AI, Actually – Episode 18: How to 10x Your Individual Productivity with AI

Michelle Hamilton (01:26)

Thanks, Pete. You know what? I tell you what, I am only just a couple of weeks into AnswerRocket But ironically, this is my division that I’m super excited about change management and AI implementation and truly helping teams to understand how to have a people first approach to AI and understand how to apply an AI lens and strategy to the work.

that they’re already doing. And so, you you talked about unused licenses and fear, and that’s where I’ve spent my last three years focused on helping organizations think about that. Answer Rocket is an astonishingly forward thinking group as they should be considering we’ve been in the AI space since 2013.

I love the fact that we are not only AI first, but also incredibly open to conversation, trying new tools, but really understanding that you have to walk before you run, understand how to use the products first in a simple everyday use case that then you can begin to apply to your own roles and KPIs.

Nicole (02:21)

Yeah.

Pete Reilly (02:35)

What are some of the tools you’ve leaned into, Michelle?

Michelle Hamilton (02:38)

I will tell you, I’m a big Claude fan. I have been using Claude for almost not exclusively because I am agnostic, but I do use it for most of my advanced work. But I will tell you, I actually do have some very strong use cases in both Gemini and ChatGPT. They all have their own features and benefits.

Oftentimes, organizations and individuals think of ChatGPT as being the Kleenex of AI. So when I’m doing coaching and consulting, we generally actually teach to ChatGPT because it’s the one most likely people already have downloaded to their phones or have on their computers. And it is also the one that most organizations should be the most concerned about, not because of the company, but because of shadow AI use.

Pete Reilly (03:08)

Yeah.

Michelle Hamilton (03:32)

And if you are telling your employees that they can’t use AI, I promise you they have chat GPT on their phone.

Pete Reilly (03:38)

Yeah, yeah, they probably have the settings wrong. So everything

they do is getting his training the model, right? We see that we see that all.

Michelle Hamilton (03:43)

Yes, everything is

creating the model. So you actually need to have a progressive attitude about a people first approach.

Pete Reilly (03:50)

Yeah,

yeah, for sure. ahead. What were you gonna say?

Mike Finley (03:53)

was going

to say, I really love that quote of people first approach to AI. Right. And I think that that’s where a lot of times we get lost in the fact that the people are the uniqueness about the AI, right? AI as you buy a chat GPT off the shelf is the same for every company everywhere. It’s your people that are going to make it better, right? They’re going to make it different and unique and differentiated. And the fact is we people learn differently than the mechanical brains, right?

Michelle Hamilton (04:00)

Yeah

Pete Reilly (04:10)

Mm-hmm.

Mike Finley (04:20)

the wetware versus the hardware learns differently, does things differently, and we have the opportunity to benefit each other. We can sort of level up each other. And we see that in some of the best adopters of AI is the people that learn to level up with it, not just to put it to use, but to really go to another level of their own performance and capability.

Michelle Hamilton (04:42)

Mike, I love that because I’m finding organizations that are the most successful are the ones that are combining that use case into an intellectual property library of AI use cases and combining the best prompts and outcomes into an IP doc that then continues to educate the entire organization to do better, be better, and get to that run phase that everybody is looking for as quickly as possible.

Mike Finley (04:59)

Yes.

Yeah, that, that corpus you’re describing, that is the new source code, right? The old source code that we worship because people spent so much time so carefully building it, that’s getting made automatically. Now the real source code now is those evergreen examples of how to do your business, right? Where you capture because of these input factors, we did this thing and that’s a unique trade secret. If you have those, if you have a trove of those, your AI is going to help you take over the world.

Pete Reilly (05:21)

Yeah.

Right, because everybody’s got access to the same models, right? It’s how you leverage them and how maybe your specific playbook and your specific strategy and how do you use the model to help you accelerate that ⁓ because everybody’s got access to the same thing now.

Nicole (05:41)

That’s right.

I think of the world now as pre-ChatGPT and post-ChatGPT. And I remember Pete, post-ChatGPT in those first 30, 60 days, you were always saying, how are you using AI today? And my goodness, I was struggling to come up with answers for you. I asked you a question today. I used it like Chrome today. ⁓

Pete Reilly (05:56)

Yeah.

Nicole (06:15)

And now I think fast forward three and a half years later, here we are. There is not a single thing that I do that is not AI first. ⁓ For example, I was putting together, ⁓ I was gathering input from a bunch of team members this week on how do we AI? And I was putting together a presentation. I used AI to do the whole thing. And like Michelle, I’d say I lean more into Claude, whether it’s chat, whether it’s co-work or whether it’s Claude code.

Michelle Hamilton (06:23)

Yes.

Nicole (06:44)

⁓ all the time up in Claude, ⁓ but I still use others other tools. ⁓ Actually, my favorite Gemini example was for a personal project that I did. I actually used Claude code to build a prompt for me that I then fed to Gemini for a landscaping project that I was doing and I had it create a visualization for me because that’s one of the things that does really well. So I just think, you know, when I think about how do you get started with AI?

Pete Reilly (07:05)

Anyway,

Nicole (07:14)

I think there is no wrong answer. But you come into work and if you’re thinking about what is it that I have to do this week that I do all the time that just drives me crazy and pick that thing and spend a few minutes doing it. That’s right.

Michelle Hamilton (07:30)

takes more than 15 minutes. If

it takes more than 15 minutes, you’ve already got your AI lens to start with. And I’m going to take it even one step easier if we’re even talking about people who are still at that point where they’re trying to think about how to adopt it. Don’t worry about work yet.

Nicole (07:37)

That’s right.

Michelle Hamilton (07:49)

Think about just your daily life, your daily use cases, and then how can you apply some similar opportunities, whether you’re planning an amazing vacation for you and your partner to Italy. Well, guess what? That entire conversation can then also be, you know.

added to a workflow for your organization in thinking about planning out a retreat, planning out a strategy. But my favorite use case is absolutely to be using it in transcription mode so that if you need to get out, you need to iterate, you need to think, or you’re just standing in the bathroom in the morning getting ready to go, hit the transcribe mode and just start talking to it. Hey, here’s what’s on my list of things to do today. I’ve got this recurring idea.

or I’m gonna take my dogs for a walk, I wanna talk through this strategy, just got off this great phone call, I wanna follow up with some additional notes, just use it as your thinking partner and personal assistant.

Mike Finley (08:47)

I love these power ups that we’re talking about because, know, it’s one thing to say, yep, I’m going to go use the tool and whatever static output comes from it. I’ll put that to use right one time, you know, help me describe this thing for somebody, help me draw something. Right. There’s another level, Michelle, you just started touching on it where, where you’re beginning to build a portfolio of how you like things done.

And now you’re leveling up your AI use because when you do say, ⁓ you know, Hey, let’s make a trip to Japan. Well, here’s my five other trips that we planned before. So I’m not going to have to tell you what kind of airlines I want to travel, what kind of dates I want to go on, what kind of, ⁓ expense level I want to have, right? It’s going to have that already. Right. And that, that, that begins to form this, this, again, this unique corpus that describes you and what you want differentiates what you’re after. And it really levels up the results that you’re getting. Right. And, that’s.

A lot of people are having that first experience of leveling up their use with some of these vibe coding tools that are out there, right? The lovables and the bolts and, you know, different ones that let you not just use AI to do something, but use AI to make something that does something, right? And that’s a really new experience for lot of people who have not been able to code before and are now putting it to use in this way.

Michelle Hamilton (10:02)

You know, Nicole brought up earlier the idea of she used Claude to then prompt her Gemini model for her landscaping project. That’s called meta prompting. And so if you just go back to your own model and say, hey, I need to figure out how to prompt you through this complex idea. I’m not really sure where to start. Let me tell you everything I know about it. And we were chatting about this before the podcast started. Just tell it who, what, where, when, why, and how.

and the ultimate outcome that you’re looking for. Now help me build out the perfect prompt. I want you to teach me how to use you so that I can teach you what I need for me.

Mike Finley (10:42)

Fantastic. And, know, just a little bit of the insides of what’s happening behind GPT and Claude and all that is they’re roughly a hundred layers that have continuously done that. They start with whatever you ask and they go through many, many layers of what’s called this transformer technology where they’re pushing through. And ultimately what you see is what comes out the other end. And so when you take one model’s output and feed it into another, that meta prompting, you’re stacking one model’s hundred layers.

on the other models hundred layers and making a 200 deep stack, which is now even more powerful, right? And, and this, this gets to some of the scalability laws, right? Like, so why don’t we just go a thousand and see what happens, right? Well, because the expense and the time and the, the, you know, the, energy consumption get too high, but, but that is a constant change as, the hardware improves, as the power solutions get better, as the models get more efficient, we’re getting more and more and more of those layers.

So we’ll have to do less of that meta prompting. But as of right now, there’s no end in sight to how many times you can do that and continue getting improvements.

Pete Reilly (11:49)

I don’t know about you guys, but one of the things I struggled with, I still struggle with this, is ⁓ reminding myself when I’m about to do something that there’s probably a better, if I’m doing it the same way I would have done it even six months ago, that’s probably the wrong way. And figuring out how to sort of interrupt myself to just sort of question that and to try it the new way can be a challenge sometimes. And often I…

I force myself, even if I’m halfway through a task, to stop and go back to the beginning, to do it what I would consider the new way. Because otherwise it’s just hard, it’s really hard, I think, for people to start building and getting into the habit. I find that to be one of the more challenging aspects of getting into this new

Mike Finley (12:39)

Well, I think you’re hitting on something

there, Pete. That’s, that’s a big part of this move, especially for enterprises. We’re not just trying to teach people a new skill. We’re trying to teach people to change, ⁓ and not just to change to the next thing, but change to the next thing and the next thing and the next thing after that. It’s a very meta thing. And that’s why there’s such a divide between people that, that, are struggling and people that are launching, right. And have moved forward.

Pete Reilly (12:56)

Right.

Mike Finley (13:05)

Because really what you’re, it’s like the old, know, college doesn’t teach you anything, but it teaches you how to learn, right? I used to hear that a lot. And, and in this gen AI era, we don’t need to teach people one set of tools. need to show them how to change and how to know when to change, Pete, kind of to your point about progressing.

Pete Reilly (13:20)

I think that’s exactly right. Yeah,

I think that’s exactly right.

Michelle Hamilton (13:22)

I might

even teach them to think, what is my current job and what are my current drivers? Now, how do I create a personal assistant type relationship with an AI model and just simply explain to it, here are my deliverables for the day, for the project. I’m going to ask you to help me with this, not because I’m asking you to replace me, but because I want you to help me both expedite as well as make my process more efficient.

you are my thought partner now. And I highly encourage people to ask the model to ask you as many questions about the process as possible. It takes a little longer that first iteration out, but I will tell you then if you continuously come back to that conversation, if it’s about the same topic, it has now learned that information from you and you’re going to be able to get to the end result much faster.

Nicole (14:18)

Yeah, I want to come back to.

Mike Finley (14:18)

The average person,

this is just an estimate, but on average, you can count on about a 10 IQ point boost over what you would do yourself plus an AI, right? And there’s just more there. You know a lot, the AI knows a little bit more in certain areas, and it’s like collaborating with somebody. The result’s gonna be better. It’s still your work, right? It’s still your work, but it’s gonna be better. And so that’s the key thing to embrace.

Nicole (14:48)

I love that. I think there’s another way we were talking earlier, I’m not sure who brought it up, but about sharing tools across the team. And I think, Mike, if we talk about the 10 IQ point boost from using AI, if we share tools across the team, that’s got to get us another 10 IQ points, I think. And that’s one of the ways that we level up here inside the Answer Rocket is we have a channel, a Slack channel called GenAI NerdOut.

And we all share, did you hear this? Did you hear that? Are you using it for this? Here’s a great plug-in I’ve discovered. And I think the reality is the people you work closest with are those whose workflows maybe are similar or at least adjacent to what you’re doing. So you’re going to learn a lot from what each other is doing. And I think the conversations I have with other team members about what they’re doing with AI today,

are probably the things that lead to the biggest leaps in capabilities for myself. think, Pete, you would love to hear me say ⁓ something about, talk, okay, I’m gonna leap topics a little bit here, but one of the big buzzwords out there today is forward deployed engineers. ⁓ And in my mind, everyone is a forward deployed engineer today. I know Pete knows, okay, once upon a time, I won’t say how many years ago,

I was an engineer, but I was never a software engineer. I have some, you know, enough technology in my background to be dangerous. But I feel like today with the tools that we have at hand, we can all be the forward deployed engineers. And I think that the key to being a good forward deployed engineer is you need to understand the basics of technology. You need to understand things like the SDLC, software development lifecycle.

Pete Reilly (16:20)

Okay.

Nicole (16:41)

and also agile concepts. But I actually think the most important thing for success for a forward deployed engineer is someone who can understand, comes at it from a business first perspective. And I think this is what makes us all so powerful as AI first business users. I can come at it and I can build a prototype. I don’t need to know how to code. I have a development team who’s going to make that production worthy.

But as a forward deployed engineer, I can be doing that POC myself. And that is a capability I would have never claimed in the past.

Pete Reilly (17:22)

What was your aha? How did you discover that? What was your aha moment?

Nicole (17:25)

I don’t, you know,

that’s a great, that’s a great question. I’m not sure there was an aha moment or I was the frog in the boiling pot. You know that it just happened every time. Although maybe my aha moment was on a call with John Workman. So I went ahead and installed VS code and I connected myself to Claude code. And then I sat here with a prompt and I said, I don’t, I don’t know what to do. So maybe it was when I got on a call with John Workman again.

taking advantage of my colleagues who are maybe a step further ahead in the pipeline. And he helped show me how to use a simple tool in Claude Code directly and get started. And then as I downloaded Claude locally and now I have chat, co-work, and Claude Code, and I am just doing everything I possibly can at this point.

Pete Reilly (18:06)

Mm-hmm. Yeah.

Michelle Hamilton (18:19)

I love that fearless approach. I mean, it is. People spend so much time being afraid of either just trying or, you know what Mike was saying earlier about the 10 IQ point boost, I also have so many people saying it’s just making us dumber. My thought, and I’m fortunate enough to have spent

Pete Reilly (18:19)

Maybe there’s a… Yeah.

We’re using it wrong if that’s true, but okay.

Nicole (18:38)

It’s crazy, yeah. That’s right.

Michelle Hamilton (18:45)

a lot of time speaking at universities, training school districts. And I want to tell you, the people that are embracing it and thinking about it have the opportunity to be some of the world’s greatest thinkers moving forward, not because of the AI, but because we’ve always been ⁓ chained to our ability to ask for more in the past and dream bigger and better. And now we have children and young people coming up who will be able to think so much bigger.

than we were ever given permission to think. And they are going to be able to lean into and utilize these tools and incredible engineers like we have at Answer Rocket to be part of that solution. But they are going to be able to say, what if we, do you think I can? And the answer is yes, you just have to dream it. Now let’s find a way to lean into it.

Pete Reilly (19:38)

It’s

⁓ have this one of my, my wife would say it’s probably my favorite movie of all time is The Matrix. And there’s a scene in the movie where Neo gets plugged into The Matrix for the first time and some guy gets on there and basically loads loads a bunch of programs and Neo wakes up and goes, because I know Kung Fu, right? So I think, mean, honestly, this is the closest thing that we’ve ever had to that. I mean, you want if you want to learn something really quickly, it’s

Michelle Hamilton (19:46)

Oh yes

Kung Fu, loads the Kung Fu

Pete Reilly (20:08)

pretty amazing if you want learn about a client and their business or just a new topic or a hobby or what have you. You can just lean into the agent and Michelle, what I’ll do, you talked about sort of the voice thing. I’m driving around instead of just listening to the radio. I’ll say, hey, tell me about this. Push me on it. Make me ⁓ smarter on the way to this meeting. And it’s absolutely incredible. And what I was going to add earlier, ⁓ I’m picking up on it. I think there are three things to maybe

Michelle Hamilton (20:23)

all the time

Pete Reilly (20:37)

there are some themes I’m hearing maybe to call out. One is it’s really helpful if you have some peers that are sort of into this and sort of sharing with you and what’s new and there’s new stuff out every day, which is crazy because otherwise you just don’t know what’s possible. My daughter is getting married in the fall and she had some spreadsheet of people and addresses and another spreadsheet of people and addresses and she had to do something to move one over the other and she goes, I just don’t want to do this. I’m like, well then don’t.

Mike Finley (21:07)

Right?

Michelle Hamilton (21:09)

You don’t

have to.

Pete Reilly (21:11)

And so we just jumped on and we did, there’s a new capability called computer use from Claude and we just told Claude what to do and 45 minutes later it was done. It’s very slow by the way, but it got done and she didn’t have to do anything. So just knowing what’s possible and peer sharing about that is super valuable, number one. Number two,

It’s knowing this stuff, mean, you have to, it’s sort of like trying to learn a bike by just looking at the bike and studying the bike. You just got to get down on the bike. And the funny thing is, is you can sort of ask the bike and you’ve talked about this, Michelle, you can ask the bike how to ride the bike. So if you’re sitting there and you want to build something and code something, you’ve never built anything in your life. You have no idea how to download anything to your computer. You just sit there and ask and say, Hey, I’m trying to get this thing done. And if you are persistent enough.

Mike Finley (21:45)

Yeah.

Pete Reilly (22:07)

and just say dumb it down like I’m five or five years old or whatever, you can get, it’s unbelievable what you can get done. So just being willing to sit there and be persistent and ask how to get the thing done is a a big ⁓ enabler.

Michelle Hamilton (22:22)

I am gonna recommend though, Pete, because Mike was touching on this earlier, the idea of using multiple models and the importance behind that. Another amazing way to, if you don’t have someone to iterate with is, not only now you’ve asked it to give you something that is factual, riding a bike safely, helmet laws, whatever that might be.

Pete Reilly (22:31)

oh sure

Michelle Hamilton (22:44)

You are also going to ask your model, I need you to give me a full citation of where you got that information from hyperlink that full bibliography. And then I want you to copy paste that entire outcome into a second model or into a doc to save context and then tell the other model, prove it wrong. Find all the mistakes. You go check hyperlinks for me. I want you to cross verify each other and ensure that I am only moving my voice and my company’s reputation forward.

Mike Finley (23:03)

Right? Right?

Pete Reilly (23:05)

It’s pretty bad.

Michelle Hamilton (23:13)

on verified facts.

Mike Finley (23:15)

And so this is one of those kind of mind blowing moments, but I want you to think about this. When you’re moving something from one model to another like that, you’re going through a very narrow funnel, which is human language, right? The models themselves, their native language is giant vectors of numbers. It has to get forced down to a little tiny fraction of what the model is thinking to make it so that we can handle it. Now imagine, imagine a world where it never has to come back to human language. You just let model A talk to model B.

directly without that that kind of narrow ⁓ human time signal in between You know Pete’s point about the computer uses super slow These problems are gonna disappear these models are gonna get if you think if you think things are changing fast now Get ready for what’s coming because now the models that are that are based on the world not based on words And that’s a huge difference right three-dimensional movement and so forth based on images based on video models that are doing real-time speech my mom

⁓ as an example, know, she’s has not been able to use AI, but I, but I used a different model. use the real time models from open AI and gave her the ability just to talk to it and have it talk back. and all of sudden she’s asking about sports scores and trying to find out about lottery numbers. And do you remember 1962? And, know, it’s just, it’s blowing her mind and, and, and she doesn’t even really know that she’s using AI, right? She’s just, she’s just really happy to have this new part of her life. Right. So so much more coming.

Pete Reilly (24:37)

Right.

Mike Finley (24:42)

from a, okay.

Michelle Hamilton (24:43)

I’ve got a fun project for your mom. So my

parents are both in their middle 80s and I have taught my mother how to upload her favorite recipes, hit voice mode and say, I need you to walk me through this step by step.

Mike Finley (24:56)

Nice.

Michelle Hamilton (24:57)

Let me know what

ingredient I need to add. I’ll let you know when I’m done. And by the way, you’ve been, you you, are playing the role of my, and her AI model is named after her mother and she literally will cook so she doesn’t have to touch her phone or her iPad. She’ll cook an entire recipe one step at a time, having fun, iterating back and forth, having this conversation and forgetting that she was ever using AI to begin with having a great time and making

Mike Finley (25:07)

Yeah.

Michelle Hamilton (25:27)

and a good recipe.

Nicole (25:28)

⁓ That’s a great example of how to bring AI to people who aren’t used to using it. I want to come back Pete to a point you made about using AI to fix or figure out something that went wrong with some other AI. I think that was you Pete who made that point. I have a perfect example of this where I was so.

I am a PC user. feel like it’s almost a confession here. think I’m all other Mac users. ⁓ And ⁓ Claude desktop when it was initially released, it was initially released on Mac. And I felt so, you know, left behind because I couldn’t use Claude desktop yet. I’m a big I use it on on my browser all the time. I probably have

Mike Finley (25:52)

Ha ha ha!

Michelle Hamilton (25:53)

So.

Nicole (26:13)

10 or 12, well, everyone here knows that I have a browser tab problem. Probably 10 or 12 of those are Claude right now.

Michelle Hamilton (26:20)

you’re in good company.

Nicole (26:22)

Good, good, good. so when I don’t when when they did finally release, ⁓ Claude desktop for PC, it didn’t work automatically. And I, I spent a good amount of time with my, yeah, I spent a good amount of time with my Claude browser, saying here’s the I would just copy and paste. Here’s the error you gave me now. What do I do next? Copy and paste. Here’s the error you gave me now. What do I do next? And I will say the first time through I spent a good amount of time troubleshooting.

Pete Reilly (26:32)

I remember. You were not happy.

Yeah. Right.

Nicole (26:52)

and it said, sorry, we’ve actually reached the end of our rope here. Go ahead and submit a bug ticket. But a week or so later when I came back in, it worked. in the end, it’s still a happy story. it is a great way. You can take that to anything. You can troubleshoot anything you’re doing in any of the ⁓ tools.

Pete Reilly (26:58)

You

Yeah,

I think Michelle was the one that pointed out the models sort of feeding on each other. ⁓ I’ll say my most common input to any model is a screenshot, I Here you go. What do I do now?

Nicole (27:18)

⁓ okay, sorry.

Yeah, yeah. You know,

Mike Finley (27:25)

Yeah.

Michelle Hamilton (27:25)

Amazing, amazing, love that. Show us how to use.

Nicole (27:30)

the other thing I love about these models, there’s no judgment. I can ask Claude the dumbest question in the world and no one else has to know that I had to ask that question.

Michelle Hamilton (27:31)

video.

Pete Reilly (27:38)

It is pretty nice. Yeah.

Mike Finley (27:41)

Yeah.

Pete Reilly (27:42)

Yeah. Yeah.

Michelle Hamilton (27:43)

It

Mike Finley (27:43)

My wife was that my wife was asking me what the difference is. Like why, why are there so many models? And, ⁓ and I, and I of course went into my normal spiel. Well, the model sizes and the parameter counts and the, she’s like, no, no. mean, what do they do? That’s different, right? ⁓ that. So I gave her, I gave her a little tool, ⁓ that I vibe coded a tool.

Michelle Hamilton (27:43)

is a great support.

Nicole (27:50)

Mm.

Pete Reilly (27:59)

Yeah. ⁓

Mike Finley (28:04)

where she can type one prompt and see what Grok does with it, see what Claude does with it and see what GPT does with it. And she was floored. She was absolutely shocked because they are very, very different, right? And they have their own personalities. They have their own political alignments, especially Grok, right? ⁓ They have all sorts of different levels of tech that they lean into. They hallucinate in different ways, right? And so, this is why we play them against each other and with each other.

Pete Reilly (28:04)

Yeah.

that’s cool

Yep.

Yeah. ⁓

Nicole (28:16)

So different.

Pete Reilly (28:21)

Yeah, yeah, yeah.

Mike Finley (28:33)

to make results come together just like any other tool.

Pete Reilly (28:33)

Yeah. People will ask me, oh,

what model should I use? I’m like, I don’t know. What kind of car do you want? What are you trying to do?

Mike Finley (28:39)

Right, right, right, right. Yeah.

Nicole (28:40)

Right.

Michelle Hamilton (28:42)

Pick one,

Nicole (28:42)

Yep.

Michelle Hamilton (28:43)

and you should probably pay for it. Free is fine to start, but it is worth the three Starbucks coffees or $20 a month to get a superior model, superior.

Mike Finley (28:48)

Yeah.

Mm.

Pete Reilly (28:53)

At least one. my point is

they’re a little different. I use Claude very differently than I use the OpenAI model, like Codex, for example. They’re very different. I personally find Claude is very good at what we just call general work office tasks. I built this ⁓ little, basically I built an assistant.

and I had it code up, it downloads my Fireflies transcripts, it downloads my emails, it downloads my ⁓ calendar, and I gave it a whole set of rules about how to process my to-dos and so on. And it’s an amazing office assistant. can leave a meeting, maybe we talked about something, and I signed up to create a presentation, and after the meeting, I can just say, can you just create the presentation we talked about?

Mike Finley (29:45)

I’m sorry.

Pete Reilly (29:45)

It’s gone through the whole meeting, I read the transcripts and it does a fantastic job on creating a, can you create the spreadsheet on the model of the thing we talked about? I think Claude is really, for me, great at those things. But if I want to lead into code and building things and so on, I tend to find personally that the codec seems to do a little bit better at those things for me. the point is that they seem to, then if you want images and so on, it seems like Gemini models are probably better in that area. So they’re definitely…

Michelle Hamilton (30:15)

we haven’t talked about is Perplexity And I will say we were talking earlier about inviting multiple models to have conversations and we are using.

Pete Reilly (30:15)

⁓ I’m sorry.

Yeah.

Michelle Hamilton (30:25)

⁓ perplexity quite a bit for some large projects in our house. ⁓ Where we are asking it to go out and call upon multiple models simultaneously. I think as well though, we’re touching on some more advanced things now, but for some of our everyday listeners, you need to keep in mind that it is generative. It is why it’s called generative AI. So if you are sitting next to someone, your outputs and answers are not going to be the same.

It is based on your previous use. It is based on a whole host of inputs and the language that you used. But don’t think you’re doing it wrong if your answer doesn’t look exactly like the person sitting next to you’s answer.

Mike Finley (30:54)

Thank

Right, right,

right. And then a lot of enterprise users scratch their head and say, well, then how can I use it for my business? And you know, the short answer is you can solve that, that difference across answers. You can solve it with AI.

Or you can have AI generate tools that do it the same every time when you run them. There are lots of solutions for it. you know, it’s not, it’s not the same as it. It’s almost like a colleague, know, depending on whether they had coffee or tea this morning, they might, you know, answer you in one or two sentences. That doesn’t mean either answer is wrong. It’s just the way that that person presented. And ironically, it is very much like a colleague in the sense that what you’re doing when you’re hitting one of these cloud APIs, you’re interacting with a cloud instance.

Michelle Hamilton (31:20)

Yes.

Pete Reilly (31:25)

and

Mike Finley (31:48)

that is interacting with a lot of other people. Now that doesn’t mean that you’re getting cross thoughts from those other people, but it does mean that it’s not in the pristine state that you left it last time, because it’s been talking to other people, just like your colleague has been home and dealing with baseball and making dinner, and now they come back to work. The model’s been off solving rocket equations, and then it went and it wrote a term paper, and then it made some jokes for a five-year-old, and now it’s back to you to helping you, right? So it’s gonna have some variety when it comes through there.

But the thing that think we’re kind of getting to, especially with some of Pete’s use cases that I think is really important, is to go one more level of meta, right? So bear with me for a second. So the first level of meta is, hey, I’m going to write a letter to my boss. Here’s what I’ve got. Help me out, right? I’m going to use static tokens that come back from that. Okay. The second level of meta is, hey, I need a program that does this thing for me, right? Write a program that helps me automate something that I do every day, right? So now,

Now it’s not going to, it’s making a tool. So one is using the model. The next one is using the model to make a tool, right? Level three, level three is using the model to create uses of models. So what do mean by that? Right? What I mean is that if you want to make a new tool, like Pete, like the tool that you were talking about designing, and then you want to add a prompt to that tool that says, find me everything in the last month that had anything to do with

Delayed orders right something like that. So now now you’re adding AI to the tool that you’ve just built So AI is helping you add AI to a tool that you’ve just built right? That’s that next level of meta and it’s really where where the future is happening right now, right? We are we’re literally the interface of software is disappearing because it’s becoming prompts the prompts are disappearing because they’re becoming voice, you know, the voices is

Pete Reilly (33:30)

Right.

Mike Finley (33:40)

going to disappear because we just connect the models to the data and ask it for advice, right? So this is the sequence. And that’s why, again, teaching people to change is a lot better than teaching them how the tools work. The tools is like learning arithmetic. It’s good to learn how the arithmetic works, but we’re to get to calculus pretty soon, right? And that’s truly where the value starts to get delivered.

Nicole (34:01)

You know, it’s so interesting, Mike, the point you bring up about learning arithmetic or calculus. One of the things I’ve observed, have two college age kids, well, one recent grad and one who’s currently in college. And one of the things I really struggle with is that the universities out there are really holding

the kids back from AI. They have to go into even an AI class and sign an AI use statement. And I really feel passionately that our universities are doing a disservice to these kids who are coming into a world where they’re expected to understand AI tools. Now I am all for learning how to learn.

and you got to teach the fundamentals. But I think it is important for our universities to embrace the fact that the future involves AI and teach these students who are coming into a workforce to be prepared for those new rules of that workforce and to actually have the skills to know how to brainstorm with that AI and to be able to use AI appropriately in their workflow.

to be more effective and more efficient. You talked about the 10 IQ points. I think this is critical for the students who are coming into the workforce tomorrow to understand.

Michelle Hamilton (35:28)

Well, know, and I’m even thinking beyond that, even from a more humanist scale, universities are also an amazing opportunity for students to learn how to work together as people.

and how to look up from their computer and have conversations with other key stakeholders, other people who have curiosity and interest in same topics. And what I love is that AI can still be your knowledge worker, work alongside you, but now as humans, you can go back to connecting with each other and stop focusing on just a keyboard 10 hours a day and actually combine those two. And from a university perspective to waste money

Nicole (36:04)

Yes.

Michelle Hamilton (36:09)

on trying to catch the cheaters is doing such a disservice to our future when instead they could be thinking about how are we educating them to work as a team and to move AI for good forward.

Mike Finley (36:24)

Yeah, absolutely. The, work with my alma mater, which is Transylvania university. It’s a really, really old school in Kentucky. And, ⁓ and one of the things about that school for me was that I just took everything. I took every class they had in everything that I could fit in the schedule. And I came away with all sorts of different kinds of knowledge because I loved learning that. And now that’s an environment where, where AI can really flourish because it turns out you need persuasion and communication and music and art.

Nicole (36:24)

100%.

Mike Finley (36:53)

and computer science, right? And psychology to make AI really, really work well, right? And so it was in a, in one of our board meetings, it was kind of a funny story. And someone said, well, maybe we should have, you know, students opt into AI and professors opt into AI and we could have AI exclusion zones. And I just said, that’s going to be at the zoo. That’s the only place you’re going to find an AI exclusion zone is a little, is a little human exhibit at the zoo where, where they don’t have any, any AI. This is evolution. We’re, this is, this is how we move forward.

as a species and I think it’s really important.

Pete Reilly (37:27)

Speaking of how we’re evolving, I was gonna ask each of us to maybe talk about how our workflows are different, maybe just over the last six months. Maybe what was it like before? How is it different? And the reason I wanted to jump into that is because I think it’s hard for people who are not doing this to understand, how is it different? How’s the day different? What are the kinds of things I should be thinking about? And so maybe, Nicole, I’ll start with you. How would you describe your just day-to-day workflow as being different?

in the last six months.

Nicole (37:59)

don’t know how it’s the same. There isn’t a single thing that isn’t different. ⁓ I start my day asking co-work what’s on my calendar. ⁓ I used to go to my calendar and I used to review my meetings and now co-work is doing that for me.

Pete Reilly (38:11)

Mm-hmm.

Right.

Nicole (38:19)

I also, you know, I spend a lot of time with customers and so I spend a lot of time building presentations ⁓ and I tend to want to tell everyone everything. So in the past, my presentations are all a lot of words and I tried desperately to put pictures in it. Now, if I have, whether it’s an internal conversation or a

customer conversation, it doesn’t matter. If I have to tell someone a story, go into Claude and I give it my either, I think we were talking about this earlier, it could be meeting transcripts, it could be I took a walk and I recorded my just, I call it my verbal vomit. I just recorded everything I said, whatever transcript it is or whatever other context I have.

Maybe I have a spreadsheet with some information. Any context I have, I dump it in there and I say, build a presentation. I do tell it, you’re going to be presenting to X, whatever audience that is, whether it’s executives at a customer or…

or an internal team of analysts or whoever it is, I’ll tell it who you’re presenting to. I’ll tell it sort of high level, what’s the story I’m trying to tell. I’ll dump all the context and I’ll say, go build a presentation. And I find myself building five, six, seven presentations a day. And whether I use them all or not is another question, but I just dump that context and build.

Pete Reilly (39:46)

Right.

Michelle Hamilton (39:49)

I love that.

Pete Reilly (39:51)

Michelle, how about you?

Michelle Hamilton (39:53)

I have to start out the same way, Nicole, in what isn’t different, but ⁓ you know.

I went from running my own two companies to joining this team. It was previously helping me run my companies, so I had many bots running simultaneously. But now the ability for me to build out far more complex ⁓ workflows and ideas. I have a chief of staff that I built out for me that, like all of you, it checks my email, it checks my calendar, it checks my Slack, it does research on every person I’m meeting with that day, does personal

a mapping to their predictive disk and then tells me how to best deliver the information back to them so that it is delivered in a way that they contextually want to receive, not just the way that I tend to word vomit on people about AI. I’ve become a really boring dinner party guest.

Nicole (40:49)

you

Mike Finley (40:50)

Fascinating.

Pete Reilly (40:50)

And Mike,

Michelle Hamilton (40:51)

you

Pete Reilly (40:52)

you live in a bit of a different world, right? I mean, you’re on the technology side of things.

Mike Finley (40:56)

Yeah. Yeah. Yeah.

Uh, you know, so rather than kind of way it used to be, which is probably every waterfall textbook you’ve ever read about, or, maybe, uh, you know, sprint oriented agile book you’ve ever read. let me just talk about what happens now. Right. So, uh, so what’s unique now is for me is that, that every day something is released. So, so I’m putting out new software every day, roughly a two week sprint, roughly.

a two-week sprint is coming out every day from, and just from me, not from me and the team. My stuff is going into the work of a team, but it’s just, it’s a fascinating pace. Now, what’s interesting is it’s not just faster. As a result of it being so much faster, it’s also no sunk cost. If what I did yesterday, I wake up in the morning and say, oh no, there’s a much better way to do that. I don’t have to say, well, no time for that because that’s two weeks invested.

We just start over. We just throw it away because it was just one day’s work. mean, mind blown, right? Like now all of sudden, if I fundamentally chose the wrong technology, I can just start over in another one and catch up in two or three days. I am working every day in a tech that I never learned to do directly. I’m working with the models and we are building together. We used to say a full stack engineer was somebody who could do the back and the front and the database.

Now, and now it’s like you can design the screen, you can understand the feature, you can build the database, you can scale it out, you can do the release, you can make it all more cost effective, you can write the marketing materials, right? Like that’s truly full stack, right? And so that is the change in how AI is impacting it. It’s basically turning electrons out of the wall into brand new software for not much money, right? And so…

That superpower is the fundamental change.

Pete Reilly (42:52)

Yeah, I would say very similar to you guys in a lot of ways personally, as we talked about sort of the chief of staff, ⁓ use that constantly throughout the day. I don’t think I’ve built a single presentation by hand in the last three to four months. I don’t think I’ve built a single spreadsheet by hand, no matter how complex it is. ⁓ Even attempted, even done some legal work. And it just feels like if there’s anything I’m doing sort of the old way.

It doesn’t make sense. And one of the things I’ve observed about my own, maybe some differences ⁓ is one, I am building a lot of tool. Like I will literally say, great, go build me a tool to do this thing or automate this process, which I never would have attempted ⁓ six months ago. Two, I have this anxiety if I don’t have at least three to four agents spinning at any one time doing something, because I feel like I’m wasting tokens and ⁓ so on.

I do notice I am constantly trying to get a lot more done by spinning up agents to do various things and feeling like I’m missing out. But I’d say the other big thing for me is I have to constantly, I’m constantly when I have something I have to do, my old habit is, ⁓ I need an hour and half on the schedule to go do the thing or at least to draft the thing and so on.

And to me, the thing that I have to do differently is say, no, no, no, no, You, you not only, okay, fine, go book an hour and a half, but you could finish the thing, right? Instead of doing the first draft and second draft and so on, just knock it out. And the number of things that I need to just even continue to challenge myself to knock out in any one week, that’s one of my bigger challenges. It’s like be much more aggressive on what you can get done in a very short period of time. Cause it’s truly, truly game changing.

Nicole (44:47)

I think that’s right, Pete. think that a lot of us are in meetings all day. You tend to have five to 15 minute blocks for actual work. And the great thing is with these agents that we’re using, you can spin it up to start a task in those five minutes. And then your next five minutes, you’re reviewing a deliverable and giving it new instructions. And then in the next five minutes, you’re actually finishing the job.

Michelle Hamilton (44:48)

Thank

Pete Reilly (45:04)

Yeah.

It’s incredible old days you would have said, ⁓ 50 minutes. I I barely have enough time to even get up, get started and spun up. And now that’s enough time to knock the task out.

Nicole (45:21)

100%. Yes, that is a game changer.

Michelle Hamilton (45:24)

Yeah, I agree.

Mike Finley (45:25)

Let me

just throw one theoretical point out there, because I think it is important for everybody to sort of get this. ⁓ The critical fuel of every successful AI project is what’s called the context, right? It’s what goes into the AI, right? That is a precious resource. They talk about millions of tokens or hundreds of thousands of tokens, but it’s a precious resource because…

that can’t increase without causing a square power ⁓ impact on the cost and time of the result. So for all of us, even as the models get brilliant and much better and bigger and faster and cheaper, it’s gonna continue to be precious to get right what you’re putting in, right? And building tools to help you get that right and building ways of cataloging good examples from the past and building up your own kind of…

⁓ knife set that’s super sharp about how to slice the problem that you’re dealing with to get the AI to do a better job. That is going to be an evergreen as far as the eye can see right now, which is some number of years. A lot of other things may change, but getting the context right is going to continue to be an important capability. That means taking out stuff that’s not relevant, and that means making sure that all the right stuff is in there, and that means checking the result when it comes out, right? None of those things are going to change, even as the models get again faster and better.

Pete Reilly (46:46)

Awesome. Well, listen, we’re approaching our time. So maybe we’ll do this is just advice you guys have on folks that are trying to get started, where they should start or how they should think about it, whatever advice you want to give them. And Michelle, I’m to give you the last word, actually. And I’m just going to, I’m going to jump in. I’ll go myself, Nicole, Mike. And for me, I think it’s just simply jump in and just start asking it how to do the things you’re doing.

and be persistent. That would be my thing.

Nicole (47:19)

I absolutely agree and I will add to that Pete. No one’s looking over your shoulder. Don’t be afraid to ask stupid questions because no one’s ever going to know you asked them. So I agree with Pete. Just start, just start asking what can I do? And then if you have a stupid question, ask it a stupid question. Can I do this? And it’ll tell you, well, I can’t tie your shoe for you, but I can tell you how to tie your shoe. Yeah. Yeah.

Mike Finley (47:42)

That’s right. I

Michelle Hamilton (47:43)

can draw you a picture of it too. I can draw you diagram on that.

Mike Finley (47:47)

think we should all ask all kinds of stupid questions because you never know what you don’t know, right? But, ⁓ know, kind of like Pete was saying, my advice is to jump in and just realize something that this is kind of a Macintosh moment. So many people are Android and PC where you can do a lot of damage with the wrong right click and that’s why people are kind of shy about it. This tech is a lot more like a Mac or an iPhone where

where you’re not gonna shoot yourself in the foot by just trying something out. So go for it, give it a shot, see what happens. It’s not gonna take your secrets away. It’s not gonna charge you million dollars. It’s not gonna infiltrate your home. Just give it a shot, try it, make it work.

Michelle Hamilton (48:28)

And I’m actually just end it by saying, I agree with exactly, I mean, I think we were all thinking the same thing, but I wanna even back it up a little bit further from there. I want you to not be overwhelmed by the Perplexity of parts of this conversation if you’re just thinking of getting started or maybe you’re just gonna take that next step.

Pete Reilly (48:28)

Yeah, that’s all.

Michelle Hamilton (48:49)

AI doesn’t care what street you grew up in, where you went to school, what board you sit on. It does not care who you are, what country you live in, or what you look like. It only cares about

the conversation that you are having with it. And I will tell you, we all came from very, very different backgrounds. And yet we are all lucky enough to find ourselves sitting here today working for an AI company. But that is because we are willing to be open and think and ask for more. And this is your opportunity to keep asking for more.

Mike Finley (49:24)

right.

Nicole (49:25)

Love that.

Pete Reilly (49:25)

Awesome.

Well, thanks everybody. Have an awesome rest of the week. Take care. See ya.

Michelle Hamilton (49:31)

Thanks, Pete.

Nicole (49:32)

Thanks.