A year ago now, and I think for any, he said two years, by the way, since Code Red, but I think it's it's only one year, right? I'm pretty sure. I
[00:56:53] Henrik Werdelin: dunno,
[00:56:53] Jeremy Utley: GPT-4 last year. Is it? Yeah, GPT four had to had to be last i'm pretty sure it's March of 2023. And if they if they declared a red, so he is basically, he said two years, but I think it's, I think it's one year. My thought is he's a year ahead. Ahead of where most folks are. A march, March 23. Yeah. Oh, 23. Okay. Two years there. There you go. Yeah, he was right. Um, so March 23 is when GT four came out. They announced their code Red April of 2023 had their first AI hackathon. April of of 2023 we're filming July of 2025. I think most folks who are listening have yet to declare a code code red, because they've never thought of it or it or they don't feel it's necessary yet. But you can kind of just start the clock from when do you declare this is so foundational and fundamental and important that we need to actually stop the gears, have folks spend a week experimenting, exploring, building, um, and then. And then. There's so much resources that he's provided. It doesn't have to be, uh, confusing. It it doesn't have to be I mean, you could literally hit rinse, repeat on a lot of his playbook, which he's graciously put out there
[00:58:07] Henrik Werdelin: to add, one thing to add on that is that he said two things, and I think you brought one of them up, which is basically. When do you tell the team that this is a big deal, right? And you using the code with us? The kind of like the moment, but I think what he then also says is like a lot of people felt internally that was a bit alarmist, right? And I think a lot of people who don't call a code rate or something similar, it's probably like feeling the same pushback from the organization. They're like, yeah. like. Emails, email, like ai, you know, whatever. Um, right. the second thing that I think is fascinating is like this idea of hold anxiety, other, his staff's anxiety on his shoulders. Like we will have many people try to do the same thing, and that is fine. You know, we will have many of these projects that we, we'll try and we accept a 90% kind of like, uh. , Success rate, you know, like even per, per chat, that is like something that a lot of organizations probably wouldn't do because, you know, like, Hey, what do say? Like, the product only works 90% of the time. Mm-hmm. Um, so I think there's probably also like, I think a, a leadership. Kind of component of it, which is like, Hey, if I am going to utilize AI and get all the benefits out of ai, I also am going to allow my team to kind of pass on some of this anxiety that they might have until me as a leader, because I am willing to be a little bit alarmist, I am willing to accept that I'm wasting resources, having many people do one thing and so forth.
[00:59:37] Jeremy Utley: Yeah, I mean he, he actually says quite eloquently in in his post, we have determined that standing still is the only sure to fail approach. And so I I think, maybe, maybe a simple way to put this is there's a binary decision before you do nothing or duplicate effort. Effort. Which Will will you choose to do? And I think for too many many organizations, the, the perception of the pain of duplicating effort is leading them to the wrong conclusion, which is therefore let's do nothing because we don't know what to do. And the truth is as with any other innovation, the only way you discover what's worth doing is by commissioning experiments, by duplicating effort in some areas in order to overcome that kind of inertia of the unknown. So Wade's a great example. This is
[01:00:25] Henrik Werdelin: You're probably a, you're probably a student of this because I would imagine this is not an AI thing. That's an innovation thing in general, right?
[01:00:31] Jeremy Utley: Totally a
[01:00:32] Henrik Werdelin: a hundred percent. And so , what is the prevailing kind of wisdom or thesis of why organizations. I guess there's written many books like The Innovator's dilemma and a lot of these different books on it. But if you were to kinda like just answer the. A leader's question now saying, Hey, I don't understand what's so difficult to get my organization to kinda like go from a zero to a five. You know, like , in this journey. Where would you point them?
[01:01:02] Jeremy Utley: I tell them to look in the in the mirror.
[01:01:06] Henrik Werdelin: Say more.
[01:01:09] Jeremy Utley: We have, you and I have had the privilege of speaking with so many amazing leaders. Who are so outspoken about their own experiments their own behavior. You know, think about Brad Anderson sharing his screen before all Hands meetings. You think about Diarra Bousso sending Loom videos to her team. You think think about, , Kevin Kelly's daily suno practice. Right on and on and on and on, right? A leader is saying, I'm trying to tell other people to.is is missing the Right. Starting with yourself, you know, Greg Shoves third screen. Right. And on and on and on. The leaders whose organizations are transforming are leaders who have transformed themselves.
[01:01:55] Henrik Werdelin: Amen. It's a good piece of advice, professor.
[01:02:00] Jeremy Utley: Well, uh, people won't log off because of the moment of of silence. It has took me to think of the answer, no, because sometimes the human, the old old gray matter can't respond as quickly as G pt can
[01:02:12] Henrik Werdelin: I, one of the most, impressive degree shows I've ever been to when was, uh, sung Martin's Lane in London where they have this kind of interactive degree show, and this student had made this kind of piece where she had recorded. Every time on any TV channel a 24 hour, uh, kind of cycle where there was silence, and our thesis was that silence in today's world is when something really powerful is about to happen or has just happened. It's just after the car accident. It's just when somebody said something outrageous, it's just when something crazy is about to happen. And so I was kind of drilling in that moment of like, yo, what's gonna come? And you delivered, sir.
[01:02:56] Jeremy Utley: Well, let's, let's, let's leave it to the listeners. Maybe the secret code word should be, uh, either silence delivered or silence did not deliver. That's the code word you get you get to choose. Let's do that.
If you've enjoyed this episode, feel free to like, subscribe, share share with a friend, , perhaps share with a leader who needs to hear the message after the moment of of silence. Thank you. Goodbye,