How Every's Head of Consulting Uses Codex Every Day

Transcript

You go and teach executives and other people at big companies how to use AI. And so I think what you're doing is a good window into how great operators and executives are starting to use this stuff. What Codex helped me do was create an operating system. My email knows what's going on more than I do. I'm bullish on all of the administrative tasks that will suddenly be taken care of because now we have this super alien tool that can support those things. [SPEAKER_03] Knowledge work now is turning into something like gardening, where when you're gardening, you're creating the conditions for the growth to happen, but you're not making the plant with your hands. Every is the only subscription you need to stay at the edge of AI. If you care about being on top of the latest models and using the latest tools, you have to subscribe to Every to separate out the signal from the noise. Go to every.to slash subscribe today. Natalia, welcome to the show. [SPEAKER_00] Thanks, Dan. Good to be back. [SPEAKER_00] So for people who missed your last episode, you are our head of consulting at Every. That's right. You are also the manager of Claudie, consulting's AI agent employee, which was the star of our last episode together. [SPEAKER_03] And I wanted to bring you on because I feel like every couple months things shift so radically. And for me, you're one of the bellwethers of how things are changing because you're an early adopter yourself. [SPEAKER_03] And you go and teach executives and other people at big companies how to use AI. And so I think what you're doing is a good window into how really great operators and executives are starting to use this stuff. So the last time we chatted, Claudie, which is the internal AI employee agent that we built to help run the consulting business, to send out sales proposals and manage the CRM and all that kind of stuff. Claudie was this nascent thing that Nitesh, who's our senior AI engineer, was sort of Wizard of Oz-ing in the background, making it work minute by minute. But I feel like now Claudie is actually working. The model releases over the last couple of months have dramatically changed how much she's able to do. So give us an update on Claudie. How are things going there? You know, it's funny with the speed of AI. [SPEAKER_01] Claudie feels just not novel. [SPEAKER_01] Claudie is an agent that does work for us every day. [SPEAKER_00] And Claudie has its own LinkedIn and Twitter feed and manages our dashboards and has a trust battery now that's new. [SPEAKER_00] So it's running on a loop to self-evaluate performance and improve itself given the feedback that we give it. [SPEAKER_00] And Claudie is thriving, I guess. [SPEAKER_00] Yes. [SPEAKER_00] But one of the things that's interesting is you hired Claudie to do operations stuff. [SPEAKER_03] But you're also now hiring an operations person. [SPEAKER_03] So what have you learned about the uses and limits of these sorts of internal agents for stuff that you might want to hire a human for? [SPEAKER_03] Yeah, you know, it's really interesting. I think as we've all been using AI more, the thing that we keep coming back to is that AI is really good at executing against a standard operating procedure. And Claudie is exceptional at that. But Claudie still needs two things. [SPEAKER_00] One is it needs constant oversight and management to make sure that it's doing those things really well. [SPEAKER_00] So the question of taste and reaching for excellence still requires direction and managerial support, which can be quite tedious and time consuming. [SPEAKER_00] And two is when you are working with people, as much as I love working with Claudie, I want to interface with people. And I find that. I can't really relate, but I see why someone might feel that way. [SPEAKER_03] And the reality is that while we do have all of these rich dashboards and all of this data that Claudie is populating, we need someone to surface what is interesting about that data, what the signals are, and to help lead those conversations. And so actually, I suspect that we will continue to expand the team to build on the data and information that Claudie surfaces so that we can actually do interesting things with it. One of the big things that you went through recently, which I think is super relevant to anyone inside of a big org or anyone running a software company, is you actually bought a CRM. [SPEAKER_03] And previously it was all Claudie glued together with Google Sheets. [SPEAKER_03] And I think there's this whole narrative running around. [SPEAKER_03] I think honestly SaaS stocks are back. [SPEAKER_03] So the narrative is a little bit less present than it used to be. But it's still on people's minds: are you just going to vibe code all SaaS? [SPEAKER_03] Fable currently is banned, but I'm sure it will be back. [SPEAKER_03] Maybe it's even back by the time this episode comes out. But if Fable can just one shot a CMS, why would you use one? You have the ability to make your own CMS. And we have enough resources internally for us to vibe code one. But you decided not to. Yeah. Or you decided to move off the homemade one onto a professional one. So why would you do that? Yeah. [SPEAKER_01] So despite my hopes and aspirations that I could do all of the things and become an engineer and maintain all of these engineering products that I've vibe coded. [SPEAKER_01] This one, I can't relate to this one. [SPEAKER_01] It turns out there are actually private and public companies whose entire business it is to do these things really well. [SPEAKER_01] And sometimes these very specific things really well. [SPEAKER_00] So we vibe coded a CRM tool that allowed us to manage our sales pipeline for a while. [SPEAKER_00] And it was managing in Google Sheets. So it was Claudie, Claudia was the glue between what was going on in Slack and the meetings and Google Sheets. [SPEAKER_03] Yeah, exactly. [SPEAKER_00] So basically Claudia had access, was able to read my email, was able to read our meeting note taker's notes, was able to digest inbound leads that came. [SPEAKER_00] And then would track this all in a Google Sheet. [SPEAKER_00] And then eventually that became a database that we were managing. And these things just require maintenance, right? In order for the data quality to be good enough that you can do interesting things with it, you need to almost, like with Claudia, be on top of the quality of the data. [SPEAKER_00] And so it turns out this is Atlassian's entire business. Yeah, exactly. So Claudia had access to, was able to read my email, was able to read our meeting note takers notes, was able to digest inbound leads that came. And then would track this all in a Google Sheet. And then eventually that became a database that we were managing. And these things just require maintenance, right? In order for the data quality to be good enough that you can do interesting things with it, you need to almost like Claudia, you need to be on top of the quality of the data. And so it turns out this is ATEO's entire business. And so I think one of the challenges with AI I certainly have is that in the era of AI, you can build anything. I think I even said this in the last podcast. The question is, should you build and maintain whatever you actually build? And I think in this case, and probably in other use cases, we also rolled out Asana for our project management system. I think we're able to do the scale of the work that we are able to do because of ATEO, because of Asana, and because of Claudia managing all of that information is much greater than if we didn't have those tools. But now we just have less burden on the team to maintain that. [SPEAKER_03] Can you give me a concrete example? Because in my head, I'm thinking, well, CMS is just customer records. And that's just a spreadsheet. So you should just be able to have Claudia do everything. So can you give me a deeper dive into what specific kinds of things came up that were harder than you expected? Yeah. So yeah, totally. Let's talk about maybe a traditional sales pipeline lead, right? So there's the inbound. You have these sales logic rules where certain things need to happen in order for them to move further down the pipeline until they are a converted client. And sometimes those things happen very quickly. Sometimes they happen over a longer period of time. With my human brain, I think I can track what's going on over a two to three month period. And then any conversations that are taking place outside of that and after a certain amount of volume, I just can't quite track. With a tool like ATIO, it has access to all of the things that Claudia had access to, but it has really robust logic. So it can basically track the movement of a deal over the course of the pipeline. And it can flag it to me in different ways in a way that I would have had to supervise Claudia to do. And Claudia was just not inherently set up for it. It could do that if I spent more time training it to do that. But it's ultimately a reward payoff thing. I think one of the things that's unintuitive about software is real software is a compilation. It's a logical machine that compiles thousands and thousands of little logical rules that you wouldn't expect you would need beforehand. And the whole job of the company and the engineers is to gather all the rules that are needed and then put it into the system. And when something breaks, change the rules. Yeah. And AI is very good at working around that kind of determinative system and writing it, but it's not going to one shot all the rules that you're going to need. Yeah. Needless to say, I've become a really big fan of PRDs. Yeah. And actually scoping what I'm building, which I think I've improved in both scoping and building higher quality things. And also making that decision earlier of whether it's going to be worth it for us to just invest in a tool versus for us to build it out. Yeah. I think a good metaphor is I'm just cycling on the difference between software and language models. But a good metaphor is software is a little bit like your bones in your body. And a language model is a little bit like your brain and your ligaments. So if you didn't have any bones, there'd be no structure and you'd be just a flopping jellyfish on the floor. But if you didn't have your brain and your nervous system and ligaments, you'd just be a pile of sticks. And I think that's a good way for software and language models. That's how they sort of start to work together. And of course, language models can grow bones, which is interesting. That's maybe a bit different from the way we're set up, but growing bones while complicated. And a whole body plan is very complicated. But you said something earlier that I think is really interesting and I want to push on, which is I see you going from not technical to building stuff. And I feel like there's a step change for what you can build and what you can attempt over the last month or two. Do you feel like that's right? And if so, tell me more. Yeah. A hundred percent. [SPEAKER_00] I would say, riffing on Claudia a little bit and the evolution of how I work with Claudia and also how I work with other tools. Codex has been maybe the single greatest improvement. I have to confess on the podcast that Dan did tell me to download Codex. [SPEAKER_01] Maybe every day he saw me for weeks. [SPEAKER_01] I'm very annoying about things I think are good. [SPEAKER_03] I did. And I think you have something that I don't have as much of, which is the fearlessness when it comes to trying out a new AI product. [SPEAKER_01] And you know, you could just feel the compute, it just wants to go. [SPEAKER_00] It's just so powerful. [SPEAKER_00] It's just so powerful. [SPEAKER_00] Mm-hmm. You know, I think generally I love learning. I'm an ambitious learner. And. [SPEAKER_03] You are really, that's something that people should know is you are the most curious learner I think I know. [SPEAKER_03] You spend your weekends having Claude or Codex building these big learning guides that you just read end to end about anything that you're thinking about. [SPEAKER_03] Yeah. [SPEAKER_03] And I love it. I think it's amazing. [SPEAKER_00] It's just so powerful. [SPEAKER_00] It's just so powerful. [SPEAKER_00] Mm-hmm. [SPEAKER_00] I think generally I'm an ambitious learner. [SPEAKER_00] I love learning. [SPEAKER_00] And. [SPEAKER_00] You are really, that's something that people should know is you are the most curious learner I think I know. You spend your weekends having Claude or Codex building these big learning guides that you just read end to end about anything that you're thinking about. Yeah. [SPEAKER_03] And I love it. I think it's amazing. And it's a superpower because AI lets you do more of it and it helps you use it better. I think it is a superpower and sometimes it feels a little bit like a vice. Yeah. [SPEAKER_03] Yeah, yeah, yeah. [SPEAKER_03] Do I need to know the history of bookshelves from first principles or something? [SPEAKER_03] Because I feel that's something you would look up. [SPEAKER_03] I would like to know that. Yes. I would like to know that. [SPEAKER_01] But with Codex, I feel the truth is that I don't have to think so much about the thing. I think about things like the file systems and the folder structures and how the scripts are set up and it just works. And so I think I have to focus a little bit less on architecting things well, which is very much a skill and something that our engineers do extremely well. And just trusting it to make good decisions and actually build solutions for me, which is really what I want. So can you show us some of your Codex workflows? [SPEAKER_03] Yes. [SPEAKER_03] Okay. Let's see. We'll start with, let me share my screen here. We'll start in Codex with, we're talking about learning. [SPEAKER_00] So I have to show you my guiltiest pleasure, which is my favorite skill I've ever built. It was originally a prompt, maybe six months ago and it codifies the way that I like to learn things, which is what is the history of this particular topic? What are the first principles that guide the physics of this topic? And then how did we get to where we are today? [SPEAKER_00] And what are the variables in the marketplace around this topic? And that is how I spend my weekends reading these guides. [SPEAKER_00] And sometimes I don't have 12 hours on a Saturday to read through these guides. And so instead, I make little cartoons that just summarize what I'm seeing and what I'm learning. So this started out, if you can see my screen here, this started out as a prompt that Nitesh, one of the engineers on our team built, based on Claudie, who we all know, works on the consulting team. And Claudie can teach principles and anything coming from this learning skill. So go back up to the top. [SPEAKER_03] Yeah. So tell me, what were you trying to learn and how did this get made? [SPEAKER_03] So in this case, one of the things that, well, one of the sad things maybe that has happened over the past six months is that as I've spent more 12 hour periods in front of my computer, I have prioritized my physical health less. [SPEAKER_01] And so I'm trying to learn about what I need to know to improve physical education, which is what I'm looking for. Mm-hmm. [SPEAKER_00] And what I need to know in order to make more strategic workout decisions. [SPEAKER_00] And so I asked Claudie to make a guide to explain what is the history of physical education? How did we find ourselves in a situation where we have to do specific types of mobility and workouts? [SPEAKER_00] And what do I need to know to make good decisions around how to spend my time on this particular topic? [SPEAKER_00] So Claudie explains how we got to where we are, basically workouts as a topic emerged about 200 years ago. Really? Yeah. [SPEAKER_03] That's actually earlier than I would have expected. [SPEAKER_03] Earlier? Yeah. [SPEAKER_03] Because I figured even a hundred years ago, we were still doing a lot of physical labor. [SPEAKER_03] I think you're right. Yeah, you're right. I mean, it really became a thing during the industrial revolution, of course, as people were spending more time in factories. And so with my learning skill, I could read all about that. But with Codex OpenAI and the visual models that Codex has, which are so powerful and so good, we could just make it a cartoon. And this is something that I could scroll through on the subway or on a walk or having coffee. And it's- What'd you learn? And so it takes these really complex concepts. I mean, one of the things that I learned that was really interesting was anatomy, which I did not learn much of in school and was really helpful to learn about. And actually one of the most interesting things that I really enjoyed from this particular zine, or set of cartoons, was learning about the time scales with which different parts of your anatomy get strong. So muscles get strong faster than ligaments get strong faster than bones, of course. [SPEAKER_00] And so thinking about progression in physical strength as something that's happening across your body from your bones to your brain. [SPEAKER_00] Yeah. [SPEAKER_00] Yeah. [SPEAKER_00] So this is one very fun example. [SPEAKER_00] Something that I will do on the go. [SPEAKER_00] You already know how AI is changing how everyday work gets done, how much ground you can cover and how fast a team can scale. To stay ahead. You need tools that give you a competitive advantage built for this new era. Adio is the CRM for the agent native world. It meets you where you work, compounds every customer signal into context and then acts on it across your pipeline to let you move at unmatched speed and scale with agents and automations for every job. Adio orchestrates your work around the clock. We use it internally at every level and we love it. [SPEAKER_00] Something that I will just do on the go. [SPEAKER_00] You already know how AI is changing how everyday work gets done, how much ground you can cover and how fast a team can scale. [SPEAKER_03] To stay ahead. [SPEAKER_03] You need tools that give you a competitive advantage built for this new era. Adio is the CRM for the agent native world. [SPEAKER_03] It meets you where you work, compounds every customer signal into context and then acts on it across your pipeline to let you move at unmatched speed and scale with agents and automations for every job. Adio orchestrates your work around the clock. We use it internally at every and we love it. It's built to handle the scale of your workloads. [SPEAKER_03] It's extensible with an API and MCP access and is built with infrastructure to keep up with your most ambitious agents. [SPEAKER_03] It's loved by high growth startups like granola modal whisper flow and every Adio runs the work behind every win. [SPEAKER_03] That's Adio the agentic CRM. [SPEAKER_03] Go to adio.com slash every and get 15% off your first year. [SPEAKER_03] That's adio.com slash every. And now back to the episode. Diving into codex. [SPEAKER_00] I will share. How do you organize your codex? Okay. [SPEAKER_00] So you have a bunch of different projects. What are the projects? So you don't use pinned or do you use pinned? [SPEAKER_03] I only use pinned for my email triage, which is the app that you so generously gifted me this year. [SPEAKER_00] And the email triage is the only thing that I really pin. Everything else I just work in. [SPEAKER_00] Okay. [SPEAKER_00] You're a codex pin. [SPEAKER_00] I'm a big pin guy. Because I find that I lose stuff otherwise. I don't have a project for everything. [SPEAKER_03] So it's just all the work I'm doing is just all pinned. But this is interesting. [SPEAKER_03] So you have a project for every sales strategy, your dad, NZQ epistemology. [SPEAKER_03] Incredible. [SPEAKER_03] Tell me more. [SPEAKER_03] These are my learning. [SPEAKER_01] I don't know what to tell you. [SPEAKER_01] I'm suddenly really excited about how Aristotle came up with syllogistic systems and how we use them today. [SPEAKER_01] We definitely don't need to go into that. [SPEAKER_01] It's incredible. [SPEAKER_01] We actually might need to. I've been spending too much time around you. [SPEAKER_01] So I basically just store and organize my codex as I organize my projects. [SPEAKER_00] So it does feel like codex does really well something that you had to do some mental organization in cloud code. In cloud code, I spent a bunch of time really understanding file systems. [SPEAKER_00] And would always have the finder open to understand where things were being saved and what was really being created. In codex, that's all happening in a really visual way. [SPEAKER_00] So I feel like there's just a little bit less of a mental load that I have to take. But I basically work in whatever project I'm prioritizing that day. Okay. [SPEAKER_03] Got it. [SPEAKER_03] And show us email triage. [SPEAKER_03] Because I've done a video on email triage the way that I use inbox sweep or now we're calling it tend. [SPEAKER_03] And this looks like you're still using the original, but I think you've made some custom modifications, which is another thing that I love. [SPEAKER_03] I built an open source app that lets you turn your emails into cards and will blur anything out that you don't want people to see. [SPEAKER_03] But this look is different from the app that I made. [SPEAKER_03] So tell me about how you use it, how you do your email now, how it has changed things for you and what modifications you've made. [SPEAKER_03] Yeah. So when in V one of, oh, thank you. In V one of the app that you shared with me, it was obviously very custom to you and it had kind of buttons in order to archive or send emails. Mm-hmm. There's a few things that I need to do in my inbox. I'm either delegating something, I am tracking it in Asana. And we work with clients that have hundreds of employees and we need to track what is going on across the different teams that we're working to support. So there's a big mental load when I am triaging my inbox and I basically created a second brain in my updated version of the inbox because my email app can do a few things. So we can maybe blur out any of this that we shouldn't be here. But as an example, my inbox was trained on a ghostwriter that I built a year ago. It was like one of the first skills or prompts that I built for myself. So it's trained on the most recent 150 emails that I've sent. And it understands all of the different contexts in which I need to communicate. And so now it is overlaid on my inbox. It has all of the contexts of the work that I'm doing across all prospective clients and existing clients. It drafts a note in my voice. And there's a few things that I can do. One is I can approve to send it. So I could just click that button and it'll get sent. I can ask to rewrite it. [SPEAKER_00] This was one of the great original buttons that you had in your app. I can archive it if I don't want to reply to it. I can archive it. Maybe this is something that I don't need to reply to, but it needs to go into its own markdown file. So every client that I work with has its own markdown file. And basically at this point, my email knows what's going on more than I do. So whatever it's drafting is probably slightly more accurate than what I would have come up with. [SPEAKER_01] So sometimes I don't need to reply. Someone else might reply, but I do want that context to go into the markdown file. We can just archive it if we don't want to reply to it. We could archive it. Maybe this is something that I don't need to reply to, but it needs to go into its own markdown file. [SPEAKER_00] So every client that I work with has its own markdown file. [SPEAKER_00] And at this point, my email knows what's going on more than I do. [SPEAKER_00] So whatever it's drafting is probably slightly more accurate than what I would have come up with. [SPEAKER_01] So sometimes again, I don't need to reply. [SPEAKER_00] Someone else might reply, but I do want that context to go into the markdown file. [SPEAKER_00] If I click the task button, it'll become an Asana task as well. [SPEAKER_00] I can click a few of these things. It can just go into spam. And then there's a save action here. So this is the kind of thing that's just so insane because you can build an app for yourself on the go, right? I was realizing I need to triage my inbox and send stuff to different places. Yeah. And I could just ask Codex to build a button that made that integration and then keep using it on the go. I remember we were sitting in the office on a Sunday and you were making this extremely complex flow chart. [SPEAKER_03] Do you have that? Can you show the flow chart? [SPEAKER_03] Oh God. [SPEAKER_03] Because that was a moment where I was like, holy shit, she gets it. [SPEAKER_03] Bring up the flow chart. [SPEAKER_03] We want to see it. [SPEAKER_03] Let me see if I can pull up the flow chart. So what you're seeing here at a high level is. Can we zoom in a little more? [SPEAKER_00] Yeah, sure. Go for it. [SPEAKER_00] A high level, this is a sales pipeline management flow chart. [SPEAKER_00] And so this is the kind of thing that Atio just does really well, right? [SPEAKER_00] You import the logic and then it could help you manage your pipeline at scale. Is this for your email or is it for Atio? [SPEAKER_00] So this is for Atio, but this is the same logic that I need to use when I am triaging my email. [SPEAKER_00] So actually it goes to both places. [SPEAKER_00] Okay. [SPEAKER_00] And when we get an inbound and it comes to my email, depending on whether it is a fit for the work that we do, there are different kinds of emails that need to be sent. [SPEAKER_00] And then obviously that advances as the conversation evolves. [SPEAKER_00] So this is the logic that enables me to do this. [SPEAKER_00] And it's the same logic that enables Codex to do this. You made this and then how'd you feed it into Codex? [SPEAKER_00] I PDF'd it and shared it with Codex. Okay. One thing that's really interesting about this is what's really hot right now is loops. And everyone's saying loops, but no one knows what loops are. [SPEAKER_03] This is an example of a loop. And the way to think about loops is I've been using this metaphor a lot. [SPEAKER_03] Previously knowledge work, whether it was code or writing or email or whatever, it was very similar to sculpting where when you're sculpting every single thing that happens on the sculpture is something that you did with your hands. [SPEAKER_03] I think that knowledge work now is turning into something like gardening, where when you're gardening, you're creating the conditions for the growth to happen, but you're not making the plant with your hands. Yeah. [SPEAKER_01] And that's what a loop is. Instead of doing any individual email, you are building the system that does your emails for you. And you're intervening at different parts of the process. [SPEAKER_03] Like one of the things we talk about a lot is the human sandwich at the beginning and at the end to say, this is maybe worth my time. [SPEAKER_03] And then I'm refining the draft or something like that. [SPEAKER_03] And you're trying to compound it. [SPEAKER_03] So you create a flow chart, you do your email with that flow chart that represents a loop. [SPEAKER_03] And then every time you're done, you can compound learnings back into the system so that it gets better over time. [SPEAKER_03] Right. [SPEAKER_00] Yeah. I mean, I think this really is just an evolution of the model manager analogy that you shared four years ago, which is we are going from using these systems effectively as individual contributors, right? Where we are asking them to do this one thing really well, or a small set of things really well, to creating a system, which is something that a good manager does when they have a big team that they need to help operate. Couldn't be me. [SPEAKER_03] Could not be me. But I'm glad that you're able to do that. But it's the same thing, right? [SPEAKER_01] Yeah. [SPEAKER_01] Yeah. [SPEAKER_01] You need to create the conditions to help people succeed. [SPEAKER_01] And similarly, you need to create that shared context for AI. [SPEAKER_01] Okay. So is there more on your email app to show us? [SPEAKER_00] I think that might be it on the email app. There's a bunch of other things that I'm doing in Codex that I can share. [SPEAKER_03] Show us some more stuff. [SPEAKER_03] Because again, the email itself, I think is life changing. [SPEAKER_03] It's been life changing for you. [SPEAKER_03] You get a lot of emails. [SPEAKER_03] I get a lot of emails. [SPEAKER_03] I think we're both getting through our emails way faster than we ever have before. [SPEAKER_03] Yeah. Which is crazy. [SPEAKER_03] So what else? [SPEAKER_03] Um, what else? [SPEAKER_00] So I can also share, maybe on the personal side, I could share a little bit of my. So maybe I'll anecdotally share my best loop that I've run is when we were setting up Atio. I basically had this moment where we're working with this really fantastic team who is helping us organize the logic of the CRM. And they asked me to enrich the information based on context of what had happened on the calls and what had happened in the emails. And my favorite loop that I've run so far on Codex is I just gave Codex a A bit of my, so maybe I'll anecdotally share my best loop that I've run is when we were setting up Atio. I basically had this moment where we're working with this really fantastic team who is helping us organize the logic of the CRM. And they asked me to enrich the information based on some context of what had happened on the calls and what had happened in the emails. And my favorite loop that I've run so far on Codex is I just gave Codex a goal, which was to set up my CRM to accurately reflect what had happened in my conversations and in my inbox for each one of the hundreds of conversations with clients and prospective clients that we've had. [SPEAKER_01] And I gave it a more robust prompt and direction in order to do that. And I think six hours later, I went to sleep and six hours later it was complete. I woke up to effectively a CRM that was fully set up and had done what would have been weeks of work that otherwise I would have had to do. That was actually only possible because of the fake jam, because of this logic, it could make good decisions, make good calls with the shared context that we had created. And it's one of those moments of joy and delight with AI where I wake up and my quality of life has improved as a result of this loop. [SPEAKER_01] I guess before we move on from this, you do a lot of consulting. [SPEAKER_03] We do a lot of consulting with executives at big companies, at tech companies, at hedge funds, at PE firms. We do a lot of training of those people, training of their teams, trying to help organizations get more AI pulled like this and do work like this. So what is the takeaway for someone like that who's listening about a workflow like this and how they should think about whether and how to start incorporating some of this into their work day? [SPEAKER_00] I think my first tip would be to start with the systems that you have already. So if you are already managing a big team and you have KPIs and shared goals and OKRs that you're tracking, the same architecture or system that you're using to guide your team, give to AI, provide to AI if that is something your company allows. [SPEAKER_00] And then think about what are the tasks that you want your people to focus on and to do, right? So at the end of the day, only I can get on calls and have productive conversations for clients. For now. Mike Taylor on my team did recently tell me he cloned me. [SPEAKER_01] What's happening? Just remember me as the original version of Natalia. [SPEAKER_01] How do we know that you're not already a clone? I don't actually know. We'll never know. I might be a hallucination. [SPEAKER_01] Yeah. [SPEAKER_01] So yeah, start with that shared context, that shared infrastructure. Think about what are the things that you want all your people to do and then start with small tasks. I think the single biggest mistake that I often still ambitiously make and also see our clients make is you want to just remake the whole thing. You want to be AI-pulled, be AI forward, just be an AI first organization. And so often that just means you need to standardize and write down how you do a single thing really well. And if you do that and do the next task and define what that looks like and what it's done when it's done really well, you can end up with these more complex systems that can do sophisticated work for you. But the work, at its baseline, it's not particularly sexy. It's just you having to read a markdown file or a very simple set of instructions and starting there. So if you're one of those people and you want to try something like this workflow, by the time this video is out, by the time this podcast is out, we will have an open source version of the email sweep app that Natalia just showed. [SPEAKER_03] We'll put a link in the description. You can just throw it into Codex or you could throw this video into Codex and Codex will just watch it and then just make something that works like it, but for you. [SPEAKER_03] But let's keep going. I want to do some more. I know you have some personal projects and other things that you wanted to share. [SPEAKER_03] Sure. I'm personally fascinated by the role that AI will have on how we run our lives. I think there's just so much that needs to get done. And so many of those things are administrative tasks that I just can't find time in the day to do. [SPEAKER_00] And so one of the most recent things that I asked Codex to do, I gave it a goal to basically create an app that triages my dad's care. My dad is 81. He's the best. He works with multiple nurses who support his care. [SPEAKER_00] And there's just a lot of health things that need to be triaged, right? Medical appointments, followups from recent procedures. WhatsApp threads with the nurses, with my family. [SPEAKER_00] And so what Codex helped me do was basically create an operating system for how, as a family, we could triage my dad's care. I had this long project, this is a 13 hour project that Codex worked on to basically help go from a prototype to creating a full app that could help us with my dad's care. [SPEAKER_00] I'll pull up the site here. It's now a live app. [SPEAKER_00] So what we're seeing here is the portal that my family shares for tracking what is going on with my dad's latest in his health. And so we get Google form reports from the multiple nurses that support him. [SPEAKER_00] I had this long, this is a 13 hour project that codex worked on to help go from a prototype to creating a full app that could help us with my dad's care. [SPEAKER_00] And I'll pull up the site here. It's now a live app pool. All right. So what we're seeing here is now the portal that my family shares for tracking what is going on with my dad's latest and greatest in his health. And so we get Google form reports from the multiple nurses that support him. And then we also have a WhatsApp thread of casual updates of how an appointment went or how his dosage on a certain medicine is going. And so what I have here is a top line, here's the latest. I'm Colombian. So usually this is happening in Spanish, but sometimes if it's the middle of the day and I need to know what's going on, I will just toggle it. And it'll just give it to me in English so that I can digest it a little bit faster. [SPEAKER_00] But really what we have is this one central place where instead of having to dig through all of these different threads and sources of information, Codex has just made it really easy to digest all of that information in a single place and to allow us to support my dad. [SPEAKER_00] And what we can do best, which is to be present and loving as his family. [SPEAKER_00] And your other family members are also accessing this. [SPEAKER_00] Are they also accessing it with Codex or how does that work? No. So this is a password protected website that we use and share. The nurses have a version of it so that they can also see what the other nurses have been working on. [SPEAKER_00] So there's continuity in care and you'll love this. Dan, there is a tracker for the different things that each one of us is responsible for and should be following up on. Right. Which are things that we all have, personal busy lives that we need to do. And based on what's going on in our conversations, these things will get either highlighted as things that have not been resolved or they will just be completed and grayed out. Really cool. So this has been amazing. What do the nurses think? Are they just like, what the fuck is this? This is the most organized family I've ever run into. [SPEAKER_03] What are they thinking? [SPEAKER_03] Do they like it? [SPEAKER_03] You know, it's funny. [SPEAKER_03] I think a really good tool is not about the tool. [SPEAKER_01] I think the nurses just feel like we are more proactive in showing up around the topics that they need help with. [SPEAKER_01] Right. [SPEAKER_01] So I think for them, we've just been better partners to them. I love it. It's just one of those things where this is so obviously useful and good for you and your family and for people. And I think that gets missed so often when we talk about it being great at coding and stuff like that. [SPEAKER_03] And it's like, actually, yes, it is. [SPEAKER_03] And you can use it to do stuff like this. [SPEAKER_03] And people don't realize they don't realize that they can do that and how available it is and how applicable it is to all of the tasks and all of the stuff that we have to do, whether it's caring for a family member or anything else in our lives, that it takes a little bit off your plate. [SPEAKER_03] Yeah, definitely. [SPEAKER_03] I mean, I'm just so bullish on women using AI and all of the administrative tasks that will suddenly be taken care of because now we have this super powerful tool that can support on those things. [SPEAKER_03] I know Claire has talked about that, Claire Vo, who we love. [SPEAKER_00] And the cut recently ran a big piece on how moms are using agents to do something similar. [SPEAKER_00] So really excited about that space. So I know one of the other things that's happening for you is not only you're building these apps, but you're building artifacts that help you, we talked about this a little bit, that help you learn stuff, for example, or just generally navigate the world. I think people think of AI as being, oh yeah, I guess it can generate text documents, slop text documents, but I think you're using it in a way that helps with rich information transfer that I think is really important. Can you show us some stuff? [SPEAKER_03] Yeah, sure. [SPEAKER_03] So maybe one example of that, I love Claude artifacts. [SPEAKER_03] They're just so cool and powerful. One example of that recently is from a trip that I took my mom on to New Orleans. So of course the thing that I was most excited to learn about was the pump system that New Orleans uses, which is just incredible engineering. And the thing that I just don't have time to do a deep research into. And so what I did going into this, it was French quarter fest when we were going over the weekend. [SPEAKER_00] And so I created these artifacts on the go as I would come across things that I was interested in seeing or learning about. [SPEAKER_00] And it would basically give us guides in Spanish so that we could both share in what was interesting to us as we were walking around the city. [SPEAKER_00] And it would also actually it was French quarter fest, not jazz fest. [SPEAKER_00] The jazz fest was the week after. What it would do is it basically I asked it to read through my Spotify playlists to get a sense of what kind of music I liked. [SPEAKER_00] And then to look at the lineup that we had for French quarter fest. [SPEAKER_00] Oh my God, that's so cool. [SPEAKER_00] And then to basically select which bands it thought we were most likely to want to see. [SPEAKER_00] And so it was amazing. [SPEAKER_03] It was great. [SPEAKER_01] You know, it's just the Timba and the salsa bands were the ones that were highlighted. [SPEAKER_01] And so we could really use our time optimally so that we could go and explore New Orleans. [SPEAKER_01] And then when we were showing up for French quarter fest, we could go and see the bands that would most resonate with us. [SPEAKER_01] Which just feels like a really fun use of AI. [SPEAKER_00] Incredible. [SPEAKER_00] I love getting to talk to you. [SPEAKER_00] I always learn something when we chat. [SPEAKER_00] And so it was amazing. [SPEAKER_03] It was great. [SPEAKER_01] The Timba and the salsa bands were the ones that were highlighted. [SPEAKER_01] And so we could really use our time optimally so that we could go and explore New Orleans. [SPEAKER_01] And then when we were showing up for French Quarter Fest, we could go and see the bands that would most resonate with us. [SPEAKER_01] Which just feels like a really fun use of AI. [SPEAKER_00] Incredible. [SPEAKER_00] I love getting to talk to you. [SPEAKER_00] I always learn something when we chat. [SPEAKER_03] And if you want this kind of thinking inside of your organization, Natalia runs our consulting. [SPEAKER_03] So if you want to get this out into your executive team, into your product teams and your engineering teams, reach out to every.to/consulting. [SPEAKER_03] And Natalia, we'll have to do this again in a couple months. [SPEAKER_03] Yeah, we will. [SPEAKER_03] All right. [SPEAKER_03] Thanks for having me, Dan. Thank you. Folks, thanks for having me. You absolutely, positively have to smash that like button and subscribe to AI and I. [SPEAKER_02] Why? [SPEAKER_02] Because this show is the epitome of awesomeness. [SPEAKER_02] It's like finding a treasure chest in your backyard. [SPEAKER_02] But instead of gold, it's filled with pure, unadulterated knowledge bombs about ChatGPT. [SPEAKER_02] Every episode is a roller coaster of emotions, insights, and laughter that will leave you on the edge of your seat, craving for more. [SPEAKER_02] It's not just a show. [SPEAKER_02] It's a journey into the future with Dan Shipper as the captain of the spaceship. [SPEAKER_02] So do yourself a favor, hit like, smash subscribe, and strap in for the ride of your life. [SPEAKER_02] And now, without any further ado, let me just say, Dan, I'm absolutely, hopelessly in love with you. Uh, I love getting to talk to you. I always learn something, uh, when we chat. Uh, and, uh, if you want this kind of thinking inside of your organization, uh, Natalia runs our consulting. So if you, if you want to get this out into your executive team, into your product teams and your engineering teams, reach out every.to slash consulting. And Natalia, we'll have to do this again in a couple months. Yeah, we will. All right. Thanks for having me, Dan. Thank you. Oh my gosh, folks. Thanks for having me. You absolutely, positively have to smash that like button and subscribe to AI and I. Why? Because this show is the epitome of awesomeness. It's like finding a treasure chest in your backyard. But instead of gold, it's filled with pure, unadulterated knowledge bombs about chat GPT. Every episode is a roller coaster of emotions, insights, and laughter that will leave you on the edge of your seat, craving for more. It's not just a show. It's a journey into the future with Dan Shipper as the captain of the spaceship. So do yourself a favor, hit like, smash subscribe, and strap in for the ride of your life. And now, without any further ado, let me just say, Dan, I'm absolutely, hopelessly in love with you.