How ChatGPT is Changing Custom Coding in HubSpot (and Beyond) With Connor Jeffers

Sidekick Strategies Expert Interviews

• Show Transcription

  George B. Thomas (00:01.796)
  Welcome back Hub Heroes. I'm super excited. Another episode of the Hub Heroes Sidekick Strategies. That's right, I'm your boy, George B. Thomas, and today we're talking about chat GPT. By the way, as a guy who has the initials G-B-T, do you know how hard it is for me to say the word chat?
  
  GPT, I literally have to pause anyway, and coding HubSpot like a pro. The good news is it's not just me. I'm here with Connor Jeffers, the man, the myth, the legend. Connor, how the heck are you doing today?
  
  Connor (00:34.25)
  I'm good, man. I feel like you should launch chat, chat GBT. It's a great one. You can bring all the personality.
  
  George B. Thomas (00:39.133)
  Yeah.
  
  George B. Thomas (00:43.444)
  Yeah, I tried to get the domain and I will not tell you how much they wanted to charge for that domain. I was quickly like, oh, I was going to buy this for an easy like, here's my socials or here's how you could talk to me. I'm not going to spend sixty five thousand dollars for people to be able to talk to me. Yeah, it was a well, it's less than what you thought, but it was a lot more than I thought because.
  
  Connor (00:53.226)
  Nah, it's good, I'm good.
  
  Connor (00:57.856)
  Yeah, yeah, yeah.
  
  Connor (01:02.062)
  It's actually less than I would have thought, to be honest with you. That's not- that's less than I would have thought.
  
  Connor (01:10.684)
  That's a lot more than we were trying to do, yeah.
  
  George B. Thomas (01:12.572)
  I was looking for the $9.99 for the first month or so. But anyway, Connor Jeffers is founder and CEO of Aptitude 8, an elite tiered HubSpot technical consulting firm, and happily, my boy Max Cohen, I just gotta throw that out there, and App Studio building and DAX, products on top of the HubSpot platform. Prior to A8, Connor led growth teams building venture-backed companies on top of Salesforce.
  
  Dirty word in this episode, but we'll take it. And operated a consulting firm on the Salesforce ecosystem. He has over a decade, ladies and gentlemen, of experience in system orchestration, digital transformation, revenue operations, and growth. Now, without further ado, let's get into the good stuff. Connor, we are talking about nerdy stuff today. And one of the things that I wanna do is be able to just level set.
  
  for some of the people who are tuning into this, either in the audio version or the video version, can you just take a couple minutes to explain what Chat GPT is and how it can be used as a coding assistant along with how it may pertain in that manner to HubSpot?
  
  Connor (02:25.91)
  Well, you have to understand, George, I'm just a large language model. And so I can't answer that question. Uh, no, I kid. Uh, so, so when we talk about Chachi BT, uh, all the rage from open AI, uh, I feel like there's a bajillion content all over the place, uh, but essentially, um, it's a tool that allows you to interface with some of the, the newest and largest and most robust.
  
  uh, natural language models, uh, via a chat interface. Uh, and most recently and most excitingly for, for me and my team has been, uh, the new code interpreter, uh, and it can do all sorts of wild and crazy things. Um, but some people are talking a lot about AI and chat GBT for, uh, content and marketing stuff, and I think that is, uh, cool. Uh, but the things that really get us super excited is some of the stuff that you can do with.
  
  technology and development and code. And I think from my perspective, the primary value is really bringing down the barrier for people to be able to interface with things that previously you'd really have to understand how to write code, build code, understand APIs, do a lot of this kind of stuff and make it a lot easier for people to be able to access and leverage, which I think is sort of the democratizing force of all of the AI tools in general.
  
  George B. Thomas (03:38.264)
  I love this. I do want to go off the beaten path almost immediately because you said code interpreter. Can you just dive a little bit deeper into that and what that is, like how we would be part of that or use it or like all of those kinds of pieces? Cause I do agree with you. The major conversation is chat GPT for AI assisted writing or research or marketing things. Let's dive into what that code.
  
  Connor (03:53.613)
  Yeah.
  
  George B. Thomas (04:06.656)
  What did you say again? Code interpreter, right? Yeah, let's talk about that.
  
  Connor (04:08.43)
  Code interpreter is what they've called it. I honestly haven't reflected on the term until you've said this. I'm like, it is a weird term. Essentially, so the code interpreter is a beta functionality that's available in GPT-4. So if you are a paid chat GPT user, which is $20 a month, which I think dollar per value is probably one of the highest dollar per value things you can do. It's instantly one thing, and you're like, whoa. Yeah.
  
  George B. Thomas (04:31.772)
  I got my 20 bucks. I- First hour! First hour! 20 bucks! Oh, paid.
  
  Connor (04:37.938)
  Yeah, check it out. It's incredible. But the code interpreter is a new beta on GPT-4. So the code interpreter does kind of two things with ChatGPT. The first one is that it lets it understand code. So if you give it code and you ask for it to edit that code, try to model that code, change the language, do something that's effective, it'll understand what that code is. It will also write code to solve problems, which is honestly really wild.
  
  It's spooky to watch when you kind of like see it actually happen. But you can ask it to do things. You can upload a file and then it will use code. It supports Python today. It'll basically use Python to do the things you ask it to do. And we can talk about what all of those things are. But basically think about it as if it's a think about all of the magic that if you've ever used any of the chat, GPT-3 or 3.5 stuff and
  
  It's that and then also combine it with like a decent software engineer who really knows what they're doing and it's just like ridiculously fast. And you can start to get a picture.
  
  George B. Thomas (05:45.468)
  Yeah, and I want to dive into that because I think you're headed in the direction that we need to unlock people's brain to. And that is what are like some of the main benefits of using chat GPT for custom coding. Again, it could just be custom coding or custom coding in HubSpot compared to that traditional coding method that somebody might be used to.
  
  Connor (06:06.442)
  Yeah, so I'll give you like a really, we can start on the basics, then we can start to kind of like go up. And there's kind of two different things that I'll talk about here, and I'll do my absolute best to delineate between the two, because they're very different. One is interfacing with GBD4 in the code interpreter through chat GBT, like through the interface as a person working with it. The second one is what you can start to do by.
  
  developing and building on top and actually piping information into the GPT models more of like an API layer and then getting things back out. And then kind of the third is like, what does that look like into the future? But let's start at sort of like v1, which anyone can do. You do not need to know development. You do not need to know how to write code. Python might just be a snake to you. That's OK. You can use GPT-4 in the code interpreter. So I'll give you a very real example of something like I did this week. And like.
  
  honestly blew my mind, which was I were working on fundraising for Happily. We're pulling out our subscription data from HubSpot and we're trying to get like, uh, reports and charting. And one of the things I'm trying to figure out is like, what gives the best looking revenue curve that I can slap onto an investment deck, right? And so it was like, do I want this monthly? Do I want this quarterly? Like, how do I want this to look?
  
  And as I was working on it with my designer, she was struggling with, hey, I'm trying to get this chart, but like it's showing MRR and I want to roll it up and I want to do ARR or whatever. So she's basically, I have this Excel sheet and I'm trying to do like VLOOKUPs and pivot tables and trying to get it to look a certain way. And so we spend like 20 minutes talking about this last week and on Monday, I had a meeting cancellation. She's out sick. I'm like, I really need this done. So I downloaded the file. I go to chat GPT and GPT-4.
  
  When you flip on the code interpreter, there's a little plus button. You can upload a file. So you can upload the CSV. And I just put in all of our transaction information. And I was like, hey, here's this. What I want is a table. And I want a table that shows me the rolling MRR for every month. And in a second, it's like, OK, cool. Here's that table. And it'll show, here's all the code. There's a little working section. And then you can click it. It'll expand all the code it's writing. And what's wild when you watch this is it self-corrects.
  
  Connor (08:20.862)
  its own assumption. So it goes and runs on it and it says, hey, I tried to do this, but it looks like the dates are actually all text and they're not all the same format. So what I'm going to do is I'm going to normalize those all to be the same date format right now. And then it just does it and it's like, cool, I did that. Now I'm going to combine everything by months. And I didn't give it, here are all the steps we need to do to get this to the right place. It's interpreting, what is it I want to achieve? What does it need to do in order to get from where it's at to that end destination? And let me compile all of that.
  
  So it gives me the table and I'm like, cool, can you give that to me as a chart? And it gives me the chart. I say, cool, instead of looking at monthly, can you group all of that quarterly? And it updates the chart. And then I gave it the brand guidelines with the hex codes of the color palette for our deck. And I was like, can you format this chart in a way that is copy pastable into my slides? And then it spat that back out. And I honestly, like my experience, this is wild. This is something that I would have done with other people. I would have taken.
  
  like at least several hours if not a couple of days between Slack messages and everything else. And that's just like, and data level stuff, whether you're, here's all the data, what's the trend.
  
  A really solid HubSpot example that we used recently is like, we have an open text field on signup for, if you install any happily app, we ask you, how did you hear about us? And what we were doing is taking all of those data points and you can load them all up to chat GPT, and you can say, convert all of these into pick list values and just figure out what pick list makes sense between the most common things with everything in this.
  
  data set and just like summarize all of it. And it'll convert all of that into, here's a bunch of dropdown values that we would suggest based off of everything you just uploaded in this file. And it's doing all of that by reading the Excel file and it can read PDFs. You can start getting into really crazy stuff. But that's kind of the code interpreter is.
  
  Connor (10:10.854)
  you can interface with it in those kinds of ways. And data is, I think, a really accessible way for people to be able to come in and have a rough understanding. Quite literally, if you were ever in a position and you're struggling with an Excel formula, just stop and literally go to GPT-4, put in your file, and just be like, I want this. And it'll just give it to you. It'll blow your mind.
  
  George B. Thomas (10:33.624)
  I mean, I'm sitting here listening to the great examples that you gave and I'm like, oh my God, my mind is blown. And I'll be honest with you, and the people who are listening or watching this, I'm like, well, what does that mean if I wanted to like code a calculator? Or what does it mean if I wanted to create this widget? Or how could I do this, power my businesses in this other way in minutes or like seconds? And the thing that really blew my mind is I was like, wait.
  
  Wait, this thing is doing something that I couldn't do in high school. It's showing its work. Like what is happening right now? And it's amazing to me. Now here's the thing. And I don't know if I wanna call this best practices yet, or if I just wanna call it Connor's brain.
  
  Connor (11:07.594)
  Yeah, it's wild.
  
  George B. Thomas (11:17.996)
  But I know my next question is when you went in there, you were thinking about it in a certain way. You knew you wanted to do it in a certain way. You wanted to talk to the AI in a certain way. So like marketers that are watching this, developers that are watching this or listening to it, and they're new to using AI powered coding assistants or what we're talking about today, what are some key considerations, best practices, how do they unpack Conner?
  
  like what should they keep in mind when integrating chat GPT into their workflow?
  
  Connor (11:54.794)
  Yeah, so I think the biggest thing with GBT that people have to remember, and I think the word AI has like, created bad assumptions with this. All chat GBT really does is it's honestly, it's a guessing machine. That's all it does. It basically ingests all of the known information in the world and you ask it a question, it reads your question and then it guesses, like quite literally a character by character and word by word, like what would a good answer to this question be?
  
  And most of the time, that's like a pretty good answer. But some of the time, it's not.
  
  And some of the time it's wildly wrong. And I think one of the things that people, when they try to dunk on GPT is they're like, oh, I asked you this question, it gave me a wrong answer. And I'm like, it doesn't know anything. Like it has no knowledge. Google is like, let me go and find all of the information and try to give you the most accurate one. ChatGPT just gives you the most likely one, which is why when you try, when there's like examples of people that come in and try to do it with like legal or medical, it'll just be like, oh, like these are legitimate sounding cases. Like this seems pretty likely.
  
  case law I made up that you should use because it doesn't know what's right. And I think what's really interesting about development and code and some of the things there is it's really not a subject to that concern. And so you can start doing really wild things especially like I would recommend enabling like the web browsing plugin and you can... some of the things that we've done is like going to a... you can go to API docs and you can paste in
  
  What I want is I want a piece of code that will create an order in NetSuite. And like, here's the link to the orders API for NetSuite. And what I wanna do is I want to put in this company name field and HubSpot. I wanna, and you can have multiple links. You can be a go read the HubSpot API docs, go read the NetSuite API docs. And what I want is something that will create an order in NetSuite and it'll set the.
  
  Connor (13:54.014)
  company name as the name of the company, but we'll look it up to make sure we aren't creating a duplicate. And like, here are all the things that I want. And in some cases, you might get something spat out that is like copy pasteable into a coded action and works. And we've done that, like no editing, nothing like it, it works. We've done it. But more than likely, you're going to get something that's like most of the way there. And it's going to be really useful. Another way that we've used it recently.
  
  as our development teams, one of the hardest things about building stuff custom, especially when you're writing code, is so much of how code works is in the eye of the person who built it. And maybe they left you really good comments, but they probably didn't, especially if it's broken. And so it's really hard to be like, what does this thing do? And you can copy code. This is the thing where stuff gets insane. You can do this with a really good example. If you have a complex Excel sheet,
  
  Like you can do this also. But if you take code, you drop it into GPT-4, and you're like, what does this do? It will tell you. It will be like, oh, this is a function. Here's what it's designed for. It powers a web app. Here's probably how somebody interfaces with it. And it can understand what that code is. And you could even say, cool, can you edit that code? So another, here's this code. Here's what it does. And what I want to do is, in addition to the function it already does, I also want it to create an order in NetSuite.
  
  Can you give me a version of this code that does that? And it can iterate that and kick it back out. And that type of workflow is already being integrated into VS Code, which is a popular compiler. And that's actually just like, if you have developers, like go have them use VS Code and have the co-pilot on it. And it's that experience, but inside of where they actually write code and it gives you insane superpowers.
  
  George B. Thomas (15:43.728)
  First of all, holy ish. I'm just gonna, I don't even know where to go. Like after that, like you had me at API documentation cause I can't tell you how many times I've gone to look at API docs and as a somewhat front facing HTML, CSS, a little jQuery like develop at mind, I look at the API documents and I start freaking sweating. So like to be able to actually just, hey, here's a URL and do this.
  
  Connor (15:45.646)
  Hahaha
  
  George B. Thomas (16:12.508)
  You had my mind was blown also I too have met that developer that does not know how to leave Comments on their freaking code. We see you stop it
  
  Connor (16:19.318)
  Hehehehe
  
  Connor (16:24.81)
  You could ask it to comment the code. That's the other thing that's insane. And it'll do it even now. You can do it in VS Code. It'll comment your code for you. Just, hey, what does this do? Can you add comments to this? And you should check them. But anybody who's managed anything, revisions and edits and new cycles are way easier.
  
  George B. Thomas (16:33.499)
  Oh my gosh.
  
  Connor (16:44.342)
  than V1. And I would sort of, I think, I think GVD is amazing at both of those workflows today. I think that if you trust it to just do everything and give me something useful, you'll be constantly frustrated. But if you look at it as, and I think that people who are really effective at managing others, just get insane superpowers because you can be like, hey, what I want is something. Here's kind of what it's like. Can you give me a version of that? And you're like, oh, this is good. Make these changes. Do these things.
  
  Great, now I'm gonna take it out of here and I'm gonna go work with it myself and I'm gonna build something that I'm looking for and do something differently there.
  
  George B. Thomas (17:17.856)
  This is so good. By the way, I also did think about probably three or four developers where I'm like, and I need to share this episode with these developers right now. Speaking of which, I want to talk about what we might call collaboration or streamlining the processes of these businesses, agencies, people, developers, marketers. And so in your mind, how does chat GPT enhance or streamline collaboration and communication between marketers and developers when working in?
  
  Connor (17:26.082)
  Yeah
  
  George B. Thomas (17:47.83)
  custom coding projects, preferably with like their HubSpot and what they're trying to do or what they're trying to create as an organization.
  
  Connor (17:56.19)
  Yeah, I mean, I think what's really interesting is if you're not a very technical person, right? And I think that most people, like I think it's like the old product manager skill set is kind of like what everyone in tech skill set is at this point, which is like, I understand how you could build stuff, but I couldn't write the code for it. And then you have a lot of difficulty of like, how do you get like scoping?
  
  like chat, it's amazing. Like you can come in and be like, okay, what we want to do is we have this project and we're going to create these things in this way and we're going to do these things in that way and we need to be able to parse this data and deduplicate it and we want to put it over here and you can give all this information to gbt like in scratch notes like go to a sales call grab your sale or like even better take the transcription like i'm not even joking like go to gbt4 here's my whole transcription like write a project brief that I could give to a developer that articulates like what we have to do here
  
  and it'll just give you something that's probably most of the way there every single time. And so we're using this a lot for scoping. We're using this a lot for project planning.
  
  were another amazing use case if you do. We actually have a project we're working on right now to make this a little bit easier that I can talk about. But one is example data. If you want tables and tables of sample data that you want to be able to just upload to HubSpot, just tell it, hey, I'm making a HubSpot instance. I'm demoing movie theaters. And I need a bunch of theaters. And I need showtimes. And I need films. And what should the data model be? It'll just tell you what the data model should be. And then give me tables that are example data
  
  So a project we're working on right now is a data model builder where you could tell it what type of organization are you and what, and it will suggest what objects you should use. But then you could also say, oh, I actually really need like a concessions object or whatever. It'll say great, that sounds good. And then it'll say, are there any important fields on any of those objects that like you really want to make sure are included? You can maybe tell it what fields and it'll just guess what fields you want. And then it'll just spin up a developer portal that has all of those objects.
  
  Connor (20:00.64)
  of those fields, all those properties, and then all of the sample data that's like pre-configured, and then you could just like go show it. And I think the craziest thing about that statement, right, that's not like a crazy advance, like you need to be like an AI developer, like that's not complicated. Like it's like a form and kick data to GPT-4 and then like use a put in the Hubs.API and like load it all up.
  
  And that's it. And you can build things like that. And so I think that's where you can start taking that conceptually and sort of extending it to, what does the AI powered CRM of the future look like? And it's, I think it gets really crazy really fast.
  
  George B. Thomas (20:45.22)
  Somebody needs to call 9-1-1. I need help. My brain is melting right now. Like I'm listening to you talk about this and I'm literally like I have had an idea for a while as a guy who loves to do HubSpot tutorials and would love things to be populated with data so people see what a real-world scenario would look like in a fake world. Oh
  
  Okay, let's continue on. I don't know viewers, listeners, if your brain is freaking out. I hope it is, because that means that you're on the precipice of change and doing something different. Here's the thing. If you are listening to this and you are ready to change and you want to start testing this, I want to make sure that we're paying attention to something that is highly important. And so, Connor.
  
  What are some of the potential challenges or limitations that users may face when utilizing chat GPT for coding, these custom codes, HubSpot coding, like A, limitations, and are they just fine, like, nope, it's a limitation, or have you found that there's ways that they can overcome some of those limitations as well?
  
  Connor (22:00.074)
  Yeah, so as of this moment, right, like code interpreter has Python, JavaScript and Ruby. So if you wanna do stuff outside of those languages.
  
  It won't do them like it doesn't understand them and it and like the time of this gets released like that might not be accurate It's like moving pretty fast. So right now you sort of limitations on coding and development So there are some really cool AI projects people are doing like if you're a marketer I would take a look at like some of the stuff that people are doing with Using AI to generate entire websites or like pasting a URL There's some that you can grab the URL of a website and just like rebuild that whole site in your own your own like domain It's pretty insane
  
  George B. Thomas (22:38.392)
  brand and domain, oh my gosh.
  
  Connor (22:39.332)
  So there's, it's crazy, but there are some limitations there, right? And so...
  
  One is coding languages. The second one is we go back to that guessing machine, right? And I think that that's the most helpful way to think about what these AI models are, is they're just really sophisticated guessing machines. And that means that it's going to be wrong sometimes. And you have to, you want to check it, you want to validate it, you want to test it. If you put something in there, you grab it, you paste it, you move on, you just think it's going to be right, you're going to be let down every single time. And so know that it's really good at giving you first drafts and it's really good at giving you edits.
  
  used it recently for like, we don't know why this isn't working. I can't tell you it's driving me insane. Copy like debugging, copy all of that, paste it in. Why isn't this compiling? It'll be like, Oh, like you have this function up here and it's recursive and it shouldn't do that. And you're like, cool. Do you have a suggestion for how I can fix that? And they'll be like, yeah, this. And you're like, cool. And it runs and it works. You're like, amazing. I'm going to move on with my life. But I think that if you're doing anything with, with technology, I think especially like.
  
  revopsy hacker folks, and if you are integrating stuff, automating stuff anywhere that you're doing these kinds of things, take like a beat and go and try to interface a little bit with GPT and see what it gives you back. And sometimes you're impressed, sometimes you're disappointed.
  
  I tried to get it today to summarize, like I ran a rev ops bootcamp with HubSpot Academy. I uploaded the PDF of the rev ops bootcamp and was like, can you write me a LinkedIn post about this whole thing? And it's like, I can't read this file. I don't know what this is. And it's so funny, like you get so jaded with technology, so you're like, ah, useless. I feel like that's like an outlandish thing to come in from. But I do know if I had the transcription and I had the recording, it could totally do it. And so I think that
  
  George B. Thomas (24:12.918)
  Hahaha!
  
  George B. Thomas (24:22.604)
  Yeah.
  
  Connor (24:24.838)
  Areas that are really accessible to anybody are anything with like data files, Excel sheets, like never make a pivot table, never do data analysis, never try to build charts, like total waste of your time. Like take any Excel file, put in a GPT-4 and just be like, I want this and it will just give you that. And that's incredibly valuable. Anything with...
  
  First drafts of code, reading API docs, is this possible? Is there an endpoint? Something we do a lot for one of our products at Happily is Zebra. It's a Stripe HubSpot connection application. And frustratingly, for every single deal, there's like, oh, but we want to do this one weird thing in Stripe. And it used to be like, oh, OK, let me write that down. And let me send it to the engineering team. And they'll go look at the Stripe API docs. And they'll come back. And then we'll see if it's possible. And something we're doing a lot right now is like, hey, here's the Stripe API docs.
  
  Can I insert a custom item into a memo? And it'll be like, oh yeah, you can. Here's the request to do that. And I'm like, cool. Can you give me like, where do I validate that? And I do a double check, but that's like a couple of seconds. And I think that type of workflow, whether you're doing sales engineering or solutioning or building stuff, incredibly accelerating.
  
  George B. Thomas (25:41.748)
  Yeah, it's so interesting. I've...
  
  I know the power on one side of this and I hear now the power on this other side that we're talking about today. Because it's funny you mentioned in seconds earlier today Connor I was talking to a potential client and they said to me, content ideas are hard. And I had them rattle off what are your services, who are you providing them to, and a couple of other things and I was literally typing it into chat GPT as they were saying this and then I asked it to spit me out 20 blog article ideas.
  
  read off the ideas to that person in the moment, in real time, and their face was like, oh my God, like you are a content God, like where do I sign? And it's just so funny because, and we're starting to hear this more and more, it's not AI that will replace your job, it's people who learn how to use AI that will, right? And so we're hearing that on the content side, I'm literally hearing that in my brain as I'm talking to you about this.
  
  Connor (26:19.178)
  And they're like, pfft, you're a genius.
  
  George B. Thomas (26:43.848)
  What I want to do is I want to make sure that we give people some actionable steps here. So when you think about this conversation that we've had, all the things that are possible, the mindsets that marketers and developers should grab as they're moving forward, can you provide some actionable strategies or tactics that these marketers and developers can implement to leverage chat GPT effectively?
  
  and efficiently in their coding process, in their day to day, if you will. Where does your mind go with that?
  
  Connor (27:15.806)
  Yeah, I mean, I think the first thing is go pay 20 bucks. It'll be the best 20 bucks you ever spent on upgrading to the GPT-4 engine. I think that that's probably a good step one. I think step two is like, play with it, try it. See what you can get it to do for you. Like honestly, we are so early. And like, I think even when you go...
  
  And what I did, like, right, I started, I went to YouTube and I was like, what are some cool things that I can do with this right now? And like all the videos I saw, I was like, these are not very interesting. And then I think just start to troubleshoot and experiment and see what you can kick in there and see what you can get back out. And I would suggest for most things you do, if any, and I'm sure that there's probably like some great word in German for this that I just don't know. But like, whatever the moment is when...
  
  You're trying to do a thing and you have like a very clear and articulate idea of what that thing is, but you don't know the steps in order to get to that next point. Stop what you're doing, go to GPT-4 and just be like, hey, here's what I'm trying to do and here's the output I want. And if you are working with a coded file,
  
  give it the coded file. If you're working with a data table, give it the data table. And you will be amazed by what you get out the other side. And I think that the more you do that and the more exposure you get, the more you start to be like, really like, oh, cool. I understand how I can do this. I understand how this can tie into my workflow and what I'm doing. And like crazy things that we've done. I mean, the talk I'm giving in inbound this year, I didn't give the title of. I gave all of the content to GPT.
  
  information about who I am and I asked it to fill out like 30 different submissions with all the table information of the inbound talk template and it gave it back to me and I was like this looks great give it to me in a table and I'm thinking about this right now I haven't thought about this since I told this. Then I went and copied all of that and put it into an Excel sheet and like formatted it and instead I could have just been like hey here's an Excel sheet like can you fill that out for me and just give me the file and now it could just give me the file but
  
  Connor (29:25.818)
  I'm stuck. Like how can I do this? This is really frustrating or what which items are in this sheet? But not this year this table any of those things where you're like, how do I do this? Like let me go Google it Let me look up a V look up like that's a very tangible example for a non developer anybody who writes code
  
  whether it's for a living or like very much not for a living, you have that experience all the time, which is like, how can I get this thing to do the thing I want it to do? And if you can articulate what is the thing you wanna do, what does the outcome look like? It can usually help you get there in really tangible ways.
  
  George B. Thomas (29:58.452)
  It's funny because when I hear you talk, Connor, I literally envision it's like your personal assistant. It's your little buddy on your shoulder that when you get stuck, you'll be like, hey.
  
  What should I do here? And then like just thinking of it that way and embracing it. And the other thing I heard in that section, by the way, it might be a rewind point of the interview, is taking time to educate yourself. That's a great step of like, go watch some YouTube videos, go find a course, go like look at what you and then play around with what you learn inside of the thing. It's so good. Now, I'd be remissed if we did this interview.
  
  And we didn't talk about the big boy and the big girl stuff, the real important things. And what I mean by that is, Connor, should people be aware of any privacy or security security, security concerns that like that the user should be aware of when it comes to chat GPT for custom coding and being a personal assistant? Like talk us through your brain on privacy, security.
  
  the sometimes boring stuff but important stuff that we need to think about at our organizations.
  
  Connor (31:13.386)
  Yeah, so the first thing I'd say as an initial caveat is I would not describe myself as an expert or evangelist for either of those topics. What I will say is anything that you put into ChatGBT, it knows now, and it knows now everywhere.
  
  for everybody. And that doesn't mean that, hey, I uploaded this file and it had emails in it, and now ChatGBT is just going to go and someone can say, hey, what emails do you have? And it'll give it to them. But they go into the repository that is the universe. And so if you have any level of data security that you're aware of, either for your own organization or inside of those rules, I would be hesitant around some of those. And I think that's the big, sort of one of the big pushes with
  
  to our mesh's chat spot project, right? It's like, how can we give that power but we can isolate the data set to like, it's secure and stored only in HubSpot and some of those other pieces. Still early days on that front, but I think that there are so much that you can do. If you were working with sensitive data, probably don't upload it to chat GBT is probably like a good idea. And I would sort of isolate it to things. And for myself, with the exception of like customer data or PII or anything that we kind of like manage,
  
  I'm not a super private guy. And so I like the robots knowing everything about me. I think it helps them serve me better. But if that's a concern for you, and I think especially if you're working with data that you have from other people, I'd be, I wouldn't be going and punching in credit card numbers or HIPAA data into ChatGBT. I think that'd be a bad idea.
  
  George B. Thomas (32:53.864)
  Yeah, and the caveat is if you're listening or watching this, seek legal advice on those things for your organization, without a doubt. I have a surprise question for you that wasn't in our outline, Connor, and that is if you could boil this down to hashtag one thing. What is the one thing that you hope people take away from today's episode?
  
  Connor (33:19.062)
  I should go and play with chat.jpt's code interpreter.
  
  George B. Thomas (33:24.352)
  Boom. Simple, easy, actionable. If people have questions for you, where do you want to point them? You know, where's all the places? Because people, listen, they're either going to want to buy you a steak dinner after this episode, or they're going to have, you know, questions about whatever, but where do you want to send them?
  
  Connor (33:24.375)
  That'd be it.
  
  Connor (33:43.338)
  Yeah, I mean, I would say LinkedIn's probably my best channel. I'm super responsive. I post a lot of stuff. It's Connor Jeffers on LinkedIn, or you can take a look at Happily or something else. And we're around, we're happy to talk to you. And if you have questions or you're excited about AI stuff, reach out to me. I guess I'm very responsive. I don't think about it that way, but like I've had people see something, message me, and I'm like, oh, that's super cool, man. Like, how are you doing this? And they're like, whoa, I had no idea you'd respond. I am not.
  
  Like I'm, I'll talk to you. I'm very accessible. So I'd be happy to.
  
  George B. Thomas (34:17.276)
  I love that as another guy who is very accessible and people say, wow, you're up at that time. Yes, I am. Hub Heroes. This has been a great, great episode. If you're listening to the audio, make sure you head over to Hub Heroes community, community hub heroes dot com. We have the video version sitting there for you with show notes as well. Until next time, I know your mind's probably blown. You need a minute. You might want to sit down. You might need a snack after this episode. But until next time.
  
  Remember to be a happy, helpful, humble human. And we'll see you in the next episode of the Sidekick Strategy... I don't know, the Hub Heroes Sidekick Strategies. Have fun editing that, Noah. Awesome. Dude, let me hit this. Holy mackerel. I am not kidding you.