Calum Johnson Show: w/ Allie K. Miller on AI Agents For Business, Content & Life (Transcript)

We’re in the middle of this program right now. I taught the first session — it was how agents work, how to build them, how to set up documents that personalize it for yourself. This whole session was 60 minutes. Within the first 48 hours, all of these people, all of whom are non-engineers, some are consultants, some are entrepreneurs, some are execs at Fortune 500 companies — they’re all building their own autonomous proactive agents. Zero code.

And it’s the ability to go into Claude Code or Codex, Perplexity, Gemini, whatever, any one of these systems, give it a little bit of information about yourself, a little bit of information about how you want it to behave, and a little bit of information about the actions you want it to take. And yeah, maybe it’s a little annoying to figure out where all those files go and how to format it or whatever. But these systems are now good enough to also talk to in natural language, to have it help you help it. And we weren’t at that level 3 years ago, right? Like, you could talk to ChatGPT and say, “All your answers are really short, make them longer,” and it would remember that for that one conversation you were having. But we have these systems now that can hold context, that can hold memory.

And so all these people, they’re building daily briefings so that every single morning they have a rundown of their calendar, daily action items. There’s one woman who put her entire to-do list in Notion and every single day kicks off Claude to literally review her to-do list and take on as many tasks as possible. There was a woman in the group who’s obsessed with this chicken truck that comes to her neighborhood.

CALUM JOHNSON: Yeah.

ALLIE K. MILLER: But never remembers this chicken truck’s schedule. And so every single day she just gets a readout of the chicken truck’s schedule. Another person loves Lenny Kravitz and gets a readout of his daily concert schedule. I mean, it’s crazy. But all these — I don’t need a chicken truck, I don’t need a Lenny Kravitz readout. I love the idea that we have proactive memory-based, context-based agents that require zero coding, that you can talk to in natural language, that any single person can build, and they can help solve the chicken or Lenny Kravitz problems that you have.

Two Worlds: Power Users vs. Everyday People

CALUM JOHNSON: To your point, I think we’re in a really interesting time, and this is what I’ve seen from just listening to people. There’s almost like two worlds. You have these AI tech-heavy, almost like insider power users who have like 7 AI agents running their workflows and their company.

ALLIE K. MILLER: 7 would be very low, but yes.

CALUM JOHNSON: Even more. You have a better sense even than me. And then on the other hand, I think you have almost like the everyday person who’s maybe heard of ChatGPT or used it, is using it really like they use Google. And they’re worlds apart. And I find almost the mentality is also the feeling around AI. I think the power user has all this excitement and this opportunity. And then the everyday person, I hear a lot more anxiety. Can you kind of just speak to that disconnect? What is it that the power user is so excited by when they think about AI and how it can be in their life? Versus more so what I hear with everyday people and also in the media, which is fear and anxiety — “This is going to take my job.” Why is there that disconnect in your mind?

The Mindset Shift: From Tool to Operating System

ALLIE K. MILLER: So first, both groups do have anxiety. This idea that the super users have totally figured it out and everything in their life is perfect and is tied in a perfect bow and they just have AI working for them all the time — it’s not quite the case. So neither group is alone in that feeling of anxiety.

I think actually the big difference is not that the super users have seen opportunities that the surface users have not yet seen. I think it’s the mindset approach for how they even look at these tools. This is what I’ve learned over the last 3.5 years of teaching these things, again, to board chairs all the way down to interns at companies.

If you look at this thing as a box where you have to memorize buttons — and you have to tell yourself, “Ah, yes, in order to have a live video conversation with ChatGPT, I must hit the bottom right corner and then hit the video button, and that is how I function” — if you treat it like a tool where you are learning the motions and buttons, then you’ll keep it in that box. And you’re going to hear these stories of people building 30, 100, whatever number of agents and going, “How on earth is that person able to do that when I’m sitting here memorizing a button?”

What the super users have taken in and spun on is that AI is not a tool. It’s AI as this operating system for how you handle your work, how you think of your personal life. And when you think of it as this wider enablement, one, you start to get way more excited to learn about the really annoying backend-type features like what should I title this file. But the second is you get a little bit more creative and wide about where to use AI.

So if you thought of it as a tool, you might realize that you can take a photo of any sign in any grocery store and translate it into 80 languages, or ask it for nutritional facts, or ask it for similar snacks, right? All of those are information gathering.

The Spectrum: From Information Gathering to Action-Oriented AI

As you move along the spectrum — because it’s a spectrum, not a binary — you might go, “Okay, first I was doing information gathering. Now I’m going to do a little bit smarter.” Let’s say you’ve got all of your podcast transcripts, right? You’ve done hundreds of episodes at this point. Let’s say you take all your podcast transcripts. If you were a surface user, which I know you’re not, you would give these systems your entire podcast set, the transcripts, or maybe even a couple, and you would say, “How many people out of this batch work in AI?” or “Pull interesting quotes about AI from these episodes,” or “How many people focused on self-help and who were they?” You would do an information gathering.

If we move up a little along that spectrum, maybe you’ll compare and contrast and say, “I’m Calum. How has my thinking changed over the last 100 episodes? Have I become a stronger or weaker interviewer? Obviously stronger. Why or why not? Give me proof. How do I compare to maybe other podcasters? What is my competitive advantage? Forecast my next 10 guests.”

And then if you really keep going up, then you’re going to say, “Wait a second, I’m going to be more future-looking, forward-looking. I want an astronaut as my next guest, right? Knowing every single thing you know over the last 200 episodes, find me the exact astronauts that I should bring on, research all of them, research all the podcasts that they’ve been on, forecast which one is most likely to come onto my podcast, run 10 different simulations by that person to see what type of conversation we can have, then run 20 cold emails based on those episodes to then figure out the 6 sentences that I should send to their publicist that’s most likely going to convince them, and send it and then follow up with them if I—”

The ability to immediately go into goal-oriented and action-oriented stuff — that is what this spectrum is. It’s going from tool into operating system. It’s going from information gathering to action-oriented. It’s going from single prompt into just wider context about your life. And that’s the shift, right? It is a full mindset shift.

There’s nothing technologically that is holding surface users back. Nothing. Because I see it every single day that non-engineers are crushing it and that they’re able to do it. It’s just they’ve set aside a couple of hours and they’ve figured out that this is a spectrum that they can surf along. And that’s literally it.

AI Tools Explained: From Chatbots to Autonomous Agents

CALUM JOHNSON: You know, when you mention non-engineers crushing it, you’ve seen so many examples of it, and I even think about, we have a lot of people that watch this show that want to start businesses. They’re either entrepreneurs or they’re aspiring entrepreneurs. Are there any examples that come to mind that you can share where actually using AI in the way that you just described has actually pushed people towards starting their first company that actually took off, or building a product? Like, what comes to mind?

ALLIE K. MILLER: Yeah, I think so. There’s a couple different types of companies, right? You could start a software company, you could start a services company, or maybe you already have a product or service company and you’re just trying to scale that.

So I worked with one founder who had about 12 clients, and she could only service 12 clients at once because her time was fully maxed out. In chatting with her and figuring out what the bottlenecks were about her business, I learned that the bottleneck was not in actually servicing them. It was the massive amount of lead intake in qualifying the right clients in the initial kind of information sharing.

And so she was able to go on to, I think she went on to Lovable and built out an entire interview system to then give to potential clients. And that’s also not really where the trust was built anyways. It was like filling out a form with the doctor. And so these potential clients go through an entire interview process, and she was able to set up a second-tier cohort-based thing. She went from 12 clients to, I think, something like 35, 36. She 3x’d her revenue just by having AI be a scaling mechanism on the things that she was already bottlenecked on.

I have other friends who are coaches who have set up AI twins of themselves so that these clients — she didn’t grow the number of clients, but she was able to increase what they were charged because they didn’t just have the one call with her every 2 weeks or month. They were helped 24/7 around the clock with every single one of her pieces of content and her private data sources that she was putting into these systems. So these people could get help, right?

I also have another friend who runs a product that is both physical and software. She has several agents that are run through largely OpenClaude that she might have a React engineer and she might have this type of engineer and that type of engineer all helping her on the software product, working alongside her actual software engineers. And then she also has OpenClaudes that help her in her personal life. And can I just tell you the craziest thing that she shared?

CALUM JOHNSON: Go for it.

ALLIE K. MILLER: She’s a very busy mother. She’s at home, she’s staring at her desk, and underneath the desk is just hell of cable management. And she has this OpenClaude on her phone and decides to just try it out. She takes a photo of the messy cables, sends it to OpenClaude, and just says, “Fix this.” Okay? This is an online-only AI agent that she has fed something physical and has said, “Hey, fix this.”

This system goes online onto Nextdoor, posts job openings to fix this cable thing, finds and narrows down potential people to help, sends those options back to my friend Catherine. Catherine reviews them, picks the guy, sends it back to OpenClaude. Then she forgets about it and at 6:30 PM someone knocks on her door. She opens the door and it’s the guy there to fix the cables. It was done in 30 minutes for $30. She Venmo’d the guy and it took less than 12 hours start to finish with minutes of her own work. And she had been putting this task off for a year.

AI Agents: Transforming Business and Personal Life

CALUM JOHNSON: Hmm.

ALLIE K. MILLER: The ability for AI agents to not just help us in our business, in a product and coaching and service scale way or a software way and scalability. But the ability for AI agents to take meaningless crap off of our plates in our personal lives so that we have more cognitive magic to be able to apply to the rest of our life. That’s incredible. That is incredible.

And so we’re going to see AI calling people in the real world, bringing humans into our space, hiring humans, helping us recruit. We’re going to see AI helping us scale software, helping us take a website and create an iOS app out of it. We’re going to see people all have their own AI personas that you can scale out your own work so that you can service clients while you’re sleeping. All of that is happening all at the exact same time with the exact same technology.

CALUM JOHNSON: Okay. You just shared it so eloquently. The agents, ChatGPT, and the fact that the way that people are using it is allowing — it allows you to customize an experience, which you can charge higher prices off the back of that. It also allows you just to save time.

ALLIE K. MILLER: Yeah.

CALUM JOHNSON: And to your point, you have more of this cognitive creative magic that you can bring to your job or your business. But you know what, because I’m so cognizant of where I was 3 to 6 months ago when it came to this.

ALLIE K. MILLER: Yeah.

CALUM JOHNSON: Can you explain to people clearly the difference between ChatGPT, which I think most people have heard of — they have over 200 million users in the US, the majority of people have used ChatGPT. And then you mentioned Lovable, which is like there’s a distinction between that and ChatGPT, and people might be hearing that name for the first time. And then OpenClaude, which is something that I’m getting my head around right now, which actually has this autonomous element.

I think one of the things when it comes to AI, the space is moving so quickly that it’s like I was just learning about ChatGPT and now she’s talking about it called the service guy and he turned up and fixed the cable. Can you explain those three different things, and then also the evolution and how quickly things are moving right now?

ChatGPT, Lovable, and OpenClaude: Understanding the Differences

ALLIE K. MILLER: So let’s first start with just what are all these different tools. You can think about it as some tools are going to be horizontal where they can really do anything for you within whatever their parameters are. And some tools are really, really specific.

In the horizontal space, you have some platforms that are more chatty. So ChatGPT is single-threaded. Chatbots, for lack of a better word. It’s an assistant. It’s the ability to ask a question, come back. It can read your email. It can access the internet. You can upload documents or images to be able to ask. You can have live video conversations while you’re walking through a store. All of that is the paradigm of the AI assistant, right? Like retrieving of knowledge. Okay.

The world that we walked into, early to mid-2025 and later, is agents — the ability for AI to actually take action. So not just read our email, but reply to the last 15 emails. Or not just look at the things that we’re looking at in a store over video, but also go ahead and figure out, out of these 5, which one’s the best and order it for me and ship a second version to my sister so that both of us can have the same product. Right? It’s the ability to make purchases, to send things, to edit things, to deploy, to post on your behalf, to literally take action in the real world, to use tools and to complete real tasks.

I would even go so far — this is a hot take — as to say that none of the assistant things completed tasks for me, right? Like, sure, I could have had it write 15 of my emails. I still had to send every single one of them. I still had to be beholden to sitting at my computer and going through this whole thing. It didn’t take meaningful work off my plate.

Because systems can now take action like OpenClaude, like Claude Code or Codex, we now have AI systems that can take actual tasks off of our plate. So we now have things that feel much more like personal assistants or chiefs of staff that can check the entire box. And that shift really started in about March 2025 because of model performance, because these models can now reason through lengthy steps to be able to complete a task. And they’re very accurate in figuring out what those steps could be. And they’re very performant in figuring out what tools to use and actually using the tools in the correct way.

OpenClaude being much more autonomous and proactive, and having systems that kind of update its own memory in a much stronger way than things like Claude Code or Codex. But all these systems are improving every single day.

From Transactional to Transformational: Making the Mindset Shift

CALUM JOHNSON: Yeah. I’m going to share my experience. And I’m nowhere near as much of an expert as you.

ALLIE K. MILLER: No, please listen. You are going to inspire millions of people just by saying the next sentence that you say.

CALUM JOHNSON: Because when I first started using, and it was ChatGPT, it was very transactional. And I remember getting one of the episodes or conversations that we filmed for this show, and the guest talking about how a lot of people use ChatGPT like Google. They ask it, I don’t know, where is this? And it spits out an answer. It’s this very input-output transactional relationship. That’s where I started.

I then kind of moved on to, still within ChatGPT, using AI as an advisor, almost like a counselor. It’s giving you advice on things, or you’re pasting in transcripts. “Oh, what can I learn from this?” So it’s giving advice and expertise.

In the last few weeks, I’ve gotten deeper and deeper into OpenClaude. It’s like we’re getting to the point that AI is almost like an employee. It can do tasks. You onboard it and then it goes, similar to what an employee would do.

And so I actually want to start with kind of that first part of things, the going from transactional to even using it as an advisor, because I’m so cognizant that for a lot of people listening, when I said that most people use ChatGPT like Google, they’re like, okay, that’s me. And so it feels like there’s this huge gap of how do I get from where I am now to the point where I’m using AI to actually handle entire tasks.

ALLIE K. MILLER: Yeah.

CALUM JOHNSON: It’s almost a mindset shift.

ALLIE K. MILLER: Absolutely.

CALUM JOHNSON: And so I think one of the things that you and I can do, which is so powerful, is how do we make that mindset shift real and practical today? And so if someone’s listening to this and they want to just start developing the way that they’re using AI and these tools, what would you say is the step one? Like, okay, today I’m using it transactionally, I’m using it like Google, but I want to just make a little bit of progress towards using it in some of the ways that you’ve described.

ALLIE K. MILLER: Yeah.

CALUM JOHNSON: What’s the step one that someone can take today?

Going Problem First: The Right Approach to AI

ALLIE K. MILLER: My step one is going to piss people off because I’m actually going to say that your step one has nothing to do with AI. I’m going to say walk away from your computer and I’m going to say pull out a whiteboard, a notepad, or Otter.ai or whatever if you want to dictate, and literally just think about your own goals and who you are.

One of the biggest superpowers in the AI age is knowing who you are and what you want and what you don’t want, because these systems can help you better accomplish your goals in a faster way, maybe a cheaper way, and in a stronger or better way. So taking one big step back and saying, “Here are my goals, here’s what I actually want,” — that is the correct step one.

Because then when you’re going to these systems, maybe before you would ask things like, “What’s the best podcast studio in New York?” But if I’m looking at this set of goals and it says I want to interview an astronaut or I want to have a global podcast that takes me to all 7 continents in the next year, I’m not going to ask a question that says, “What’s the best podcast studio?” I’m going to give it all of my goals and all of my context, and I’m going to say, “Okay, now you’re my COO. Let’s go back and forth and figure out — think through 20 possible actions that I could take. Score them on likelihood of revenue, likelihood of impact, and creativity. Walk me through your prioritization and ranking. Pick the top 3. Walk me through the first 5 actions for each of them. Tell me exactly how much each one of those steps should cost. Tell me how long each of those steps will take. Take on as many of those tasks as you can find. Tell me what you need help with.”

And so suddenly we went from going tool first and then asking, to going problem first. And then we just happened to have this unbelievable tool with PhD-level intelligence that we can lean on and use and help ourselves with. But so many people read the news and see some awesome feature and start tool first. If you took a big step back and you said, the most successful people in the world are using this and going problem first, the most successful businesses are going problem first, and you flipped it on its head, you would be in a completely different place with your AI usage in days.

CALUM JOHNSON: Yeah. I can feel the passion when you talk about it — the energy.

ALLIE K. MILLER: Yeah. And also the frustration when I hear that people are still using it as Google, because I just want to be like, this could change your life.

CALUM JOHNSON: Can you share with people — and I love the way that you framed it, instead of going tool first, it’s actually problem first — can you almost take us down memory lane? The first time that for you, you actually made that shift, like you saw in your own life going problem first with an AI tool and the impact that that had, where you had this moment where you’re like, “Oh, this is how I’m going to do it from now on.” Like, what was that light bulb moment for you when you were like, this is the right approach to take to get the maximum benefit with AI?

The Parking Lot Moment: Allie’s AI Light Bulb

ALLIE K. MILLER: I’m a former product manager. And so when you’re dealing with engineers and you have this unbelievable team, you still have to say no to some features, because they only help a couple customers, they would take too long, they’re too expensive, whatever. And so you have this thing called a parking lot, which is like, “Wouldn’t it be cool if we could one day maybe make this feature happen?” And you just keep it as this graveyard of brilliance that you can’t address.

And so GPT-4 comes out and I have a parking lot for my own business, my own ideas, anything that I want to take on that I didn’t have time for, that I didn’t have people for, that maybe I didn’t have skills for, money, whatever the thing was that was holding me back. And I look at that parking lot and I go, “What if I just took a picture of this and just gave it to this model and just said, what can you do to cross off as many things on this as possible? Get me to the highest possible level of completion on any of these things.” And then even when it says, “Okay, I’m done,” say, “No, I know you can do more. Give me at least 10% more. Give me at least 50% more. Get more creative, do more.”

So it was this idea that I already had this waiting room of things that I had been aching for in my life. Again, goal-oriented, problem-oriented. And the models had this moment where they got good enough to start with a decently vague description, break it down, prompt itself, reason through what it could actually do, and then in recent months, be able to take action to actually cross it off my list.

CALUM JOHNSON: I think it’s so powerful because I’ve realized this even in myself — a lot of the times in life you can tell when someone’s on the offensive and being front-footed versus on the defensive and waiting. And I think one of the things that can be demoralizing is when you have so many aspirations and ambitions, but because of time and commitments and responsibilities that you already have, whether that’s your family, your friends, your job, maybe it’s your relationship, you constantly feel on the defensive, like you’re waiting, like, “I don’t have enough time to get to these things.”

And so I think about the power and that shift that you mentioned of being able to go on the front foot, partially because it’s saving you so much time, and it’s also giving you a level of insight that you can action on these things right now.

ChatGPT vs. Claude: What’s the Difference?

CALUM JOHNSON: One of the things that I wanted to talk to you about — you mentioned that shift, and I agree. I think most people have used ChatGPT and had that kind of light bulb moment where they’re like, “This is different. This is something a bit different.” Right now we’re going through, or at least I’m seeing it a lot on social media and I’ve done it myself, people are shifting from ChatGPT to Claude. Can you give that context? First of all, the differences between the two. And then in your mind, why is that happening? And is it even justified? Like, is there an actual benefit to someone that’s listening at home taking everything they have on ChatGPT and putting it to Claude?

ALLIE K. MILLER: Big, big questions. First, I would just like to say whatever AI you want to use, just use something. Like, just for the love of God, please start. So ChatGPT comes out end of 2022. Claude was a fast follow. They’ve been out for basically the same amount of time. ChatGPT has hundreds of millions of active users. Claude is a lot smaller. 2 years ago, 99% of my work was in ChatGPT. Today, 99% of my work is in Claude Code.

I think if you’re trying to be this crazy optimizer, sure, you might be switching between tools. I literally use all these systems. But I think what a lot of people are realizing is that as we move into this — like, AI is an operating system, AI is a coworker, AI is a teammate — vibes matter. They’re all trying to go toward the same zones, but they’re all built a little differently. Vibes tend to be what people are picking one over the other for. And all of these companies are trying to roll out the same features and they’re all a couple months behind each other.

CALUM JOHNSON: It’s so interesting when you share the context, and the point about vibes.

ALLIE K. MILLER: Yes. And it’s actually — there are metrics for vibes.

CALUM JOHNSON: Yeah.

ALLIE K. MILLER: Like, yes, we can measure exactly how good it is at coding, and exactly how it compares on over 1,000 tasks across 9 different industries and 44 occupations. And there are people that are literally just measuring vibes.

CALUM JOHNSON: And what you say is so true. It was my experience of when I first started using Claude versus ChatGPT — ChatGPT is far more agreeable. It was—

ALLIE K. MILLER: Maybe they’ll change that in the next week, you know, who knows.

CALUM JOHNSON: But for now, yeah, it’s so fast moving to your point. Whereas Claude was more like pushing me on my point of view. And so—

ALLIE K. MILLER: Can I just say one other thing?

CALUM JOHNSON: Go for it.

ALLIE K. MILLER: The thing that makes all these systems personalized to you is, yes, you giving it a document that says, “Here’s what I believe in, here’s what I want to do.” But also, you might have had 3 years of conversations with ChatGPT where it built up a whole lot of memories about you. And people were having way more personalized conversations with ChatGPT versus any other player.

What we thought several months ago was, “There’s no way that we will be able to replicate this. There’s no way. If you picked one model or one system 3 years ago, there’s no way you can switch.” Claude proved that you can switch. And it’s literally that you give one prompt into whatever program or system that you’re in today and you just go, “Hey ChatGPT, can you tell me every single big thing that you know about me and the way that I work and what I like and what I don’t like and what was really important?” And ChatGPT gives you a very long document readout and you take that document, you download it, you paste it into whatever system you want to move into, and you go, “Here’s me, let’s rock.”

Like, that ability to port things over — you can’t do that with Gmail into Yahoo or vice versa. It is so hard to move data from one place to another, and it is shockingly easy to take years of personalization and customization and move it and get, I don’t know, 85% of the performance in any other tool.

CALUM JOHNSON: It’s literally 2 minutes, the process that you just outlined.

ALLIE K. MILLER: Yeah. And there are prompts that you can already copy and paste. You don’t even have to write your own prompt. These systems want you to switch onto them. So of course they’re going to make it as low friction as possible to not only use the tools, but to switch from whoever their competitors are.

What Is Claude Code and Why Does It Matter?

CALUM JOHNSON: You know what, because we’ve spoken about the LLMs — ChatGPT, Claude, even you mentioned Gemini — I wanted to talk about vibe coding for a second, and the fact that you even got a bit excited when I said that.

ALLIE K. MILLER: We’ve just been talking about the word vibe so much. You’ve got a shirt that says “stay delusional.” I love it.

CALUM JOHNSON: The vibes are good right now.

ALLIE K. MILLER: The vibes are great.

CALUM JOHNSON: I was on your Instagram this morning and I saw an article that you did with Inc. about how you’re using Claude Code in kind of your everyday to help you run almost like your day to day. And it’s one of those things where, when I looked at Claude Code and I came across it, it looked so technical — like the terminal and even the terminology that people are using. Can you describe for the everyday person who’s non-technical, what is Claude Code? First of all, what’s the distinction between that and just Claude? And then what is the significance of how they could use it in their day-to-day?

AI Agents in Action: Claude Code, Cowork & Real-World Use Cases

ALLIE K. MILLER: So Claude Code, Claude Cowork, Codex, Gemini, Perplexity’s computers — all these things are wrappers with a lot of complexity, but wrappers around the LLM. So around these models like Claude Opus 4.6 or GPT-5.4. And it gives it a zone to run and take actions for you.

It’s just a zone where it can connect into tools like Gmail and Notion and Asana and Shopify. It’s a zone where you can talk to it and ask questions. It’s a zone where you can set up things to be scheduled so that something can happen every day at 7:00 AM. And whether that zone is a desktop app or that zone is a terminal or a mobile app or whatever else comes out, it’s just giving it a box to be able to actually run. So that is what these systems are.

What can they actually do for you? It’s going back to what we were talking about, about assistant versus agent. The ability to access all the documents that are sitting on your literal computer. The ability to be more proactive because you can schedule these things. Up until now, you couldn’t schedule things in Claude. You could kind of schedule things inside of ChatGPT where you could have it run a Google search basically every single morning. But we now have the ability to set up agent-based tasks.

So let’s say that every single morning — let’s say that you’re launching a secondary podcast. So you have this amazing one and you want to launch a second one that’s all about underwater adventures. Maybe you want it to research brand new things in the marine biology space. And so maybe that is automated, but you’re not just having it research marine biology and come back. You are saying, here’s the exact format I want it to come in. Go ahead and use these specific tools. Actually send me a Gmail every single morning with the recap in this exact format in HTML.

So these systems, because they can write unbelievable code, can connect into tools and can complete tasks that require code — like giving you a beautiful email back or editing your website. There are people who every single morning, either screenshot or automatically, will go onto their Google Trends or Google Analytics page, screenshot how their business is performing, put it into a Claude Code or a Claude Cowork, say, “Here’s my business today, give me 5 suggestions on what we should do.” And then maybe you’re like, “I really like suggestion 3. Do it.” And going back and forth and actually taking action. That’s the big change.

CALUM JOHNSON: Hmm.

ALLIE K. MILLER: Okay.

CALUM JOHNSON: There’s so many ways I could take this.

ALLIE K. MILLER: Yeah. And by the way, terminals, yes, are ugly and gross and plain. And even just the font is kind of annoying. Again, I’ve taught now thousands of business professionals how to use the terminal, and within 48 hours they figured it out. I think Claude Cowork and things like Claude Cowork are now good enough that you can start in Claude Cowork and actually not touch a lot of code and you’re fine.

So Claude Cowork is a beautifully designed desktop app that has nice little boxes everywhere. And so you can give it a task and you can track all the steps that it’s going to take and the progress through those things. You can literally have it take action for you on the internet. I can say, “Go and find me 5 Airbnbs in Cleveland for a friend group of 8 people, and we need this many bedrooms or whatever.” And I can literally watch it navigating my Chrome browser for me, but I can do all that from within Cowork.

So just think of it — instead of you telling a single-threaded boring chat thread, “Hey, what’s the question? Here’s the answer. Awesome.” — you’re instead queuing up tasks. That is the core component: not questions, but tasks. And so you open up Cowork and every single time you open up that blank thing, just think: What task would I give to a chief of staff? What task would I give to a senior sales rep? What task would I give to my financial advisor? Along with an actual professional, and just queue up tasks instead of questions.

How Allie Uses Claude Code to Run Her Day

CALUM JOHNSON: Yeah, it’s so powerful. And the Instagram post that I read, the caption was, “Here’s how I use Claude Code to run my day.” So can you just share with people some of the examples? Because you shared a number of them across both your content and in your business that you’re using Claude Code daily. Can you just share some of those examples and why it’s been impactful to you?

ALLIE K. MILLER: Yeah.

CALUM JOHNSON: Can you just share some of those examples and why it’s been impactful to you?

ALLIE K. MILLER: I’ll start with one and then depending on how long the answer is, you can tell me if you want to hear more. I create all my own content and it is a lot to be able to crank out daily content on AI that is always changing, for the last decade.

So I might read some news article. I might read a 200-page research report, but I want a way to put the actual idea out into the world. And maybe sometimes I want it to be a short tweet. Maybe sometimes I want it to be a long LinkedIn post. Maybe sometimes I want it to be an article. Maybe sometimes it’s just a recap for a few execs that I advise.

So I built out an entire app that sits only on my laptop. Very easy — it’s locally run, so I can’t go on my phone and use it. You can’t use it either unless I give you the code. But it sits on my laptop. I open it up and it’s called Voice to Social, which I think makes it obvious what it does.

So I might dictate out 1 or 5 ideas — like, “This is the exact thing that I want to get across. I want to talk about the brand new Claude Mythos model. And I want to talk about how it was very overeager and destructive in early testing. And I want to get across these 3 stories.” I am the one putting in the actual idea. I’m still the one crafting the idea.

Formatting it, changing this little bullet to that bullet — I’m fine if AI helps me scale it out. So I put it in once. And it gives me back 9 different artifacts that can go in all of these different places. And all 9 are happening in parallel as well. So I am not sitting there waiting, going, “Now can you make it into a tweet? Now can you make it into an email?” You put in one and you get back in parallel all of these things that are based on my actual brain. I didn’t lose any critical thinking. But it is immediately enabling me to spread that message to the millions of people that follow me in whatever format I need.

So that is something that has saved me hours and also lets me share content faster. If you had looked at my content calendar 4 years ago, even before all this craziness started, I was writing so many more things than ever saw the light of day. Hundreds of posts and tweets and thoughts and ideas never saw the light of day. And now because the cycle is so much faster, I can get anything out that I want. It’s actually disarmingly easy.

And so that entire interface was built in Claude Code. I made it black and yellow to match my brand guidelines. I added in the ability for me to dictate in the actual program itself — that was built in with the Whisper API from OpenAI. I built in the ability to click one button and it sends to Notion without me copying and pasting into Notion. I built an integration into Slack that allows me to press the button and immediately assign demo creation to our intern. Another button direct to Slack into a different channel that assigns it to our designer to make a graphic that goes along with it. A button that sends it into Gmail drafts so that I get 20 different versions for different clients that I might be pinging this to.

The dissemination and some of the customization can be AI-enabled. It’s that core idea, that critical thinking, the actual initialization — that’s still all me. But that entire process, all Claude Code.

Content Across Platforms: Staying Native Without Losing Your Voice

CALUM JOHNSON: That’s fascinating. And it’s interesting as well, because selfishly, I look at your Instagram, your Twitter, your LinkedIn, your YouTube — you post on all of those. Maybe not YouTube, but basically every day.

ALLIE K. MILLER: Yeah.

CALUM JOHNSON: And all of the posts are contextual, feel native to that platform.

ALLIE K. MILLER: Yeah.

CALUM JOHNSON: And it’s interesting that it’s actually Claude Code that’s helped you get to that.

ALLIE K. MILLER: And my followers are a little different and I had to tweak — “Oh, my voice isn’t quite like that, it’s more like this.” My followers on X are a lot more engineers, and so that’s going to be a little bit more tailored. My followers on LinkedIn are executives, and so I might be talking a bit more around the overall business impact of something. My clients — they’re extreme experts in some of these fields because we’ve been talking about things for years. And so I can kind of skip some of the basics and jump into depth. And so every single one of those should be customized.

YouTube — I just, for whatever reason, every time I have laryngitis or no voice, I’m like, “This is a great time to record an hour-long video.” And I just let it roll. But the core idea is still me. That’s the big thing I want people to hear: Claude or ChatGPT or Gemini or whatever is not rewriting my creativity. It’s rewriting the manual labor of the last mile of customization and deployment.

Getting Started with Claude Cowork: A $60 Solution

CALUM JOHNSON: You know what, because there’s going to be someone that’s listening and they’re going to get inspired by what you just said. And I think it’s such a common pain point, especially with content — like, “I want to post on all of these different platforms, but I don’t have the time to do YouTube, Instagram, Twitter, TikTok, LinkedIn.” I hear about that all the time. I feel it even in my own business now.

What people are going to do next is they’re going to go and try Claude Code, and they’re probably going to have a similar experience to the one that I had, which is as soon as they open it, there’s going to be a level of intimidation of the way that it looks. And then from there it’s going to be like, “How long is it going to take me even just to get started?” And set up on this thing. Can you speak to — if I almost want to make the perfect first step — like I want to go in with the right mindset.

ALLIE K. MILLER: Yeah.

CALUM JOHNSON: And I also want to start in the right way so that I can keep going. It’s not like I start doing it for 10 minutes, then I’m like, “This is too—”

ALLIE K. MILLER: I’m going to give you a solution. It’s going to cost $60 total. And there are ways to make this cheaper or whatever. But if you’re like, “Give me the easiest possible freaking things,” these are the things. They each cost $20 a month.

First, you’re going to get a ManyChat account, a Repurpose account, and a Claude Pro account — that’s the $20 a month one. That’s the $60. You are going to download the desktop app — you just Google “download Claude desktop app,” download it, switch to the Cowork tab. That’s step 1. You are going to enable the Chrome plugin, which is literally just a toggle. You are going to create the Repurpose and ManyChat things.

You’re going to record your reel and whatever as you usually do. And when you post it — like all those replies where it’s like, “Reply with the word Calum and I’ll send you my awesome getting started guide” — you can literally ask Cowork to go into these two systems, Repurpose and ManyChat, and just do all the clicking for you.

When you enable that Chrome plugin, that means that it can take action in web browsers on your behalf. So even if you’ve never used these tools, you can go into Cowork and say, “Every single day I’m going to give you a reel. I want you to take that file. I want you to put it into Repurpose,” or “Hey, I’ve already uploaded this file, manage it and disseminate it on these 4 channels.” And then on ManyChat: “Can you set up a new automation where if people reply with the word ‘workshop,’ I’ll send them the link to my workshop?”

Claude in the background is going to make all those clicks happen and navigate the literal apps for you. So yes, you had to spend maybe the $20 to get it started, but even if those $60 got you one more client, it would be worth it. So you do have to make a little bit of a trade-off, but Cowork, I actually think, is the right place to start for most people. And I wouldn’t have said that a month ago because it was feature-light, but that is where I would start now.

CALUM JOHNSON: Yeah.

Progressive Trust and Getting Started with AI Agents

ALLIE K. MILLER: And turning on Chrome means that I can take immediate action inside of those settings and every other tool plugin in CoWork. Like, I don’t know if you’ve gone to the connections page, but if you go into customize and then connectors, it’s literally like a list of 100 toggles, right? Like when you set your cookies on a page and it’s like, “Is it okay if I sell your marketing? Is it okay?”

CALUM JOHNSON: Yeah.

ALLIE K. MILLER: They all just look like that. It’s 100 toggles that just goes: Google Drive, Notion, Asana, Shopify, Stripe, whatever, whatever. And you can just toggle on whatever the heck you want. And you just look at it and you go, “What are the 10 tools I use in my business every single day?” Toggle, toggle, toggle, toggle. Assuming that you’re okay with these systems reading and taking action.

And you, in the span of 4 and a half minutes, have downloaded the desktop app, switched to the right tab, and enabled the right connectors. And you open up new task top left, you type in a task, and if anything breaks, you just yell at it and you just say, “You like Allie on Cal’s podcast told me I could do this. How come you’re not doing it right?” And it’ll walk you through it. But the likelihood that that first one gets meaningful work off your plate is very high.

CALUM JOHNSON: Yeah. So is the— because you mentioned even just yelling at it— is the functionality of using Claude very similar to then using Claude CoWork?

ALLIE K. MILLER: Is that why you recommend that almost as like the first— Claude CoWork is like a nicer looking version of Claude Code with a few missing features, but not features that I think the average person is going to care about.

Where People Go Wrong with AI Agents

CALUM JOHNSON: Okay, yeah, that makes sense. I’m curious to find out about where is it that people go wrong. And the reason that I say that is because as much as I see testimonials and people saying how it’s changed their business or it’s changed their workflows, I also hear cases of, “It deleted my entire content calendar,” “It sent this email I didn’t want it to send.”

Where is it that— because there’s going to be a bunch of people that listen to this and then get started— where is it that most people are going to go wrong? And how do they protect for that upfront?

ALLIE K. MILLER: So there will continue to be loud stories that make it into your feed that show these things. The first thing to know is that those are increasingly rare. But what I would enable is progressive trust.

Like, I’ve had several assistants in my life, several employees in my life, and I haven’t given any of them email access to my own email in the first day, right? You slowly work with them, build trust, then you give them a corporate card, keep going, then you give them— And so we should apply that same schematic to how we’re working with AI agents.

So you’re not going to— I know I kind of said to toggle all these things on, figure out whether you’re cool with that. But at the very least, you can toggle all these things on and then go into the settings and only give it read access. So if you toggle on Gmail and you just say, “I only want you to read my Gmail,” the worst that could happen is that it summarizes your email and gives it back to you. And maybe you’re screen sharing in some meeting and it’ll show something.

If you put edit or delete permissions, that’s where you’re going to have a potential for something that deletes a file or edits a file. And it’s the same across all tools. So just start with read-only access and then earn trust, figure out what is working, what is not, and go from there.

I think the other thing that people are going to run into is that just sometimes these things don’t work. Like all of these things are in beta preview/research alpha. Like they all have a million words after it that basically says, “If it screws up, we’re sorry.” And I want people to know that things break all the time and that the super users are also dealing with this. And you just kind of have to say, “Okay, it’s broken now. I’ll come back to it in an hour and see what’s working.” Like you still have to be able to do your job even without access to these systems.

The Future of Work and AI’s Impact on Jobs

CALUM JOHNSON: Yeah, about 6 or 7 weeks ago I read this article on Twitter, and I’m sure that you’ve seen it because a bunch of people saw it. 86 million people saw it, which is something big is happening. And it was written by this guy Matt Schumer, who’s worked in AI and almost was like sounding the alarm on what is coming.

And so I wanted to read out something that he said in that article. He says, “I’ve seen AI go from helpful tool to it does my job better than I do.” And everyone else— he works in, he’s like an engineer.

ALLIE K. MILLER: I’ve known him for 6 years.

CALUM JOHNSON: Yeah.

ALLIE K. MILLER: Yeah.

CALUM JOHNSON: But he says, everyone else is about to have this experience. And he lists out: law, finance, medicine, accounting, consulting, writing, design, analysis, customer service— not in 10 years. The people building these systems say 1 to 5 years, some say even less. And given what I’ve seen in the last couple of months, I think less is more likely.

You mentioned that you’re friends with Matt. Can you give people the context of what makes him say this? What is it that he’s seen to make him say this? And then ultimately, where do you land in terms of— we just shared all of these capabilities of like a Claude CoWork can do things on our behalf. What does that mean for our place?

ALLIE K. MILLER: In the last couple of months, software engineer hiring has actually started to come up. So whether or not he’s right, I think that the economic impact of AI on any job, especially code-related or tech-related, is going to be felt by everyone.

Like, I do not believe when people are saying, “AI won’t replace you, someone using AI will.” There’s no way. There are some roles that will be replaced by AI. And by running away from that sentence, you are putting yourself in a much worse position because you’re basically putting on horse blinders saying, “No, that’s not the truth. It’s not the truth. It’s not the truth.” And you’re going to prepare for your career futures in a different and worse way.

Whether it happens in 1 to 5 years, up for debate even among AI experts. What I believe will happen is that the next 20 to 30 years are gray area muck. Like mass industrial and job shifts are not done in 2 years. They might start and be triggered in 1 to 5 years. The time it takes to figure out how systems adapt, how people adapt, how our tax system adapts, whether we work 9 to 5, 5 days a week anymore— that is not done in 1 to 5 years. That is over several decades.

So predictions are one thing, overall impact is another. I think if you are Gen X, you’re likely to retire before we get perfect answers to things. And if you are Gen Z or millennial, you are going to probably see the other side of what happens when we have an entire digital workforce working.

CALUM JOHNSON: Yeah. Just when you share that statistic that these best-in-class models, 80% of the time they were found to actually perform better than someone that was doing the role.

ALLIE K. MILLER: And that’s on overall tasks. In software engineering, we literally just went from an 80% benchmark to 94%. These models aren’t yet being released to the public, but 94%— I mean, that is better than the average engineer.

What to Do Today to Thrive in the AI Era

CALUM JOHNSON: Yeah, it’s stunning to hear. And a lot of people, I think, process that in different ways. There’s a lot of anxiety. And there’s a lot of people that just— it’s almost like, “Okay, this is inevitable, I’m going to just put my head in the sand.” And you even alluded to that.

If that isn’t the right way to go about it— for the individual today thinking about what they can do— I’m even thinking about the person that 5 years, 10 years from now actually looks back at this time and thinks to themselves, “That was almost like a golden era. I made a ton of things happen in that space of time. I started this company. I got more time with my family. I got more agency”— to what we were talking about earlier.

That human being, what would they be doing today so that 10 years from now they have that feeling, that they’re looking back with that mindset and perspective? What are they doing today?

ALLIE K. MILLER: First, I hope that everyone listening goes into that group. Like, even just hearing that description, I would love if everyone had that transformation.

Here’s my current belief: exponential change is nearly impossible for us to wrap our heads around. When we were looking at COVID, in late February, there were people that thought that COVID wasn’t going to impact the U.S. in a big way. There were people by March that also were like, “Oh, this is a 2-week thing. We’ll be back in no time.” Right? I was talking to ML engineers that were running simulations in March saying there’s no way this is done before October. And so because I had that extra information, I could make way better decisions for myself, for my team, for a multibillion-dollar business line that I was running.

So the first thing is, maybe rather than wrapping your head around exponential change— because I truly think that that is not possible for our heads— it is almost better to assume the exponential change and then ask yourself, “How would I adapt if it is true? And how would I adapt if it’s not true?” And then look at the list of actions in both and say, “What is true in both? What is the least risky action that I can take across both?” Because then you’ve basically said, regardless of what happens, I’m in a much better place. So that’s theory one: diving in gets you information asymmetry and leverage.

Number two is you actually don’t have to calculate exponential change. You can just assume it’ll happen and assume it won’t and figure out your action.

The third— and this is again not what people want to hear— is that I think there are just some people that take action and there are just some people that don’t. In 2020, it was very obvious that TikTok was blowing up, and yet not everyone started a TikTok and grew to 17 million followers. There have been other golden eras, and to many it has been obvious. And those people still didn’t carve out time.

I think reducing the action potential, reducing the amount of energy it takes to do one little thing, is the best thing that someone can do. Because once they get that first action going, it is so much easier to do the next 50. The second you learn Claude CoWork and do that one task, it’s so much easier to do the next 500. The second you take one vacation, it’s easier to book more because you’re going, “Oh, it wasn’t as hard as I thought. Now I know the system to actually apply for that time. Now I know that my boss won’t yell at me.” All these walls come down.

And so find the smallest energy to take. For some, it might be signing up for an account. For some, it might be making that document that’s a little bit about you and asking for that first task. For some people, it might be spinning up your first AI agent. For some people, it might be getting a group of 6 people together and saying, “Every Thursday night, we’re talking about AI agents. This is the new book club.” Come together. We’re all going to start finding that little action potential.

And surrounding yourself with people who take action is going to set you on this flywheel that puts you in a completely different career position, financial position, and probably psychological position in a year.

Even with that, I want to empathize with anyone who’s feeling these feelings and saying: all of us feel it every single day. The only difference between someone who is in the superuser category is that they’ve just decided, even with that anxiety, to take action. That is literally it.

The Human Side of AI: Hope, Agency, and Taking Control

CALUM JOHNSON: I think it’s such a powerful perspective shift. There was actually a tweet that I read on your account. You were talking about, I think you’d been in an AI agent workshop and one of the small business owners in that workshop had shared something with you. And I just wanted to read it out because I thought it was powerful and it was personal.

He said, “I’m 53 and I work 13 hours a day, no holidays, no vacations. In 10 years I’m getting older. AI has already cut my judicial briefing time from 45 minutes to 5 minutes. But what I really want is for it to give me my life back. I want to read again, be with my family again, travel slowly. I want to know what that feels like because I’ve forgotten.”

So I wanted to — I think so often we hear about the doomsday scenario, like this apocalyptic version of reality. And I think the thing that stood out to me when I read that, I was like, it was a very tangible example, a human example and personal example of how someone could actually imagine AI transforming their life in a positive way. And so I’m curious to hear, even I want to hear more about AI agents and the workshops that you do, but even when you hear a testimony or someone share their experience, and also just in the short time that I’ve known you, how does that hit? How does that make you feel?

ALLIE K. MILLER: Can I ask you a question? Which is, so that’s from an Anthropic research report where they surveyed 81,000 people, and that quote smacked me upside the head. I mean, I stood there in front — they have a poster in their office, and I just stood at it staring. Do you read that as hope or desperation?

CALUM JOHNSON: Hmm. I read it as hope.

ALLIE K. MILLER: Interesting. I read it as the other. I think, well, first I default to just assuming that things are bad. I just want to fix everything. And I have known you to be a very optimistic person.

CALUM JOHNSON: I’m more an optimist. Yeah.

ALLIE K. MILLER: I think I read that and I read it as someone who is very nervous, anxious, scared that making these shifts in their life and adopting AI agents and figuring out how to do this hopefully will help them get back time, but that they have not yet made that decision for themselves. That’s what I hear. I hear, could, you know, throw it up in the air. Why not let the universe decide? I’m 53, someone else can figure out my future for me.

I read that and I worry for the author behind it, right? It’s a 53-year-old lawyer, I think, in the US that said that. What I have learned from seeing all these people use AI in their lives is if you haven’t figured out ahead of time what you are trying to do and what you plan on doing with the 5 or 50 hours that AI saves you, you will sound like the 53-year-old lawyer that is leaving it up to the universe saying, “I hope I get to vacation with my family.”

That should be your literal goal. Book the vacation now. Book the vacation now and work backwards and say, “Now the problem that I have to solve is that I got to get 3 hours off my plate every single week.” And then you can figure out what briefing you want to use and you can figure out how you want to use AI for due diligence or research or whatever.

I read that and I hear low agency. That’s what I hear. And I think hopefulness comes from having a goal, having a plan, and realizing that you have agency toward that plan. And in my mind, building up that goal in context, writing that down, and leaning on not just your friends and your family and resources that you already have, right, but adding in AI as an additional resource. We’ve never had such high agency as we do now.

So I ache with a lot of people that are in this situation where they’re saying, “Hey, I’m saving all this time, but where is it going?” And I want to tell these people, it’s not easy at all, but you actually need to decide that ahead of time. You really, really do.

The universe is going to keep filling your cup. The universe is going to keep asking for things from you. Your bosses, your boss’s boss, whatever. There will never be enough time to complete things. You will always have tasks that do not get done, right?

And so your first — it’s like a Captain America shield. Your first blocker actually has to be you and not having these crazy strict boundaries or whatever. You got to be a little flexible, but your first Captain America shield has to be, “No, I know what I want. I know why I joined here. Maybe I’ll work till 10 o’clock one night, but then I’m not going to do it 5 more nights in a row.”

And if you have no shield and you let that be built of vapor and not of steel or whatever, Captain America, magic, whatever he builds it with, then you will be in such a bad position with AI because it will take work off your plate. And then the flood is going to continue to fill in with other things that maybe you don’t even want to do.

CALUM JOHNSON: I agree with you in the sense that I think it’s important to make a decision, right? You almost have to make the decision first and you have to have that clarity first, and then everything follows from that decision.

I think the reason I said that I hear hope is, one, I am optimistic, and then I think the second thing is, in my experience, what I found is that we fall into this state of dread as humans when we can’t envision a better future. And every day looks like the previous day, and it’s like this Groundhog Effect to it. And so just the fact that when I read it, he could envision a better future — actually having time with family, right? Or not working all the time, getting vacation days. I was like, that feels like step one to me. You have to envision it first, and then I think from there there’s the decision that follows.

ALLIE K. MILLER: I’m with you.

OpenClaw: The AI Agent Taking the World by Storm

CALUM JOHNSON: Where I wanted to go before we get out of here is last week I had a conversation on this show and it was my first time really going deep on OpenClaw. And I know for you, about a month ago you actually went to a conference in New York, a sold-out conference in New York specifically for OpenClaw.

And just to give people the context, and you can even do a better job of this, this technology, this tool has kind of taken over the world in many places. And the reason I say the world is it’s getting mass adopted, it’s getting adopted in the US. It’s actually getting adopted even faster in China. And so right in the beginning —

ALLIE K. MILLER: Did you see the photo, by the way, of people lining up in China to be able to set up their OpenClaw?

CALUM JOHNSON: No, I didn’t see the photo. I couldn’t believe it.

ALLIE K. MILLER: And it’s like people in their 60s and 70s lining up in line and the younger generation is literally helping them get onboarded onto OpenClaw.

CALUM JOHNSON: Yeah.

ALLIE K. MILLER: Yeah.

CALUM JOHNSON: And you know what? Even as we talk about this, I’m so aware that for some people listening at home, this is like their first time that they’ll even be hearing that term. They’ll be like, “Is this similar to Claude or what? OpenClaw? What’s the thing?” Can you give the context? And even earlier on, I spoke about this — two groups. You have these power users, and then you have the everyday person, and it constantly feels like we’re behind, like we’re running one step behind.

Can you give the context? Why are people so excited about OpenClaw, number one? And then number two, when you step into that conference and you spend time there, what is it that you saw?

ALLIE K. MILLER: So this was — these meetups are happening everywhere. I kind of last minute heard about the New York one, went to it. It was hundreds of people with thousands on the waitlist. Everyone’s wearing lobster headbands and lobster stickers going everywhere. It is a movement. And this tool came out November 2025. This was very, very, very quickly adopted.

At its core, actually at the core of what all these systems are trying to be is a little bit like Her, which is just proactive, something that you can have a conversation with, something that feels natural, something where you don’t have to code to talk to it. Something that is keeping its eye out on your behalf, right? Like if it notices a scam, it’s going to delete that for you. If it notices that your calendar is really busy, maybe it’ll pre-book you a vacation to Portland a month later.

So we all want this proactive helper that gets us and is holding us. I think a lot of people felt like OpenClaw was the first thing that did that. And it’s still the most autonomous AI buddy out there because it was started by one guy. Peter Steinberger built this because he wanted it. And Claude Code and Codex are nowhere near that level of productivity and memory maintenance.

And so this idea that you could first access it from your phone, so you could be walking your dog and say, “Hey, can you email Calum and just ask him what he wanted to talk about on this podcast?” And it would draft the email based on your voice. It knows your email. It has access to it and it can send it. And then you can say, “And tell me when he writes back,” right?

That idea that an AI is checking on what tasks you’re asking for every single 15 minutes and just waiting to help you is a really warm feeling. I think every single person wishes that they had just a 24/7 helper. And so OpenClaw was kind of the first one to do that.

What did I actually see at that meetup? So first, tons of people went on stage to share what they were building with OpenClaw. One guy who works in finance built out an automatic trading system and hearing all of the decisions that he made — it’s one thing to say, “Oh, this thing trades for me.” That’s a vague sentence. Doesn’t help anyone. It’s the fact that he was like, “And I learned to not trade in the first 15 minutes of the market. And I learned to only have 2 windows of trading. So it can’t just trade throughout the day, right, to reduce the actual risk or to reduce the volatility in the opening 15 minutes. And I only gave it $100 and not $12,000.”

So it was really interesting hearing the tradeoffs that people are making with these systems. There’s another company that has 25 OpenClaws sitting in their Slack and they literally spun up a new Slack channel where only the OpenClaws hang out and they’re all like, “Hey boss, nice to meet you. What’s up, Aaron? Hey.” And they’re all teaching each other skills and passing things along. That was a really interesting thing to see.

I think agents talking to other agents is going to be massive this year and next. And I think just seeing the amount of extreme energy in the room for what I would call quite risky AI systems was strange, right? I talked to someone who is deeply involved in the OpenClaw space and I was like, “Well, what do you think are the top safety tips for working with OpenClaw?” And he said, “If you are not okay with all of your files being exposed on the internet, you shouldn’t use it.”

And that is someone who has 20 years of software engineering experience, who has been working on this for months and months and months, who’s an OpenClaw expert saying that. So it is very high risk. That’s why it’s still not adopted by hundreds of millions of users, but it is right now the ideal that all these other companies are racing toward.

Peter was hired into OpenAI because of that project that he built out and because of the unbelievable traction it got, the community that it got. Lobster headbands aside, it was really cool to be part of.

The One-Person Billion-Dollar Company: How Close Are We?

CALUM JOHNSON: You mentioned that it’s the reality that a lot of these companies like Anthropic or OpenAI are running towards. And I remember about a year ago was the first time I heard this possibility, someone talking about the future, that you could have a billion-dollar company which has one employee. And at first when I heard it, I was like, this seems completely implausible. But then the more and more that the developments get made, and the speed of development, and when you even just hear about OpenClaw and what some of the use cases you just described, it feels very feasible that you could have one person with a bunch of agents just running these workflows.

I guess in your mind, do you think that we ever get there? And how close are we to getting there? In the immediate future, whatever the way that you define that — I guess the next year, 2 years.

The AI-First Entrepreneur & Business Opportunities

ALLIE K. MILLER: My immediate future is like the next 5 minutes. Like, is it happening? Is it happening in the car ride home? I think that every single one of these systems is trying to be proactive and helpful. I think the difference between whether or not that dream comes true right now does have to do with cost. So when I was at that OpenAI meetup, there are people that are spending over $100,000 a year running these systems.

So, even if I do say yes to a billion-dollar company run by one person, it might be that the cost to pull that off is not yet open to every single person. I feel pretty confident, and we’re already kind of seeing some signals of it, but I feel pretty confident that we’re going to see companies with just a few people that are already worth a billion plus. And again, we’re seeing signs of it. That are kind of middlemen sort of companies.

But anyways, I think net new companies starting from scratch, 2 to 5 people, a billion plus will happen this year. I have a hard time imagining a one-person billion-dollar from scratch happening this year. Happy to be proven wrong, but if you said you had to bet on it happening in the next 3 years, I would definitely take that bet and say yes on it.

There are people who already have teams of 5, teams of 25 AI agents, like full personas working on their behalf. And if you’re OK with a much higher risk profile and spending potentially $100,000 to $250,000 — which there’s ways to fund that from venture capital — that is absolutely possible.

I also, I don’t know if you saw the research on it, but Citadel and Citrini, there were these two research reports that came out around the same time. One that was like, “AI is going to be the most incredible thing ever.” And one of them was like, “We’re screwed.” And they both came out within like a week of each other.

And the one stat that I loved — that every single person who feels left out of their company, or they feel like their company is slow, or they don’t know what they can do in the space — this graph showed the number of businesses that exist today and are likely to be forecasted out in our AI future. And it basically showed that the number of businesses is going to skyrocket.

I think we’re moving into the age of the AI-first entrepreneur because what has held people back — again, not at the billion-dollar level, but at a meaningful lifestyle level — what has held people back doesn’t exist anymore. Like, the ability to spin up your website used to take $20,000, $10,000, $50,000 depending on how good you wanted it to be, and it now costs a $20 to $40 subscription and like a day.

I also want to just say that the best leverage points are when things don’t take 5 minutes for everyone but take like 8 hours for everyone. Like that’s actually the secret to all this stuff. When you find that — like asking ChatGPT a question and getting it back — now everyone’s got that. So if you made a platform that allowed you to ask a question and get something back, it’s not going to be all that helpful. You have to find things that the average person is not going to dedicate hours to.

So maybe your job is just to learn Claude or something deeply. And then you go to your small medium businesses nearby and you teach them. For thousands of dollars. That is a real business. There are ways to make six figures right now just learning these things really deeply and teaching others about it. But that’s because it’s something that might take a few hours and the average person is not going to pick it up.

So finding those points of leverage, whether it’s installing OpenAI, whether it’s figuring out Claude, whether it’s just building one awesome agent that you can then deploy for whatever, moms in your neighborhood. Like that is a massive point for leverage that can lead to a side hustle or a full-time job.

The Grant Cardone AI Consultant Clip

CALUM JOHNSON: So you mentioned the AI-first entrepreneur, and I don’t know if you have you seen this clip that’s kind of going around at the moment with Grant Cardone?

ALLIE K. MILLER: I don’t think so.

CALUM JOHNSON: Grant Cardone, kind of controversial figure, rightly so. But there’s this clip that’s kind of going viral on social media where he was talking about the AI consultant. And how he sees this opportunity that exists right now. I think the question that was asked was like, if he had to make his first million today, how would he go about it? And he said he’d be an AI consultant, which is basically help like small, medium—

ALLIE K. MILLER: I’d answer the exact same way.

CALUM JOHNSON: Yeah. And he basically said that he would charge companies $8,000 a month. He’d get like 10 clients or so. And he would help them with like AI implementation in their businesses, finding a few workflows that they could implement. And so when I heard it, I was curious to get your perspective on it, given what you do. Also given the fact that you’ve been in this industry and learning about it and using it for so long.

Is this one of those things when we hear “AI automation expert” or “AI consultant” — in your mind when you hear it, is that like a pipe dream for the majority of people, or is that something that you actually think, “Oh yeah, we’re going to see a ton of those and people building, making their first million doing that?”

Can You Really Make Money as an AI Consultant?

ALLIE K. MILLER: I have faith in an open market. Like, if you go out and you’re like, “I’m an AI automation expert,” and you have zero credibility, zero credentials, zero proof points, zero social proof, testimonials, whatever, no one that’s worked with you before — the likelihood that you get customers, unless there’s extreme desperation in your area, is very low.

So I think there’s always going to be a market feedback that people that you’re pitching to are kind of going to come back to you and be like, “Ah yes, you do have these skills, so let’s come in.” I think if you spent many, many hours, you would learn enough to be dangerous, but I don’t think that would get you to a point to be able to build an entire consulting practice around it. I think it’s meaningful hours and staying up to date on it.

I don’t know if $8,000 a month is the right price point, but probably in the $5,000 to $10,000 a month range. And maybe you bring on one other person who can do things in a different time zone, like a West Coast, East Coast person.

I think, again, going back to where that leverage point is, it’s never going to be on the thing that takes 5 minutes to figure out, which is downloading ChatGPT and using it. And it’s also, if you’re a one-person shop, not going to be on something that’s going to take 6 to 12 months. You have to find something that you can figure out in, let’s say, 8 to 25 hours and can teach a bunch of people that thing.

And figuring out what your core modules are — whether it’s “I teach Claude and I can teach it for everyone at your company as a workshop,” or “I build workflows for marketing agencies and here are the 3 playbooks that I run,” or “I’m going to get your finance team or this accounting team spun up in 3 sessions over a week.” Like, you have to figure out what your modular components are, cadence, and how much to charge. But I think you could make significant six figures off of that if you have connections already in the space.

The one thing that I’ll hedge against Grant’s part is that what you deliver has to keep changing. I started in AI almost 20 years ago, have worked in it every single day for the last decade. The things that I was advising on 10 years ago look nothing like what I’m advising on today. So 10 years ago I was explaining computer vision models. I was explaining conversational AI. I was explaining this idea that if you ask it a question it can respond, and it’s not just like copying and pasting from a document that’s sitting off the side. That level of education isn’t required now. And so if I was still teaching that, I would be out of date, out of business.

So yeah, you can be an AI consultant. There is so much to go around. Oh my God, please take some of these clients off my plate, please. And small medium businesses in particular are underserved. But you are also saying that you are agreeing to continued education for yourself so that your customers can continue to be helped.

Like, you don’t want to squeeze them dry of this money. I feel extremely proud when I see my clients doing amazing work, and it makes me want to work harder for them, and it makes me want to stay up late to figure out what’s the next thing that I could advise them on. What’s something coming in 2 years that I should be talking about with them now? What is the new potential growth line that they could launch as a brand new business that if they did it now, they’d actually be ahead of every other industry? Like, I am brainstorming on their behalf. AI consultants — I just — Grant made it sound too easy.

Behind the Scenes: Allie’s Day-to-Day Life in AI

CALUM JOHNSON: I’m so happy that you said it — I think sometimes when these things come up, like these opportunities, these trendy opportunities that come up on the internet, the feeling is, “Oh, it’s so easy to do this right now.” And you mention how much this space is changing, even for you. Like, what you’re advising on now is completely different. It’s a 180 from what you were doing previously.

Can you take people behind the scenes for a second of like Allie’s life? Because I see you posting so often on your social media. You’re obviously — even the energy that you’ve spoken with in this conversation, you’re so into this. What is happening behind the scenes? What is your day-to-day like, so that you can even have this perspective, so that you can go to a client and actually advise them? Can you take people behind that curtain for a second?

ALLIE K. MILLER: Right before I start my day, AI has already been working for me for hours. So I have that morning debrief that I wake up to and it runs through my day, or it gives me all of the assets that I have to figure out that day. At the very end of the day, I’ll queue up a to-do list and kick it off to two different AI systems that can help me get some stuff done.

Sleeping is the thing that I’m struggling with the most. I think last night I went to bed at 4:00, so improving every day. But I was also up reading a 200-page system card on a brand new model, and that’s why it spiked.

I think staying up to date is definitely helpful. I think knowing your client’s work areas is very helpful. So I have some clients that are in the finance space, some that are in sports, some that are in retail, some that are in tech, some that are in pharma. So getting a taste of what’s happening in those industries is very helpful. But yeah, the majority of my day is actually on the advising side. I get to see the future every single day.

Opportunities Hiding in Plain Sight

CALUM JOHNSON: You say you get to see the future every single day and you have such a wide purview — between the advising, and you even mentioned that small and medium-sized businesses in particular are underserved. But you also work with a range of clients, like big companies, and then even influencers in the space you’re talking to. You mentioned there’s a tool that—

ALLIE K. MILLER: That’s because I’m like lonely and I need friends.

CALUM JOHNSON: You mentioned that there’s a tool that you’ve been using that’s kind of exploded your mind. If you had to almost share with people an opportunity that from your perspective is kind of hiding in plain sight — just because you’re having so many conversations with different people — like a very specific actionable opportunity that in your mind, if someone approached it with the right application and mindset and hunger and energy, it could be like a huge opportunity for them in their life.

I’m just curious, based on all these conversations and data and the reports that you’re reading and the people that you’re speaking to, like, what comes to mind? Like, what’s the opportunity?

Building a Business with AI: Opportunity, Effort, and the Future

ALLIE K. MILLER: My, again, extremely simplified answer: if you are looking for an opportunity to leave your current job, or maybe you’re unemployed and you’re just trying to figure out how to make sense of the time that you have, the way that I would spend my time right now is going deep into Claude Co-Work, learning every single setting, every single way of managing memory, of managing and building skills of your Claude.md file, spinning up different projects, like really go deep into that, which is going to take you hours. Test it for yourself. Get really, really good at that. Teach 3 people in your life that know absolutely nothing about what you’re doing. Start to get feedback on your teaching skill level. Get feedback on whether you make sense, right? And do a little bit of trialing on that.

And then go to your friends’ parents or your friends that are running small businesses that are somewhere between 50 and 500 people and say, “I can get every single person at your company trained up on this with two 90-minute workshops. I charge this much for the first set of 2 workshops and then this much for monthly continuance, which allows you 2 flex hours and 1 additional workshop over the next 3 months or whatever you want to structure it as. And for that package, it will be $50K for the next 6 months.” That easy box and just being able to serve that to maybe 100 potential clients and maybe 5 bite. That’s already you making 6 figures a year.

And the simplified version of it sounds easy. I just want people to know that it will take work. And that’s actually the thrilling part. When something takes work, that means not everyone else is going to do it. We can’t be afraid of effort. We can’t be afraid of work. You should be looking at work and effort and going, “Ah, that’s leverage.” And again, you could do this on Claude Code or Codex or whatever, but small, medium businesses will benefit the most from Copilot Co-Work or Claude Co-Work or these more non-engineering, more visual interfaces. And that’s where I would double down if I was trying to scale on SMBs.

CALUM JOHNSON: Yeah. And I hope that someone — we’re going to have someone that takes that and actually goes and does it. And I hope that they reach out to either one of us after hearing this conversation, because it would just be cool to see that impact.

ALLIE K. MILLER: Yeah.

Allie’s Vision for the Future of AI and Human Potential

CALUM JOHNSON: But if you had to project 5, 10 years into the future, in Allie’s perfect world of how things end up — because I think people have so much anxiety and they go, we hear an optimistic take and then we hear a ton of pessimism and doomsday, and probably it will land somewhere in the middle. But in your ideal reality, what is your hope for what people’s lives will look like in 5 years from now?

ALLIE K. MILLER: First, thank you for just noticing things in people, in me, in all of your other guests. I think I’m a very intense, impatient, and justice-fueled person and entrepreneurship-fueled person. And by the way, maybe that’s not the vibe that people even want to be listening to. And so first, I just want to encourage everyone to gain more hope, to gain more agency — listen to tons of people in the AI space.

My hope is that in 5 years, the average non-engineer understands how to use these tools and very specifically understands not just how to do current things faster, but understands how to do things better or do new things. My biggest fear would be that all of humanity is granted this unbelievable elf in a box that can do really, really complex PhD work, and that they go, “Great, I will write podcast scripts faster.” Chuck it out the window.

My hope is that people are given this system, this operating system, and then they go, “Oh, the world hasn’t been written for me for the last several decades,” or “This process has never made sense,” or “Everything that we do in order to get an invoice signed is so manual,” or “The way that we reach out to potential new clients, even if it’s for this new AI consulting business that everyone’s going to build, is just so slow. It’s so manual.”

My hope is that people go, “If I was an alien born right now and I had access to this tool and I didn’t have anything bogging me down from how the system used to work, how might I reinvent it from scratch? How might I reframe how I look at this system — not tool, but full operating system — and how can I reinvent that?” And that takes years. So in 5 years, my hope would be that all bad processes that ruin our mind and happiness are gone and dead, and we have all these incredible things that make us feel good every day and that we love our role in our work. But even if that happened in 30% of what we do, I think we’d be markedly happier.

That’s one. The second — I really do believe that the number of businesses is going to go up a lot. And I want to see more entrepreneurs in the world. I want everyone to have more agency in their lives. I want people to build the consulting company they want, the coaching company they want, the side hustle they want, the software tool that maybe only 1,000 people in the world need but would actually improve those lives. I want all of that to be happening in the next several months, but I’ll deal with several years.

And then I think if I really step back, in 5 years you’re going to have — the likelihood is high, so this is a gamble — but the ability to have an AI researcher that can complete scientific research on our behalf feels very, very high. Did you read the story about the guy with the dog with cancer?

CALUM JOHNSON: Yeah.

ALLIE K. MILLER: That’s —

CALUM JOHNSON: Oh, share that context with people.

The Dog Cancer Story: A Glimpse of AI’s Medical Potential

ALLIE K. MILLER: Yeah. So there was a gentleman in Australia that had adopted a dog and then the dog gets diagnosed with a form of cancer and couldn’t walk, couldn’t jump. And this guy gets his dog’s DNA fully sequenced, puts it into a combination of ChatGPT and AlphaFold, which is from Google, and basically tries to figure out what vaccine the dog could take to improve and gets an answer — probably also based on previous research — and goes to UNSW and says, “I want to save my dog, can you help me make this?”

And so a research team at UNSW said yes and took it on and made the vaccine, and then they gave it to the dog all in the span of 2 months. Tumors shrinking, dog jumping around everywhere. The idea that that could happen — and it cost a couple thousand dollars for that to happen and took 2 months — but I’m imagining that that could happen at scale for many more people, and not just for our wonderful canine friends in Australia. That is very possible on a research side in the next 2 to 3 years. And at scale in 5.

Closing Thoughts and Where to Find Allie

CALUM JOHNSON: Yeah. Allie, first of all, just thank you for coming on the show. And even more than that, just your intention, the energy that you bring — yeah, it’s been a pleasure. So thank you.

ALLIE K. MILLER: Thank you, and thank you for having me.

CALUM JOHNSON: Of course. Appreciate it. And before we get out of here, can you tell people where they can find you?

ALLIE K. MILLER: Yeah, I’m @alliekmiller on just about everything — Instagram, LinkedIn, TikTok, and X. A-L-L-I-E-K Miller. And also, funny enough, AI, no matter how many times you say Allie Miller, it’ll always spell it as A-L-I.

CALUM JOHNSON: Allie K. Miller. Thank you so much.

ALLIE K. MILLER: Thank you.

CALUM JOHNSON: That was awesome. So if you enjoyed this conversation and you want to hear even more stories like this, then just click here. And also, my team is going to put some more videos that you can watch here. Thank you.

Related Posts

• Diary Of A CEO: w/ Jeremy Grantham (Transcript)
• Transcript of Andrew Ross Sorkin’s Interview: Piers Morgan Uncensored  
• Fed Chair Kevin Warsh on Interest Rates Policy Decision – June 17, 2026 (Transcript)
• Founder Mindset: w/ Reese Witherspoon on Hello Sunshine (Transcript)
• Death of the Middle Class Debate: Daniel Priestley v Nick Hanauer (Transcript)