John Ladaga - Building a Virtual CFO with AI

John Ladaga - Building a Virtual CFO with AI

Episode - 020: - Where we learn from John Ladaga about building a remote business, organizing with tools like Slack, Trello, and Loom, fostering a positive work culture, AI's impact on creative arts like music, connecting financial and non-financial data to make decisions with CFOChat, the evolving role of AI in knowledge work, and the security of AI technologies.

John is an experienced entrepreneur and the Co-founder of Sturppy, a company dedicated to helping users build financial models fast using proven templates. With a background as a digital strategy consultant at Accenture, John, along with his co-founder Filippo, established Sturppy in 2021.

In his free time, John embraces the great outdoors of Colorado, enjoying activities such as skiing, hiking, and playing Pickleball.

Audio Podcast Links

Show Notes

Where to find out more about John & Sturppy

Websites Mentioned

Tech Mentioned

Books Mentioned

Other Mentions


(0:00:00) - Building a Business Remotely

(0:14:09) - Organize With Slack, Trello, Loom

(0:18:50) - Building Human Connections

(0:22:40) - AI's Impact on Music

(0:36:07) - Connecting Financial Data

(0:45:54) - AI and Knowledge Work

(0:52:55) - The Benefits of AI Self-Reflection

(1:00:02) - Building Live Systems With AI

(1:03:43) - The Security of AI Technologies

(1:12:18) - Optimistically Leveraging Technology for Success

Chapter Summaries:

(0:00:00) - Building a Business Remotely (14 Minutes)

In this episode, we explore the process of building a remote company and the challenges faced when working across time zones. Our guest, John Ladagar, co-founder of Stirpe, shares his experience of finding a co-founder through an online platform called IndyHackers, where he provided free feedback to various founders before finding his current co-founder, Filipo. John highlights the importance of being transparent, open, and moving slowly when finding a potential co-founder.

(0:14:09) - Organize With Slack, Trello, Loom (5 Minutes)

In this segment, we delve into the organization and management of remote teams, particularly through tools like Slack and Trello. We discuss how Slack is used for communication with contractors, while Trello helps keep tasks and projects organized. Moreover, we touch on the benefits of using Loom videos for asynchronous communication and overcoming language barriers between team members. Additionally, we explore the impact of AI translation tools like GPT and their availability in different countries'

(0:18:50) - Building Human Connections (4 Minutes)

We examine the importance of maintaining engagement and fostering a positive work culture in remote teams. The conversation highlights the need for human-to-human connections and creating opportunities for team members to showcase their personalities through informal channels like Slack. Additionally, we discuss the value of organizing in-person gatherings for remote employees and touch on the challenges of translating a unique office culture to a remote setting. The discussion also delves into the increasing impact of technology on various industries, using the example of an AI-generated song featuring Drake and The Weeknd'

(0:22:40) - AI's Impact on Music (13 Minutes)

We discuss the potential impact of AI on the creative arts, particularly in music production. The conversation touches on the importance of human imperfections in creating authentic art and the potential loss of this authenticity with AI-generated content. We also explore the use of Loom for internal communication and customer support, highlighting its capacity for conveying tone and humanizing interactions. Additionally, we consider the benefits of treating internal peers as customers and utilizing technology to streamline communication and improve efficiency in finance teams'

(0:36:07) - Connecting Financial Data (10 Minutes)

In this part of the conversation, we explore the potential of AI in the finance industry and its implications on traditional roles. We discuss the use of AI-powered tools like Stirpy Plus and CFO chat, which can help automate tedious tasks and provide real-time insights based on live data sources. As AI technology continues to advance, it raises questions about the skills that will be essential for professionals in the future and the potential impact on traditional roles like CFOs. We also consider the evolving landscape of chat interfaces and their potential to become an all-encompassing ecosystem for businesses.

(0:45:54) - AI and Knowledge Work (7 Minutes)

In this section, we consider the evolving role of AI in the finance industry and how it could potentially displace knowledge workers. We discuss the use of AI technology to surface KPIs and proactively monitor and report on data in a living model. Although AI can provide valuable insights, we acknowledge that human intuition and decision-making are still essential in determining which insights to act upon. We also touch on the topic of AI self-reflection and its potential implications for the future of AI in various industries'

(0:52:55) - The Benefits of AI Self-Reflection (7 Minutes)

In this discussion, we explore the concept of self-reflection in AI and its impact on the accuracy of AI-generated content. We mention a study that found a significant improvement in accuracy when AI models self-reflect before providing answers. The conversation also covers the differences between chat GPT and GPT and how these AI models are evolving rapidly, raising concerns about keeping up with their development. We touch on the idea of pausing AI advancements and whether it is a good idea or not'

(1:00:02) - Building Live Systems With AI (4 Minutes)

We explore the distinction between GPT and chat GPT, and how it is crucial for companies to approach AI implementation in the right way. The conversation highlights the importance of training AI models on relevant context, particularly in the world of finance, where data is constantly changing. We also discuss the challenges companies face when attempting to build their own AI systems, emphasizing the value of expertise in building algorithms and platforms that can adapt to evolving data and context'

(1:03:43) - The Security of AI Technologies (9 Minutes)

In this part of the conversation, we discuss the emergence of autonomous AI like auto GPT and the potential implications of its development. We consider the balance between advancing AI technology and ensuring its safety and alignment with ethical standards. The conversation also touches on the idea of an FDA-like model for regulating AI, highlighting the challenges of implementing such a system, given the fast-paced nature of AI development'

(1:12:18) - Optimistically Leveraging Technology for Success (6 Minutes)

In this portion of the episode, we discuss the positive side of AI advancements and the potential for individuals to use these tools to become entrepreneurs and follow their passions. We also touch on our mission to help small businesses and the goal of leveling the playing field with AI technology. The conversation concludes with a discussion of a favorite piece of software, Licecap, a tool for recording screen captures as GIFs'


Transcript generated by

Help us spread the word by tweeting about us at @PodiumDotPage and including us in your shownotes!


Well, you know what's even faster to solve that problem? Well, what we're building with chat, with CFO chat. Because I'm I'm just listening to you explain, like, how finance teams are overwhelmed. Folks just lobbying things over over email to them. I need this board. I need this to be changed. I need that. And you know who doesn't care about stupid questions or repetitive things or minor changes to grass.


AI? Hello, and welcome to Tech Forfinance where we help finance professionals leverage technology to level up their lives. I'm your host, Adam Schultz. And in this episode, we're gonna be chatting with John Laddager. John is the co founder of Stirpe a company focused on using proven templates to help build financial models fast without any prior experience or expertise in finance.


Before venturing into the world of entrepreneurship, John spent five years as a digital strategy consultant at Accenture working in various roles and accumulating wealth of experience in the industry. Originally from Atlanta Georgia, John attended Orbi Renewable University where he obtained his bachelor of science in marketing and business engineering technology. That's right. Isn't it? And he laid some boots to Denver, Colorado in twenty eighteen where he eventually started therapy with his cofounder, Salipo, Bien, twenty twenty one. The two of them met on lines through IndyHackers and worked together remotely for the first twelve months of building the company before ever meeting in person. In his downtime, John loves to embrace the outdoor activities. Colorado has to offer from skiing and winter to hiking playing pickleball in the summer.


So before we start, if you like what you hear today, please make sure to subscribe to tech for finance on your favorite podcast platform and on YouTube. But thanks for joining me today. John, it's great to have you on. Thanks so much. Glad to be here. Thank you for the intro. I think you covered it well. Appreciate it. We we we do our best day. We we we've we've gotta, you know, we've gotta make it, you know, a a good intro to keep people listening. Not that people don't don't listen, but you hear what I'm Right. So Absolutely. Absolutely. There were no. No. Absolute. I mean, I've I've been keeping up to date with with therapy. I am pronouncing it right. On that exactly? You are.


And and maybe it would help for a little bit of context if I tell the story of the name because we get asked you quite a lot. And I was just chatting with someone and he's you need to have this on the website somewhere. So my cofounder and our whole technical team actually is is from Italy. And so it was actually my cofounder that came up with a name. And in Italy, they say to do things in a a startup y way. And so startupy became a slurpee. And so that's that's the the reason behind the name, the kind of the story there, and We thought about changing it, but we said, you know what, let's stay true to our roots. There's nothing wrong with being a little bit unique, and I like the story behind it. So that's our name. But that's it.


And and it's reminded me of the book I read recently called ideas that stick And this is a complete side though. Right? You know, this this this is bonus content for our listeners. Right? But simply ideas that stick is is about copywriting. So I like LinkedIn. I like writing posts. I like doing the analytics. I thought, right, well, what if I learned a little bit more about copywriting? And he just reminded me because you spoke there about it being a little bit different. And and in I in ideas that stick The acronym that they use is success just without the s on the end.


And it stands for simple, unexpected, credible, apartment, what the other c is, but but you get the district. So -- Yeah. -- so you fall into the u of that category with the ideas that stick, which basically, an unexpected names. And so, yeah, I think I think it's memorable and I think it will stick with people. Once once they understand it, you know, which sounds good. And then the brand is is very recognizable. So It's all good. It's all good. It looks yeah.


It wasn't a hard domain name to get. So that was as nice as it was like, you know, that one was available. Yeah. That's it. And I know it's yeah. It did it years ago. Remember listening to all sorts of conversations saying it it's sad when generating a new business idea is limited by the domain names that are available. It's the Yeah. Will do we live in?


I just saw somewhere someone was saying their domain register was the dot ai domains. They're increasing the price of them. And I'm like, oh, yeah. Totally makes sense. There's probably then more dot ai domain names registered in the past, like, six to twelve months than ever before. And whoever's running that registry is probably, like, yeah, we need to up this price because these are hot right now. There there must be.


And and again, I've I've not really done much work with with web web crawling and scraping, but there there must be a way that you could do like filtered search of domain names with an AI on it. And I know there's probably millions now. But but I've been thinking about this recently because it a database that I use to to keep an eye on them guide tools. Is a site called Des and AIs for that? I think it's I think it's fairly recent, but it's it's growing all the times. Yeah. And of course, they're they're compiling that information. Into different categories, you know, is it chatty boutique or spreadsheets and and that sort of stuff. And but there's no way of sort of filtering by industry or or use apart from the the tags, the use case at the moment, I guess. So I was thinking, you know, is there a ways that you could literally just do an Internet search for all of the AI tools that there are. And and I guess that might be a business idea in itself. Right? But I always don't think about it. But, yeah, that definitely would the AI domains, I agree. He's a lot more now than there was a few years ago. That's for sure. Absolutely. Very good. So The first bit is is separate to to what you guys are doing at at Stepping and I and I think it will be useful for for listeners because And I've actually been doing some training recently on on managing remote teams.


Mhmm. And I think when we look at employee retention, and and when we look at the way that we we communicate flexibly as as managers and business owners, I think there is still a bit of a barrier when it comes to remote work. And of course, it's supported by the tools that we use. You and I are speaking on Zoom. Now, I can hear you crystal clear no no problem at all. But having a conversation on Zoom is a little bit different to running a business on search.


Coming back to the fact that you didn't meet your business partner in person for for twelve months. Was that a struggle? And Were there any takeaways from building a business that's all amongst remotely that you think would be useful to not not people that necessarily building businesses, but people that managing remote teams are working remotely. Yeah. Absolutely. You know, I think before COVID, it was an extremely bizarre, rare sort of thing, to to for for two people who had never met before to come together and start a business. Now, you know, I I tell that story and and folks say, oh, I know, you know, the founders at such and such company also met in the same way. It's becoming a little bit more mainstream. And I think because of technologies like Zoom, Slack, etcetera, it's becoming more of the norm.


I do have some advice about, you know, for anyone out there that's thinking about starting a company and finding that cofounder, maybe your technical and you're looking for someone that's on the marketing and sales side or vice versa, and what that process should look like. So this is actually this is my second startup. And with this second startup, I was really diligent about the process of finding a co founder. And and so actually what I did and and kinda how I met my cofounder today was on a form called indiehackers dot com, really cool website for for folks, mostly technical people that are trying to build kind of a side project or side hustle with the dream of turning that into, you know, their full time living. And so in that process, I I wanted to yeah.


I I was kind of ramping down my my previous company, and so I I wanted to be involved with more of a technical audience and I wanted to kind of, for lack of a better term, kind of interview some folks. And so the way that I did that was I I went on to Andy Acres and I started saying, you know, hey, I'm I'm a marketing and sales professional. I have experience with tech companies. Let me let me provide some feedback for free completely pro bono. If you need some help in that area, let me let me help you there. And so I actually worked with about ten different founders just completely free, giving them feedback about their landing page, their website, their marketing, their copy, etcetera.


And through that process, I got to work with a whole bunch of different folks, and I got to see the things that I liked, the things that I didn't like, and the ones that just came the most natural. And for me and and Filipo, my cofounder from day one, it was incredibly natural. I remember I sent him all this feedback around his landing page and I get an email back like a day later and he's like, Great. Done. Implement it. What's next? And I was like, oh oh, shoot. Alright. This guy's serious.


And so actually even before we officially made our partnership official. We had this kind of dating process of for a while, we were just working together completely with with no talks of equity, with no talks of roles, and it and it was pretty natural. And I think going through it that way, and then also explicitly once we kinda got further down the process talking about our own experiences with cofounders in the past. He came from another startup as well. He had a set of experiences. So we got aligned on on kind of that process. And we were very transparent in opening with each other in those early days. But so I guess just my advice to anybody that's trying to find a potential co founder for a company is to, you know, just like dating, you know, Take take a couple shots, talk to some different folks, see what resonates, move slow, and and you might be able to find your your perfect partner there. As far as running remote companies, I think it's a lot easier at at a small size.


So from Denver, Colorado where I live, to Encona Italy where my cofounder and our developers are is a eight hour time change. Which that ate our time change, I'm not gonna sugarcoat at its stinks. But I'm an early person, so I'm up early. And actually so so we get an overlap from, like, six AM my time till about noon my time. So I get about six hours overlap with the team. And they start their morning completely uninterrupted from me and then I end my day uninterrupted from them. So to actually sort of works out nicely having those pockets of time where we can come together and then those pockets of kind of deep work and and solo focus. Right.


And and, yeah, that's that's definitely glass half full mentality, which is, you know, Let's let's let's not think about the difficulties in working across different time zones. Let's let's think about the positives. And, you know, providing you've got your your phone notifications turned on, of course, you know, because if you wanna lie and then you might not want somebody pinging you on Slack at, you know, six or or whatever happens to be Yeah. Using the, like, the focus mode on my iPhone during during all hours is is is key. Otherwise, I start getting pings at two AM from Trello and from Slack and and all that. So I'm definitely, like, a big user of the focus mode when I'm going to bed for sure. My cofounder does the same. So focusing on Slack and trial items. Yep. And, again, this this wasn't in any of the question topics dip that I've sent across.


But so I I I I use Slack but not for work. So so Slack for me is community based, you know, so it's some part of a community called GrowCFO that's got a pretty good Slack network. And I've got a a couple of I guess, experts in the finance and AI community that I've got a Slack channel running with, but I'm not, you know, it's not managing a business. It's more sharing insights and becoming part of a conversation. For for work I use use Teams, Microsoft Teams is is achieves the same ends. You know, you've got different channels, you know, different teams and and you can you can work it that way.


So I'm just curious about your setup. So you're still a small business. Right? So you've got developers. You know, you've got people that that do different things. So and it doesn't have to be detailed, but I'm curious to know how you you structure your Slack. Is it is it siloed in? Right. Sales marketing content, you know? And is there a channel that only you and your founder have access to that you have a conversation with? And then is a completely separate development channel. And you had an approvals built into Slack, you know, to say that when people are completing work, you have to have a look at it before it moves on to the next. So I'm just curious and I'm thinking about my listeners, you know, that are either using some light teams in Slack.


A lot of people use it in a disorganized way into my so I'm just curious to know what that setup looks like for you? Yeah. I would say that we're definitely we're more organized in Trello for for kind of the approval process, things that are in the backlog. We we kinda have a meeting to kind of review the backlog. Say what we're gonna put together in in this week's sprint. So a lot of our organization is more so in Trello for kind of organizing work.


But Slack also, we we do have some organization there. So we have we also have some contractors that we work with. So I have a specific channel specific to kind of one of the marketing and agency companies that that that I work with. So they're in one channel. And mean my cofounder have access to that channel, the rest of our team, they just don't need access to that. They would just get bogged down in it. So we do have it siloed into different channels. We kind of have a general channel. We have a a kind of sales and marketing channel. And then we also do have some integrations that are helpful as well. So we have intercom. We use intercom for kind of all of our live chat support, help documentation, and we have a chat bot that will ping the Slack channel and say, hey, listen, someone's knocking on the front door. You need to answer them. So we do have some kind of automation built in there. But to be perfectly honest, Slack is more of the place where, like, the the memes get shared, the the AI trends get shared and and it's more of our kind of community forum and less about the staying super organized where we do a lot of that in in Trello.


The other tool that I would say that I use a lot in in kind of an asynchronous way with my team and also with customers for customer support I'm obsessed with loom. Do you know what loom is? The the Yeah. It's been around for forever, but I love it because a video can you know, a video is worth ten thousand words. And so often, whether it's a customer that that wants to do something with our software and it's kind of confusing to say, oh, click here and then do this and then then that all with text or I can just send a ten second loan video and record my screen, record my audio, and then and then I'm off to the running. Same goes for also sending design and kinda UI UX feedback to the team as well. If I'm up late at night, the team's asleep, and I'm kind of tearing down a particular feature. If I run into a bug, I'll oftentimes just record a loom video. And then throw that Intrello attached to kind of a card in the backlog.


And it's just so much easier, especially there there's also with with my team, there is some level of language barrier So my co founder speaks beautiful beautiful, beautiful English, but that's not the case for all of our Italian developers. And I speak a little bit of Italian. I'm I'm trying to learn Italian more and more, but video, you know, is is kind of synonymous and I can kinda shoot over a loom video of a particular bug and it's a lot less up for interpretation than than using text. So we we use I use loom a lot. Yeah. Yeah. But you could not you get child TV things that translate for you now. You can, but but my team is in Italy and GPT has been banned in the country of Italy. So so, you know, they are they are up in up in arms about that.


One of you like, a a couple of my developers that aren't as fluent in English. We we we try to do this thing where The developers are only allowed to speak to me in English, and I'm only allowed to speak to them in Italian. Me, my cofounder, we speak English, straight back and forth. But we both my team, they wanna learn English. I wanna learn Italian. And so we we do this back and forth. And a lot of it still, you know, just translating, and my developers were using chat GPT for that. And then the Italian government came in and banned chat, GPTs, and now they're back to Google translate. So But but, yeah, pretty funny. Yeah. He he he is a pain, and and I other podcast guest I had on Angel Lan, she she comes from a executive search, you know, sort of headhunting background. Mhmm. She was doing a bit commentary on one of the posts I did to that. And she she decides that it's still not available in in China and Hong Kong as well. So -- Yeah. -- there are pockets that aren't able to to use it separate from whether that is the right or wrong thing to do. For countries to to ban large language models.


Coming back to the point there that you said about of the year, the Trello being more of the operational side of things, you know -- Mhmm. -- which is the, you know, I guess, Kanban stage process, you know, information against the card and then that's moved where it needs needs to get it. That's fixed. Just wanna touch on the point that you made about, like, gifts and community and Slack. Because coming back to to the question that I asked that the answers very nicely.


Beginning, you know, how does that remote work piece play out? How do you make sure people are still engaged. People are still, you know, having a good time, you know. And if people can't have a lot and have a bit of a giggle, then, you know, what's what's the point? You know, so so I'm I'm a full advocate of having a, you know, a random Slack channel you know, where by people could just be a bit silly and share jokes and share the share the funny stuff because it's, you know, it it makes you more human and it does really needed in the world that we live in right now.


You know, human to human connections is a mustard thing that you really struggled to achieve work from. Oh, totally. Yeah. It's it's really important to us and and kind of the culture that we're building to, you know, we have a small office in Italy and lunchtime in in Italy. It's it's interesting too just like some of the working culture differences between the United States and and and Italy. But lunch is at the same time every day. The whole team takes lunch at the same time, which I feel like in the United States, we're less kind of structured with our our lunch. It's like different teams will take lunch whenever sometimes meetings get but lunch is kind of a sacred time. One o'clock is lunch. And then after usually the game cube comes out and and and it's thirty minutes of of playing Super Smash Brothers and whatever other video game is is brought to your attention.


But kinda having that, you know, and and translating that to a remote culture is tough. You know, you don't have the same kind of, you know, luxuries of being in person. So one thing that we definitely try to do I try to make it over there kind of once a quarter to spend some time with the team in person. And I think as we scale, as we continue to grow this business, we will make an effort to kind of try to at least bring folks together once or twice a year from from kind of wherever they might be in in in in kind of the world to to come together, have that team building time. So I think between having some contacts, like some actual physical contacts where folks are in the same room a couple times a year. That's important, but then also allowing to, you know, for the folks that are remote to kinda have those opportunities to share a little bit of their character, be a little bit goofy, and so, like, having a Slack channel for that, I think, is is great.


And we always kind of, you know, start off our our scrum calls and we we we we shoot the shit a little bit and and talk about, you know, whatever's going on in the world, funny things. The the latest topic we were talking about was this new AI song for Drake in the weekend. Have you have you seen this kind of going viral on Twitter? No. Oh, goodness. So I will love it. Yeah. It's interesting now. Right? Because music be fake. Yes. Yeah. And it's, like, it's good. It's a banger. So, like, So someone took and, you know, cloned Drake, cloned his voice, and then closed the weekend's voice. And then made a synthetic d with AI and then created a whole, you know, song with it and slapped it all together, and I think it has, like, you know, ten million views on TikTok and a quarter million views or streams on on Spotify and I think that the music industry is kind of freaking out a little bit. So we were kind of discussing that and and and that that was that was one of the discussions in our in our chat this morning. And excuse, how are you pleased that sort of stuff is then? Yeah. Isn't it? It was And unfortunately, I think as we've seen historically, you know, when something gains enough momentum online, it's incredibly difficult to take it down.


You know what? Especially music. Yeah. I mean, we've never we've been a culture globally of of not wanting to pay for music Yes. So so I I it'll be interesting to see kinda how how this gets handled because I think it's coming. I I I think within the next six months, we'll we'll have like a number one global chart hit that's completely AI, whether it's ripped from some other artist or completely new. I I wouldn't be surprised if if that that industry kinda starts to get disrupted a little bit more. Yeah. I don't I don't think I don't know.


Then people will still have a preference. I mean, you you probably see in the back kind of sky junkets and some guitars on the wall, and people listening can't see, but as we mentioned before we started recording, my my degree of the music productions, I'm gonna call it. Right. My career did not end up there, you know. And Part of it was the corporate ladder, you know, was following the money, you know. Mhmm. And if I hadn't needed the money, then maybe I would have done the the musician thing. But yeah, all too familiar with the whole sort of you know, musicians being paid nothing for a substantial portion of their life before actually making any money at all. I think people will still have a a preference because people will to me, auto generated stuff. I mean, taking ChatGPT when it arrived at content you. Right?


There's there's something that's almost too polished about it. You know, with with the literally having zero grammatical errors, you know, Same with music. You know, people already get a bit fatigued by fight with commercial music. You know, especially the the heavy produced you know, really loud, you know, something in the chest that ends up, you know, is is great on occasion, but some people get really fatigued by that and and they want more of the naturally produced Where there is the mistakes? You know, the the mistakes kind of make up part of the authenticity of it? Totally. And and that that's my concern with with AI is that we get to the point where, you know, that there just isn't any room for, a, the ability to make mistakes be that authenticity anymore. And and it'll be a chance to lose that. I I think it is gonna be a while, especially for the creative arts, you know, before Put it this way.


I think there's more states that AI to replace certain operational roles, I. E. Admin heavy tasks -- Right. -- like like what you were doing with with your business at the moment, then there is replace a metal bound, for example. You see what I do? I agree. I agree. Yeah. Yeah. It's interesting. And I think you're right. I think those imperfections is what makes art and what makes music beautiful. And so it's it is kind of scary to think about a world where things are are kind of run by AI and it removes that level of imperfection that makes things beautiful. So Yeah, that is kind of a scary dystopian thoughts that that I hope we can humanity can avoid and we can hang on to those things. Yeah. Yeah. So yeah. Let's let's not not go too far down that rabbit hole today, hey, segue.


But then coming back to the point about making sure that that human element is built in and and quite, you know, quite rightly, you know, you you made the effort to go and actually spend FaceTime with the team, which I I wholeheartedly agree with. You know, we we do a similar thing once once a quarter as well. But for for the remote stuff and I'll just use an exam whether it's useful for listeners or not. But I mentioned that we use Teams. And I I did try a a while, but And it's so so subjective, it's difficult to actually do a bug.


I tried to write a cheat sheet for what form of communication would be best for what purpose. Yeah. So if it's super urgent needing an immediately response, pick up the phone. Right. And and if it's low urgency, but doesn't warrant an email. Is it is it an I'm, you know, either a Slack or is it a teams, you know, at if it's long form meant to inform, you know, is it an email? You know, so so I kind of tried to apply some thought because the initial disruptor came when the likes of teams and these slap type applications came up because it kind of shifted from traditional email and phone calls into more of a gap from Calabrio where working, which I think worked through the story I'll share is is with a member of mine team saying, we were going through some some analysis, and and I noticed something, and it doesn't matter what I noticed but essentially something that shouldn't have happened.


And I sent a sure Teams message to that individual to say essentially look into this because it shouldn't have happened. And and it was the wrong it was the wrong medium. Because immediately that instant message came across as cold and as hypothetical. Yeah. So so I got because because they were tired, you know, it was it was coming towards the end of the week, you know, and and I've got really short snappy response essentially, you know, I'm doing as best as I can. And don't give you right out. I'm by no means a dictator or cracker of the whip or anything like that. You know, I I do make a concerted effort to be, you know, a patient, you know, all all of that sort of stuff where where I can't. But in this instance, I've got it I've got it wrong. Right. As a learning aside. That has got me then refocusing on. Right? Well, with the technology that we have available, what is the correct medium? You know, and more often than not, I'm now leaning more towards face to face calls like this, you know, rather than that instant message. But obviously, that has to be balanced by how much time we've got because some people like to talk more than others as well. Right? So how are you gonna cut people off after after fifteen minutes? But but that decides But the reason that I'm speaking about that again comes back to your point about using Loom.


Yeah. Because I use Loom as well. And But both of it, you know, take a finance and end work just because it's it's super easy. But some listeners that don't know about it just to it's found on that a little bit. Loom is it means of recording yourself and your screen. Yeah. You you can use it to record just yourself if you want to, or you can use it just to record your screen. The the choice is up to you. But the theory is because they're automatically loaded to the cloud, you don't have to produce them. You don't have to make you don't have to render them or or do and give that any of that post production stuff, you'd literally end up with a video where you can copy a link and send it to somebody else in seconds. So it's the speed that's the real attractiveness of Lou.


And the reason that I'm focusing on this is whether it's a finance team or an operations team or a sales team, coming back to your point there about what is the correct medium, you know, A video says a thousand, ten thousand words in the space of three minutes. Right? So really good use cases are for stuff like onboarding. Quickly producing the stuff that you repeated a hundred times for every new starter that's come onboard, you know, just just put it on a shared Slack channel or put it on a shared teams channel, you know, tech training. You know, if if if there is onboarding that's normally done one to one with internal systems, whether it's a finance system, you know, whether it's Use of internal systems like Teams or Slack. Again, you can you can do a loom. You do a three part loom that just shows the bits to a business specific to you and then share them in the user guide for the rest. See what I mean. And I think there's some really quick wins that people don't necessarily think about all the time. There could be really good ways of speeding up process and just making sure that time is spent in the right areas? Totally. Another one that is is always good too is is using Loom for for customer support.


And I think it it it goes back to what you were saying with with kind of text based communication. You can lose that tonality You know, it's hard to kind of it doesn't carry the same level of tone or or or kind of really how you're trying to convey that message. If I say, hey, I need this. Is that I need this now? Is that I need this later, like, it it doesn't come across in text, but with a video it does.


And I found with with customer support, You know, if I'm just chatting someone on our live chat, it's easier for them to kind of dehumanize the the chat with me and and folks can get a little bit snappy. And we've opened there. Right? If you're chatting customer support, you're you're not happy in the first place. Ideally, you never have to chat customer support for any of the things you do. But when you do, we we, as humans, I think, have associated with such a negative experience that that we're oftentimes nappy talking with customer support folks. And sometimes, I'll just send back a loom video. And I'll say, hey, John Laddaga, co founder, Stirpy, just wanted to show you quickly, you know, how to do this one thing in the software. People's response to that is like, oh my god. They're like, holy cow. This is a real person. This is a human behind the screen. It's the co founder talking to me. And I don't know. It's just a little bit of a hack because I think one good customer experience can can keep a customer for life in a lot of cases.


And and have a customer that will actually refer you additional customers. And I think Loom is a super easy tool for for doing that. And we don't use it all the time, but sometimes it I I'll just hop in. I'll even do it with once a week, I'll kinda go through and look at new sign ups. And I'll pick three, four, five new sign ups from the week. Just completely randomly, I'll kinda search their email. I'm like, Dan and and and kinda figure out a little bit about what, you know, what the company is, what they're doing. And I'll shoot them just a a quick thirty second loom video, welcome to Stirby email, And the responses I get from that are are are crazy. So, yeah, it's a it's it's great for internal communication, but also for customer support and outreach like that too. I think it's just great asynchronous tool for those things. Yep. And I think just expanding that a little bit to also be considerate of internal customers as well? Yeah. And and and I know so slight distinction between what you were talking about with the team in Italy and what a finance team may talk about in terms of another department. Yes.


And and we've spoken about this talk a couple of times on the on the podcast. It's, you know, you need to treat your peers as customers, you know, not just from a service level and quality of the work that you do, but also from a gauging feedback. Point of view. Hold on. And the thing that we we get caught up on as as finance seems quite a lot.


And this This is from the perspective of somebody who's nicely worked in a finance team, just works with a lot of finance teams, is that the challenge is that the the numb the number of things that just get logged over finance, you know, oh, can you do this for me? Can you get that report for me? Can you do that? And immediately, you end up with an email thread. It's like, twenty and twenty first of the long set, you know, old key just changed it to do this or, you know, oh, I didn't mean that can you do this instead.


So coming back to your point there, you know, if we've got, you know, a tax savvy finance pro, and I like using, you know, tax savvy finance probes, you know. This this is my sweet spot in terms of audience, but techs have a finance pro. And if they're using tools that enable them to quickly share and relay information to their internal customers. So so somebody comes to to me as a, you know, either somebody in in FPA or fine controller, you know, or something like that and says, I could really do it with some help with this, you know. And and then they do a screen recording.


If a couple of the reports, they've got maybe they've got some dash boards, maybe it's just an Excel pivot chart or something like that, and just walks them through it to say, well, here's a couple of things that might help with that, which which of them is is closely aligned. That way, you're not making any assumptions. You're not having to make sure that you email, like, backwards and forwards. And instead, you're able to say, look, here's a thirty second video. What do you want from me? And I think immediately that removes the barrier of what we're talking there about just mindless chat messages and emails going backwards and forwards, and it's focusing more on the value and an actual proper nuance to something instead of just being a a zombie sending reports here there and everywhere. Well, you know what's even faster to solve that problem? Yeah.


What we're building with Chad with CFO chat. Because I'm I'm just listening to you explain, like, how finance teams are overwhelmed. The folks just blobbing things over over email to him. I need this report. I need this to be changed. I need that. And you know who doesn't? Care about stupid questions or repetitive things or minor changes to graphs? AI. Yes. So So it's interesting, like, that I think a lot of that back and forth, a lot of that problem that you're talking about yes, video is great in in a team scenario, but what's even kind of a step further is is kind of what we're what we're building with with CFO chat. And and maybe I can just explain just for context. I know you're aware, but maybe for your audience of what we're building with with CFO, Chad. And I think it's just a good segue for for in the in the business. Fili boots. Yeah. But for the nations here. So nobody's gonna tell us enough for talking about a particular solution. So So Yeah. But Although, I don't know. Maybe Loom wants to be a sponsor of the pod app. Yeah. Of some Let's But yeah. So just just to kind of bring it full circle here.


You know, I'm a cofounder of this company, Survey, and we build financial planning and analysis tools for founding teams. We have two tools. We have Stirpy Planning. That's an early stage tool. It's all driver based, all assumptions based. Not connected to any sort of live data sources. Really great tool. It's been used by thousands and thousands of founders in in in ninety eight countries build simple forecasts. We have this new tool called Stirpy Plus and it's connected to live data sources. And so what we do is what we We connect all your live data, so from your GL. So whether that's zero QuickBooks and NetSuite name it, and then we also layer in operational data. To build what we consider to be a full living model of the business. And then we've actually introduced GPT four into that in the form of what we're calling CFO chat. And so CFO chat is a Always available.


Twenty four seven doesn't care about your dumb questions. AI, they can answer, create charts, do analysis and and and really cut back on a lot of what you were just discussing of kind of some of that tedious back and forth. For internal finance teams or for founders that are trying to get a better understanding of their finance. Yeah. So so it's just kinda interesting to hear you hear you talking about that problem because it's it's it's similar to to kind of the problem we're trying to solve here. Yeah. Yeah. A hundred percent. And and and it all comes back to that, you know, what's the stuff we don't wanna do? Well, how do we get rid of the stuff we don't wanna do, so we can do the stuff that we do wanna do. In some instances, and it depends. So We've also talked about, you know, personal brands on the show before where people internally might want to build their profile, especially if they work in a larger company.


Right? So there's there's a decision to be made, you know, do I want to take this on? Is it going to help me build influence and credit impact? Yeah. So so in which case, you know, maybe I organize a presentation, maybe I send the label, you know, maybe I made people very aware that I'm an expert and that I can solve their problems because I want to be known as the pro problems solved for my career. Sure. Yep. What you're also talking about there is on the other side of that, which is that it this isn't gonna help me, you know, build influence, create an impact, or enhance my career. This is just a question that I'm asked you know, two or three times a day, and I could do without going back back to all text. Now, I'm gonna give you a bit of a plug because everybody knows I I I love the shit. Right?


So the the the video you put on LinkedIn and and and I'll put it in the show and so so people can see. And so so as we said, it's it's called CFO chats. And and the example in the video was was a was a a SaaS business. So so pro probably using a bit of your data, I'm guessing, or, you know, you can't download your data in in the demo system there.


So the question you asked was and find a month where subscribers grew the most. Yeah. And and and that wasn't revolutionary. Right? Because he just came in a month by month break down if, you know, smaller number, bigger number, smaller number, bigger number. And and you could get that information without having to ask, you know, a chatbot. That's not a particularly good use case.


But then the next bit and this is the mind blowing bit. You know, not just with with therapy, but just with, you know, AI models connected to data -- Mhmm. -- is you then asked. Right? Well, why why do you think that's the best month. Why do you think it grew that much in in the months?


And the AI came back saying, and and it was like four or five parts to the response, but the one that jumped out to me was one of the last points that surged. It appears that there was an increase in page search traffic. Yeah. So so and it comes right to your point there about being tied up to operational stuff because -- Yes. -- if you would just If you were just latching on to finance data, any GL data, you know, unless it was live data, whereby you had a GL code or a report metrics that said, you know, paid advertising spend, you know, and all that sort of stuff. But what it very much seems me, you can correct me if I'm wrong, that because it reference paid search traffic that it was connected to something like a a Google Analytics or another platform that was feeding it, that information as well. Is is that right? That's correct. Yeah.


So the the kind of secret sauce behind what we're building with Stirpy Plus and and CFO chat is this concept of building a live financial model? And when you think about the data, if you think about, you know, how do we build kind of a digital clone or a digital twin of our business? The GL is just gonna have kind of part of the data. It's just gonna have kind of and and and and and to be honest, the GL is just gonna have what happened in the past. Already happened. Right? It's it's not for facing but it's also missing a lot of the context for why sales increased, for why expenses went up, for why, you know, x y z And so what we do is we build this concept of a a live financial model or a living model, and we layer in operational data.


So you can connect. And in that use case, it was connected to I think zero for the GL, Google Analytics, and then some Stripe data as well. Was all pulled in to that. And when you feed that level of context, now the AI is able to say, okay, you know, based on this results in the GL, why could that have occurred? Well, let's go look at our traffic sources. Let's go look at what our conversion rates were. Let's go look at all of these things to try to get to that level to that understanding behind just surfacing that KPI. And that's really where we think this technology gets really interesting because if you're just servicing KPIs, if you're just saying, you know, what was my revenue last month? It's cool, but like a dashboard or even just kind of like, you know, a P and L from your your GL software is gonna provide that. But doing that analysis or at least guiding a financial analyst or founder towards the areas that they should probably take a double click on is what we're really excited about. And and I think that's really only possible from having this concept of a of a living model of your business where all that data is in real time, and it's all aggregated and pulled together. Is Yeah.


I I I often pause to let these these things sink in because not because I'm struggling for stuff to say. Right? Even though sometimes I do struggle for stuff to say, but it's as soon as I look at this sort of stuff, my brain just goes into, okay, and then next and then next and then next and then it starts sort of bringing in all of these. And what if you connected that, know, and what if you connected that? Could you you're getting to the point where you've you've got a complete ecosystem where you could do everything with with the chat interface. Right? She will naturally ask chat anything.


And and again, a slight side note as well. And The co founder of HubSpot has just bought the domain chat dot com for eight figures. I think -- Yes. Yeah. I I saw that. Yeah. Insane. Yeah. That that's it. I mean, I I I think about prompt prompt dot com for, like, seven figures too. This is what you get to do when you're a billionaire. Yeah. If you were. Yeah. That that's it. And that that that podcast that was actually my first million. So so that that they they and we we spoke with that on on LinkedIn chatting with and the reason that I mentioned that is because, you know, there there are some seriously stuff smart people that are a lot of monies that are putting their cards, you know, that that they could put in their eggs in this chat basket. You know?


I am spoken to some guests that don't necessarily believe the chat is the be all an end all. I think he's easy. Yeah. Don't think we should limit ourselves to just a an I'm based chat, and it would be interesting to see about how that evolves. And I know with GPT four, especially with the with the new API. It's not just text. It's image audio. I've I mean, maybe maybe you could tell me personally, I I haven't been granted access to the GPT four API yet. So I haven't been able to use any of the image recognition or any of that sort of stuff. I don't know whether whether you've had opportunity to play around with that.


But where I'm going with this is if it gets to the point where we've got lots of live data points feeding into a chat interface, a user example, you know, sturdy plus with your CFO chart. We are then into proper replacing people's territory. You know, but or or maybe not replacing, but seriously reducing the need to hire new people. Right? You know, so so it means that if you are a smaller business and you're thinking about writing, well, do I need to build a financing? Or do I need to build this? You might think twice before you invest in people because you've got technology that can aggregate aggregate a lot of the data for you.


So it's I think especially traditional roles whereby a a more traditional CFO who just expects to be the front of all knowledge whereby everybody comes to them, for answers to the question, it's not gonna be the case anymore. People can be able to self serve a little bit more. So I don't know what your impression is. It's the way that people need to start thinking. But if if I was in that position, as I kind of am because I don't think anybody's job is completely safe so safeguarded against the devolution of AI. But for me, I'm just thinking about what skills people need to to invest in and what people need to be focusing on if they are not the resource for this information anymore. Yeah. Yeah.


It's interesting because the the technology is evolving. Right? It's like, there's there's kind of, like, at at the surface level, there's a technology like ours can be used to to surface KPIs. And it's like, okay, that's great. You can, you know, surface KPIs, see, you know, chat with your data in the past and and and and kinda see what your revenues were last month under or whatnot. That's kind of level one. And then there's kind of this level two of, you know, the example that you shared of, why do you think we grew this this month? And then a bot being able to kind of say, okay. Well, here's some some feasible reasons as to why.


And then there's kind of a next level of proactively monitoring and reporting. So kind of where we're taking this too is we have the live data built in a living model, and it would be extremely cool if you could just say, Alright, CFO. Just monitor my company and you surface the insides. You tell me the high level surface data because I don't have all day to be looking at, you know, I I don't wanna be looking at my financials financials all day, nor do we have a full time hire to be doing this. And so will that then replace the need for, say, a financial analyst or a fractional CFO in for for small business, it very well could.


And so I, you know, I just actually, I was having a conversation with for an interview with Morning Brew that that just got released about this exact topic of is AI going to displace knowledge workers. And it's interesting because I think that if this technology continues to grow at the exponential pace that it's growing, the world could look so different just in in in twelve, twenty four, thirty six months. And, you know, it it's just really interesting. And I think that it's something that as knowledge workers, whether you're in in legal or you're in finance, you should be thinking about how you can position yourself to to to kind of leverage these tools and and not be replaced by them in in in ways.


I still think we're a ways away from kind of fully replacing the role of a CFO. Right? Like, I don't think that the AI is gonna you know, I think for the same reason that we were talking about where with with team building, it's still important to be goofy and be human. We still value those things. And so I don't think if I'm a company and I'm pitching a set of investors, they wanna sit down with our chatbot and and just ask the chatbot questions? Maybe they do. I don't know. But my thought is that you still want that, you know, you'd rather have a human CFO that's sitting in the room with these investors saying, hey, you know, let's discuss this over lunch.


Versus versus a chatbot that it's kinda cold, but it's it's interesting. And it it is that perception as well. Isn't it? It is And and it's also experience, etcetera. I I focused on one example there in your demo of one of the points coming back being, you know, it was potentially increasing paid whatever apps, but let's let's not forget the fact that it actually gave us quite a lot of reasons.


Yes. It's just that there is still a person on have a look at saying, do I agree with those results? You know? And what what I often do? Is I will I will now always ask chat t p t to give me a certain number of results when I'm asking it, Edison. So if I'm asking it to model a certain scenario or, you know, give me some inspiration for stuff, it's, you know, I need at least three to five because then I tend look and see whether I think that that's accurate, what it's being produced, because again, accuracy is still still a thing. I mean, it's gonna be more accurate if it is working on your dates or assuming that your data is accurate. Right. But then that at least gives you at least gives you the ability to to pick and choose to say me as a human with experience, I'm going to choose this outcome, you know, way against the data that I've got myself.


It isn't what generated and then comes to a conclusion and is very similar topic I had in Chris with Chris Riley in the last podcast, whereby we gave the example of stuff like weather. And farmers that have been farming land since the twenty years you know, twenty years worth transit. Human based gut feeling and experience can't yet be put into a into a language model? Yeah. I I I think so. I think I think you're exactly right.


Like, with our example of CFO chat coming back and saying, these, you know, here's five potential reasons. It still takes a human to say, oh, you know, let's look at all those five let's see the ones that are worth double clicking on, and then being able to do their own analysis. And then maybe, you know, if you do your own analysis and you see, okay, subscribers were up quite a bit from paid traffic sources. It's still gonna take someone analyzing that to then tell maybe the CEO or the head of marketing that, hey, we think there's a reason here to continue doubling down. So there's there's a handful of steps to take that insight and and actually act on it. And I think that humans, you know, if we can kinda help brainstorm those potential insights and then let humans do the double click, and then communicate that into kind of a business strategy.


I think that's gonna be the role that that kind of AI plays specifically in the in the finance world in in kind of the short term. But it's it's very similar to what you're saying with with with using just if you're using chat, GPT to research blog article ideas or or for any of these tasks where it'll give you, you know, five or or ten kind of potential topics for a blog article, but then you still have to be the one to say, okay. Yep. That one, I think, will resonate based on my gut feeling and my understanding of my audience. And I I think that level of comprehension is is just gonna be that nuance is gonna be something that a human is gonna be able to do better than an AI for at least, you know, kudos I might get my words in twelve months, but but but at least for the foreseeable future, there are some interesting things. I know this is, you know, the audience on the podcast likes technical stuff.


There's some interesting stuff going on with AI with with reflection, with self reflection. I don't know if you have you dug into that at all? Do you know topic by reflection, deliberate offering a little bit. I've I've I've I've kept loosely up to date with the experiments that I think Stanford did when it it did like a sim scenario where -- Yeah. -- multiple blocks were placed in a -- Yeah. -- started communicating with each other. It sounded self reflection. No. I'm not. Yeah.


So so there was a paper that came out, and I'll I'll send it to you later if we wanna throw in the show notes or something. But academic paper that came out that said, They tested I think it was GTP-four against GTP-four that self reflects. And basically, you you might have you might have experienced this yourself just with playing with with chat GPT. Sometimes, chat GPT comes back with an answer. And then you say, you know, and you notice that there's something kind of hokey about the answer that it gave you and you say, you know, think about that answer. Was that answer correct? And then remarkably, it will reflect on that answer and then fix its mistakes. And they found through this academic study that the level of accuracy is extreme improvement with self reflection versus kind of the first attempt at answering a question. And and so I think From an accuracy perspective, it'll be just be interesting to see how that starts to get implemented as well in the world of AI of of kind of these AIs before they actually spit out what they're thinking to you, having them internally self reflect on that answer, say, is that correct? And then giving you that that because it seems like the performance is extreme there's just extreme improvement there with the concept of reflection. So, yeah, it's just interesting how fast this stops moving. It you'll be it'd be interesting to see whether you actually get the the full trail of the reflection as well. But it looks like my sort says that you probably won't see it. So that that reflects probably before it actually gives the answer because -- Right. -- the end result is something that's more accurate. Right? But we will have to we've done Yeah. And, yeah, just about that outcome. Yeah.


Obviously, the from q t three to q p t four was was a big one. Yeah. And in terms of accuracy, and it's it's a bit slower, but, you know, it's passing a lot more tests. Yeah. And but it still isn't great when it comes to a hundred percent accuracy. And sometimes, you know, IT a little bit. So a LinkedIn post is three thousand characters. And I'm just just using this as Nick Dumpsters. K.


And and sometimes if I write something naturally, I get to the point where I'm doing, you know, three and a half thousand, maybe even four thousand, which is it takes a long time to think about how to cut because you don't really wanna cut any of it out. That's the big problem that we got as people as you know, which we like for being right. You know, oh, this is all we got stuck. You know, I don't I don't wanna have to save any of it. But given it to an AI, it doesn't count, you know. It's just like Yeah. Do it.


So so if you say, you know, cut this into three thousand characters without losing any of the meeting, and that's the poor because you got the -- Right. -- structures, and it will come back and it will it will rejig it. But I can't remember. I was right into a certain length It wasn't linked to it. It was something and it and it was in excess of ten thousand characters or something like that. And and I think the limit was ten thousand characters isn't and again, I'll set a particular specific example. But I actually had to tell it twice. So so it it produced me a bit of content, you know, and I gave the limit of saying, you know, it's ten thousand characters or whatever the limit was. Right. And had produced the result. It wasn't ten thousand carriages because I dropped it into a god, this it must be a really boring conversation. It's coming in at this point. Right? It's just a bit of a kick.


Anyway, so is your job to install just a parallax account for online. And it came up with four two thousand three engine whatever it is. So I went by as a chat t p p p p p p p p p p p p p p p p p p p p p p p. P. P. P. P. P. P. P. P. P. P. P. P. P. P. P. P. P. P. P. And it came back and he said, yes, I can confirm. Oh, you've been asked it again. Are you sure? You're sure? And only then did it come back saying, apologies for the mistake. It's actually four thousand six hundred and whatever. And and and I don't know whether it was just, you know, a ghost in the machine that it fought back twice.


But, yeah, it was it it kinda does make you question again, coming back to the accuracy piece. So if you don't do that validation and you you train yourself based on the fact of just asking again. Because nine times out of ten, you know, that if you ask chat GPT again, it's gonna correct yourself. Right. Wonder how many times you'll miss the second inaccuracy. You you see what I mean? So it's Yeah. Again, the mind boggles and, you know, is again, it's still a development mega quick, but, you know, it's Yeah. Developing at the pace that they are once ensuring accuracy is maintained.


We don't think it's an easy job. No. I I would agree with you. I think there has been a huge step from GPT three to four in kind of what's called, like, the adherence to system prep. Right? So when you're creating an agent. So and this might just be helpful to folks in understanding these things.


There's fundamentally a big difference between chat, GPT and GPT. Right? So if you're a business wanting to leverage this AI wanting to use open AI's large language model, you're using GPT. You're not using chat, GPT. It's not gonna interact the exact same way because chat GPT has been trained in a special way to be a q and a bot. It's it's been trained in for this specific use case. When you start using the underlying technology, GPT, you the the kind of programmer, you the the business operator, you are creating the system prompt, which is basically the system message of saying, here's what you are, chat GPT or here GPT rather. Here's here's what you are. Stay in your lane. You're you're a fractional CFO and you respond in this way and all of these things.


And GPT four has gotten so much better at that adherence than three. So it's it's getting better on the first try, but I still think you're exactly right with, like, the reflection piece of asking things the second time. It it it improves. But you gotta just think that if they're moving at this pace, some of that is gonna be fine tuned, you know, in in subsequent releases in the future as well. Yeah. Well, if if there are releases because, again, a lot of heavy weight saying, oh, pause. You know, we we can't keep up and Yes.


But that that is completely unfounded. I mean, it did make me smile because one side of me did say Yep. This is a competitive. Right? You know? Yes. We we need to stop the momentum because they are now so far ahead, you know, and and the the best quote I saw online was like, Karen, I'm a billionaire and I can't keep up, please stock six months. It was it was one of those. And and again, that's complete unfounded because I actually have no knowledge because as much as I love talking about AI and I I use it as part of my word. And I I can't credibly be able to say whether it's a good idea to to pause or not, you know, so so that's probably a conversation for another day. Right?


But but it's interesting that point that you made about the distinction between GPT and chat GPT because -- Yes. -- I think this is where a lot of the the process is either that is kind of the make or break moment is and you and I, again, before we started the the the show, we talked about companies like yours that haven't approached things in the same way as you have in the right way, basically just trying to produce stuff as quickly as possible without checking whether it's decent or not, you know, basically could be built on a a really rubbish ideas that they're putting out into the world. So Putting something, couldn't out in the world if it's rubbish, doesn't necessarily that you can make any money from it and and so on and so forth. But I think it's interesting, and and it will be the same as as as normally. Right? You know, Some companies that do have large IT resources might be able to use a GPT internally without needing to procure an external software I don't think that's gonna change. It's the same as if you did in house development of a legacy system as opposed to upgrading to it. You see what I mean? It's it's exactly You can build your own CRM system. Right? Yeah. I mean, you could. You don't need a HubSpot. Yeah. Absolutely. I've influenced it. I've sport, but my God, I wouldn't go. No. We don't But it it comes down to that you know, the the the biggest gap with the large language large language model is it doesn't understand the context unless it's trained on the context.


You know, so so your point it is just an algorithm before you say it, you know, acted this way, acted as a fractional CFO. You are a fractional CFO. Right. And I think it's gonna be interesting because this this is where all of the various offshoes, you know, and like like you guys, they're doing it very well and and some of the others is how good they are at controlling that algorithm and giving it that context. And I think that's the biggest limitation. Companies trying to do in house or have at least initially is because they haven't been at the bleeding edge of building these algorithms into platforms to they're not really gonna know what the starting point is. You know, what what does make a good instruction on how to control and guide this this large language. Yes. It's gonna be a real real learning curve. Yeah.


I think one last thing too. It's it's interesting too because in the world of finance, specifically your your data is it's ever changing. It's evolving. Every single day, there's new transactions, there's cancellations, there's new sign ups, and sort of build a system that can take all of that context in and continually evolve as well to that ever changing context, that's where we focus and that's where we kind of differentiate than a lot of some of the the other kind of proof of concepts you've seen out there potentially where it's kinda like upload a ten k from Tesla and be able to talk to that document with AI. That's great. But business finance the data is so granular and changes and evolves. So how do you continually synthesize all that data and then feed it to a to a to an AI that can kind of evolve with it as the context changes. It's it's updating forecasts. It's changing things around. And so that's where we're kind of focused with with our technology. And and that's that piece is quite a lot harder to do than than to say, you know, create a a chatbot with our support documents that are static, you know, and and and you can kinda do it one time or upload a ten k a and be able to talk to it. So having that live system that's, you know, at the end of the day, data is critically important. And and so having that good data pipeline, those live data is kind of the basis for for what the model is trained on is is is hyper important.


One other thing just worth mentioning on kind of the difference between GPT and chat GPT is that unless you specifically opt out, if you're a chat GPT user, the data the prompts that you're entering in is being fed back to open AI -- Mhmm. -- to potentially retrain future iterations of the model for GPT four point two and four point four or whatever's comes next, whatever you're feeding it could potentially be used. There's a form that you can actually basically send in that says, hey, I don't want my data to to be given back to OpenAI. But if you're a finance professional Just be very careful about what financial data from your company that might be sensitive that you're feeding to a chat. GPT because that might be going to open AI and we just we just don't want that. You know, when when you're using a a software like ours, where we're using GPT, that data for us. It lives in Microsoft in a secure data warehouse. It never gets fed back to open AI. It never goes anywhere else. Stays within a container. So I think another thing that's just important as as companies start to try to leverage these technologies is how do we do it in a secure cyber way as well from kind of a data security reason? Yeah. Totally agreed, you know, and we touched on, obviously, you know, Italy and China highlighting it as a cause of concern, you know, mainly around around data.


But Chris Chris and Martinez has been on the podcast before. And the finance automation at Kraft Heinz, where where where he's not He's he's given presentations. He absolutely loves Chat GPT. One of his posts, and I'll I'll put it in the show notes as well, relate specifically to Chat GPT security. Yes. So one of the ways and it's what I do. You know, I I I basically give it hypothetical scenarios instead of actual scenarios. Right. Yeah. So so I I don't give away company names, you know, and and all of that sort of stuff is, act as if you were, you know, working for company of this title. Producing this sort of stuff because in that way, there's there's no way to attribute data to a name. Right.


But with the third bit, and this comes back to your point. And again, I think it's only really businesses that have got a certain level like the expertise in house that could do this, but it goes back to what you say that, you know, don't use the chat interface, you know, use it in raw with Microsoft Azure, you know. But again, you've you've gotta set that up somehow, which isn't the easiest thing to I mean, he made it seem easy, but he's very good at making things seem easy. Right? So Last and so, I have two more questions if you got five minutes. Yep. Absolutely. Yep.


So the one one relates to auto g p two. Yes. Have you used it yet? I have it. Well, yeah, I've used it in the browser. I've never done anything with development, but I used agent GPT, which is like a early browser version of it. Yep. So so yeah. So so I've been playing around with that. The only reason I asked the question is because at the moment, I don't have access to the GPT four API. So at the moment, with the auto GPU, I can only use three point fives. Yeah. You know, but but my initial impression is see it, It's still tripping over itself quite a lot. And and and for people that are listening that don't know auto GPT or agent GPT compared to normal are more autonomous. So we're moving into autonomous AI whereby instead of giving it specific prompts and it's a q and a, you give it an end goal and it will self prompt. Right.


So so the example that I gave today was create a financial presentation based on being the CFO of a hospitality industry wanting to do a bench fight report or something like that, you know. Then find industry KPIs, find industry benchmarks, and then create be some present presentation slides. So you'd be able to do that in charge EPC. You wouldn't be able to do piece because, of course, it's not connected to the Internet. Yeah. I mean, there's there's a way of connecting it to the Internet. Right? But but the point there is one prompt could potentially facilitate quite a lot of rounds of q and eight.


And it did it pretty well, you know, so until the incident, it found the KPIs that were hospitality specific that was great. You know, and it came back with some some percentage benchmark even though you'd have to double check them again, that's the accuracy piece. And it did a reasonably good job of producing the presentation slides or at least the the high level stuff, but it wasn't as accurate as GPT four. Hence why I'm really keen to try it with the GPT a GPT four API. Even though the GPT four API, as you probably know, is more expensive to run than three point five, you know, so significantly. So Significantly. Absolutely.


So, yeah, don't just let it run continuously because otherwise you might and and I think, you know, with with what I've I've only spent, like, ten p you know, with -- Yeah. -- chat. TBT three point five, you know, in playing around with it this week, but it's just for people to be aware you do need an open AI page for API and and can quickly start running up some cost steps. But, yeah, I think it's it's gonna be a while before it can really do the autonomously. Think there's some really smart people that are getting better and better with it, but I think for me at least initially, my initial impression is is is it a step too far and I wouldn't necessarily go to that.


But this is just using chat t p t at the moment. I don't know whether that's been your experience. Yeah. It's interesting. Right? Especially, like, in the Twitter world and LinkedIn to everybody's if I have to see another tweet or another LinkedIn post that starts off with, chat g t p is old news. Auto GTP is the new thing. And I'm like, come on. Like, if they're all start the same, which is hilarious because I'm also thinking in the back of my mind. I'm like, are all these people just using AI to create the same tweets? Yeah. Yeah. That's that's that's a that's a link. I can send you a link because they are. I know the blog. AI. Anyway. Yeah. So that's that's one kind of concern there. But, yeah, it's interesting. I I think it's a it's a cool implementation of the technology. It is scary to a degree because it is connected to the Internet.


It is semi autonomous in ways where it can be, you know, there's there's some bad use cases out there of of you know, someone that they're being facetious and they're pointing this out to the general public to make kind of a stance, but they created that that chaos GPT, which is basically auto GPT, but with the the intent of kind of like, you know, destroying the world, They're doing it. I'm not saying that. Yeah. Yeah. So if you if you search chaos, GPT, there's there's tons of of stuff on it. And the whole bunch of those. Yeah. You're right. I know. Right.


But but it it it's to make kind of more of like a political statement or a point as to we should be careful about what we're doing with these technologies because in the wrong hands, they could be used for harm. And so as far as kinda like the pause situation and and and that controversy, you know, I don't think it's it's it's necessarily the right thing to do to to pause AI, but and and kind of some of the evolution of it. But I I do think we need to put heavier emphasis and and put strong emphasis on, like, the safety and the alignment of these models in in the way that we're interacting with them because we could, you know, people don't understand exponential growth and and when you start to truly understand exponential growth and when you think about the capabilities of today and what's capable tomorrow, it's crazy. Right? Like, I mean, it's it's so fast. And and so I think we do need to be careful. And so I think it's finding some sort of approach of of kind of moving forward, but also ramping up I don't know if it's some sort of governing body or some sort of way to regulate in without kind of hindering technology, you know, and and stopping us from moving forward. I know, like, on the online podcast, I know they were talking about like an FDA, essentially kind of looking at the FDA, which is the Food Drug Administration here in the United States. I know the UK or EU has a similar thing But taking a model like that and potentially applying it to AI could be used interesting, but I don't know. I think that the speed at which governments run and the speed at which AI runs are on completely different playing fields. And I think it's gonna be a difficult challenge for us to to handle as as humans. Yeah.


I mean, look looking at things on again, look at things on the positive side because, of course, that, you know, there's always two ways of looking at things is. Yep. I do often make myself smile, but there's we are almost getting some of that sort of iRobot type. Yeah. WiFi movie, aren't we? And and and I think to myself, like, you know, even if the world were to end, it's like, I kind of lived the sci fi movie. Do you know what I mean? Yeah. We we were alive at a time where basically technology f stuff happened. Do you know what I mean? So that's that's how I look at things positive. And that's even with the die that's looking at the die consequences and looking at them positively, but On the other side of things, as as you guys approve it, I really hoped is that if there is more governance and you know, as people are able to use these tools in a safe way, that people can become so entrepreneurs, people can build their own businesses with, you know, quickly and start putting their value into the world.


That's I really do hope. And and and I'm hoping to hopefully, I can talk more about this fun day is that I'm hoping that my kids, ten months and two and a half years of age, never have to go through the decision that's as, you know, to live I need to make x amount of money. I I don't want them to have to think about getting onto the corporate ladder to sustain a way of life. My my wish for them is that they can decide what they're passionate about. Yeah. And they can use technology to do it. And make money from Yeah. And and I think without needing to know how to code, you know, without needing to have experience with all the Visa for this gonna come down to how well they can use a combination of tools essentially. It's the same for everybody. But that's my hope anyway. That's that's my optimistic side saying that the technology is there.


I quite like everybody not to be subject to the golden handcuffs anymore and being wedded to a job that they hate. Right. I'd like them to be able to take this technology and say that I'm gonna follow my passions, you know, whole wave of digital nomads, you know, from the beach. Yeah. A hundred percent. And I think that's that's that's part of our mission too is that, you know, we work with a lot of small businesses. And I always I always make the joke that Founders wear a lot of hats. Right? And finance is not the sexiest hat in a founder's closet. Okay.


And so, you know, if we can use AI to to not I'm not even gonna sugarcoat it. It's, like, I think it's goofy when when especially, like, larger corporations say, like, you know, we're gonna make payroll fun. It's like payroll is never fun. But, like, if we can if we could use technology like GPT, an AI to to make you understand, like, wow. Okay. Why is this important? Why, you know, how can we use finance to our advantage? How can I now play on the same level playing field as a company that has ten times the amount of resources us? That has a fully dedicated FP and A team and a CFO and all that. And I'm just a single, you know, founder, that's the goal. Is to level that playing field so that folks like that can compete. And and and so I'm optimistic as well. I think there's so many strong positive use cases for the use of AI to to to kind of better the lives and and and of of the many and and and create economic opportunities.


For folks that maybe wouldn't have had it, you know, had they not had access to these sort of tools? Hundred percent. Hundred percent. Fine. So so the the last question I always ask is and and we've spoke spoken about the systems that you used. So we spoke about Slack, Trello, chat, TPT, all of that sort of stuff.


The last question I know it does is irrespective of maybe the tools that use of business, even though it could be a business tool, is the an piece of software. It could even be a gadget. I see if it's a diffuser that you've got in the background. It could be something like that, you know. That's what what Please attack or gadget it. Could you not live without? Yeah. So well, you know, we have already this is basically, this podcast has been a a promotion for Loom. I can't mention Loom anymore. That's a little bit mess up there. It's where I'm not affiliated.


But I'll give you I'll give you one that that folks probably haven't heard of. It's actually it's old fashioned software that you actually have to download. So if you're a Mac user -- Well, I I know. -- I know. Right? Like desktop software, what? It's completely free. It's it's made by kinda some some indie developers. It's called Licecap, l I c e CAP, and it's just a simple tool for Mac users to record your screen and make gifts. And I'm constantly like I'm going through all of our documentation for just help articles and stuff and rather than using screenshots. I think gifts are so much better. And and so I used LifeCap to just simply record a gift go down the desktop, and then you can kind of, you know, upload it to wherever you're at. Simple little tool, completely free, indie developers, old school, but I use a lot.


Yeah. We've we've I I loved that. We've never had that one before. I figured that one that wouldn't be mentioned. Yeah. And then googly, it doesn't appear that there is a Windows version as well. Yeah. Okay.


On a on a soft comic, you can get it. I don't know whether it works or not. Most people can have a very good And I'm I'm sure there's tons of other tools out there that do the same thing, but I don't know. I found this a lot of ideas. This is one that works. And yeah. Yeah. Perfect. No. That that's great, John. I I appreciate that we've run over time a little bit, but I I really enjoyed that. I think it was really good good conversation. So really appreciate coming on as I say, may maybe in in twelve months time when you you're a massive organization and, you know, and it hit with your targets. We can have a conversation. We can talk about how how you did it. And it might be a conversation that says, well, I don't do anything anymore because they are. I does all. We'll have to see it. No. We we agreed. So I'm John. It's been amazing. Yeah. Absolutely, Adam. Thank thank you so much. Really enjoyed it. Thanks so much. And, yeah, have a great rest of your day. Yeah. Well, thanks so too.

Shownotes created by


tech for finance

©2022 by Adam Shilton. Privacy Policy - Terms of Use

©2022 by Adam Shilton. Privacy Policy - Terms of Use