179: AI Sells You on AI

22 May, 2024

On today's show, Tim gears-up for a farm insurance conference out in Nashville where he's hoping to educate farmers on the pros and cons of artificial intelligence (AI). But, ahead of his talk, he'll be using AI voice technology to call the conference attendees and convince them to attend his presentation. And then, hopefully, weave statistics and sentiment analysis insights—from these automated calls—back into his slide deck.

Follow the show and be sure to join the discussion on Discord! Our website is workingcode.dev and we're @WorkingCodePod on Twitter and Instagram. New episodes drop weekly on Wednesday.

And, if you're feeling the love, support us on Patreon.

With audio editing and engineering by ZCross Media.

Transcript

Spot an error? Send a pull request on GitHub.

[00:00:00] Highlight

[00:00:00] Tim: maybe in the future, I think I could probably just do this podcast with AI, just, just my voice and we'll see. I think that'd be great.

[00:00:28] Intro

[00:00:28] Adam: here we go to show number 179. And on today's show, we thought we would buck the trend, do something that nobody else is doing. We're going to talk about artificial intelligence.

[00:00:37] Tim: What?

[00:00:37] Ben: So brave, so brave of us. Hmm.

[00:00:42] Tim: yeah, yeah. Oh, you know, Tim is all in on this stuff. So we thought we would dive in, have a, have a conversation here. So anyway, Carol's not able to join us. She is once again moving, because she's living that army life now. but, so it was just, just the boys tonight. Army life. Army wife.

[00:00:58] Adam: Yeah. Would have been Carol's turn to go first. So Tim, you get to, to go first, sir.

[00:01:03] Tim's Failure

[00:01:03] Tim: Well, I got a failure guys.

[00:01:05] Adam: oh.

[00:01:06] Tim: Yeah. Yeah. Yeah. So it's been a while since I've had to fire somebody.

[00:01:11] Adam: Oh no.

[00:01:13] Tim: And so we had, we've had a sales guy who's coming up on his third year anniversary. Next month in June. And, he just wasn't, you know, they say salespeople, they gotta eat what they kill and he just wasn't killing him. Wasn't killing him. He was getting a couple of birds here and there, but you know, no big game. So yeah, I had to let him go and I felt really bad. It's like, it's like anytime you let somebody go, it's hard, but it's like with salespeople, it's like, you know what your targets are, you know, you need to be bringing in X amount of revenue every year and you've never, ever done it.

[00:01:49] Ben: Well, that's what I was about to ask because in other verticals within a business, there are so many soft targets that if people aren't performing, you can put them on. We've talked about PIPs before, performance improvement plans, but in sales, it's very well articulated, I think, what is expected of you.

[00:02:10] Ben: So it's almost like. You can just say, like, do better. Like that's your performance improvement plan.

[00:02:18] Tim: quarter, it's like, yeah, so you're supposed to bring in xxx in net revenue and you brought in less than. . So we need to improve, you know, those are made up

[00:02:29] Ben: Yeah, yeah, yeah. But

[00:02:31] Tim: and it, to be honest, I thought he would quit on his own without me having to, typically these kinds of guys are, they're very type A personality, alpha driven, you know, and, and I also think we paid him too much salary.

[00:02:47] Adam: I was just wondering, like, I wonder if there was like a, you know, if base pay was too high, if it's base plus commission or

[00:02:52] Tim: I think, I think base was too high because we were trying to hire right out of COVID. And so it was like, they had all these sales guys that had just kind of been shelved for a while because there wasn't a whole lot going on during COVID. It's like 2001, 2002. And for the love of money, I mean, you couldn't, it was a bidding war on salespeople.

[00:03:14] Tim: And so we kind of got caught up in that and paid, I think paid too much. But yeah, I think he was, I felt really, really bad. He seemed a bit surprised, you know, it's like whenever you get on a call, your boss and someone from HR is there, you know, you're in trouble.

[00:03:29] Ben: yeah,

[00:03:29] Tim: Right. Cause I'm like, I'm like, this is the first time I've ever had to do a, an exit with an official AR person on the call.

[00:03:36] Tim: And so she was on the call and, you know, he's looking at that and, you know, it was pretty, you know, he seemed surprised, but then I talked to someone else, like one of our Customer service people that usually do the demos with him. And she said, you know, I, after I explained that my feelings to her, she said, I don't think he was that surprised.

[00:03:55] Tim: She said, every time we would do a call, he'd be like, I'm not doing my job very well. Am I? I'm like, okay. So he knew, he knew, he knew. So yeah, but that feels, it just feels bad. It's like, you know, Have I been fired? No, I've never been fired. So I don't really know what that feels like. I'm sure one day I'll get to experience that.

[00:04:18] Tim: But, I figured out that, you know, in our company, you get a two week severance for every year of service. And I've been here 24 years. So that's like 48. Weeks. That's like an entire year's salary. I decided I'm not going to retire when I get 65. I'm just going to make them fire

[00:04:36] Ben: You get a quiet quit,

[00:04:38] Tim: yeah, you gotta fire me.

[00:04:40] Tim: I want that severance.

[00:04:42] Adam: Seriously.

[00:04:44] Tim: So anyway, that's me. I'm just feeling bad. I'm, you know, we'll hire someone else, but in the meantime, I'm actually doing all his sales calls and everything. And I'm actually kind of looking forward to that because when I was doing sales, I had to like, do everything like round up the people and I've, I've Discover it.

[00:05:02] Tim: I enjoy presenting. I enjoy performing. I enjoy like, you know, talking and like, you know, doing the dog and pony show. I don't like, like cold calling and things like that, but we have a service that does that for us, right? They call and call and call, and then they set up a meeting and then it's okay. Now it's Timmy time.

[00:05:19] Tim: And I put on a show, so that I'm going to do, but eventually we'll probably hire someone and, and, yeah, probably do a lot more junior kind of person.

[00:05:28] Adam: So you had mentioned, you know, how difficult it is to, to fire somebody. I got to imagine I've never fired anybody and I, I can only imagine it's a very emotional, Thing on both sides of the table. But like you were saying with sales and with such a black and white clear cut expectations, it's got to make it a little easier, at least to just get through that meeting.

[00:05:51] Adam: Right. You gotta be like, okay, you know, these were the numbers you consistently didn't meet them. So we have to let you go.

[00:05:57] Tim: 100%, probably one of the most awkward ones. This was probably 10 years ago.she was on, this is when I was kind of doing the, our operations infrastructure team, security team. And, so, you know, that team is usually the ones when there's someone who's about to let go, you let the team know, cause they got to shut off their access to everything.

[00:06:17] Tim: So you're like, we're about to go on the call in 30 minutes, cut everything off, their email, all their SSO, but she was on that team. And so she sees like, sees like people, like, she just hears the chatter at the office. She's like, she comes to my office. She's like, so who's getting fired?

[00:06:35] Ben: You're like awkward.

[00:06:37] Tim: actually it's you.

[00:06:40] Tim: And she's like, what are you, what are you talking about? I'm the, I'm the best person you got on your team. Like you sleep like half the day, every day in the office. How do you think you're the best person on the team? So

[00:06:55] Adam: well, I mean, so real quick aside here, I found out after I took my job at, University of Pennsylvania, the Wharton Business School, that the guy that I replaced got fired because he would play World of Warcraft in his cubicle at work. Like, didn't even try to hide it. His back was out to like the middle of the room.

[00:07:14] Ben: Oh, come on.

[00:07:17] Tim: for the horde.

[00:07:18] Adam: Yeah. Like, I don't know if he, maybe he was, he was probably using headphones. Everybody in that office used headphones, but still like, seriously? So that's pretty bad.

[00:07:27] Tim: How often do you play?

[00:07:28] Adam: I, you, you mean that while I'm working? no, I don't currently. I did so very early on, in my current job before we like were a product company back when we were consulting.

[00:07:40] Adam: I would, you know, you know, we would get a project and they would be like, okay, here's the work we need to do. And here's the deadline. And I would get it done. You know, if the deadline was a week away and I'd get it done in a day or two. Okay. And now I'm going to make it double good. Right. I'll take a day and like super polish it, try to improve the documentation, write some tests, whatever.

[00:08:01] Adam: And then that, like that extra day or whatever, I'd be like, okay, I'm going to sit here and get paid to play video games.

[00:08:10] Tim: I think about it sometimes, like I go to the bathroom, I'm like, I'm getting paid to poop right now.

[00:08:16] Adam: Yeah, well, you know, honestly, I thought about it as like. The way I thought about that was like, it was like bonus pay for being skilled enough to get it or get it done early,

[00:08:27] Tim: Mm hmm.

[00:08:28] Adam: right? Instead of paying me in cash, they were, I guess, just paying me in free time. That's, I mean, that's probably not, the moral high ground, but, it's the truth.

[00:08:39] Tim: Yeah, for sure. Anyway, that's me in my 7 poop. Um,how, how about you, Adam?

[00:08:45]

[00:08:45] Adam's Triumph

[00:08:45] Adam: Oh, how do I follow that? so you remember last week, we talked about my code rot problem. I had that, that Lambda function that was running node 0. 10, as a binary included in the zip file. so, and we talked a little bit about like kind of how I was trying to wrap my head around fixing that code.

[00:09:03] Adam: There's a lot of different problems with it, whatever.And ultimately I fixed it in about 10 minutes by just taking out the binary. And well, so, you know, obviously we just had to, we, this was a, sort of a Hail Mary, like maybe, maybe this'll work. Right. So just deleted the binary from the package. And instead of doing like child process, child process.

[00:09:21] Adam: spawn, it's just like, Require this file and execute it, right? Like just run the code on the native node and on like, you know, node 20 or whatever, and hopefully there's nothing that's different and broken and it's, it was fine, actually, there was one thing. And that was, so it was a week ago that, that I made that change.

[00:09:41] Adam: And today we found a bug that it's like, unhandled promise rejection causes. It's like handled differently in node 20 versus 0. 10, which is okay. If that's the one thing that catches us. That's, you know, really not that big of a deal. And so, and we got that fixed today and yeah, yeah. Now the code is still atrocious, right?

[00:10:00] Adam: It's still full of classes and

[00:10:02] Ben: Yeah, but you, you modernized it.

[00:10:06] Adam: code on a modern platform instead of bad code on an old platform.

[00:10:10] Tim: I feel struck by that.

[00:10:13] Adam: Hey, that's, that's your problem.

[00:10:16] Tim: Bad code compiles.

[00:10:18] Ben: it is a testament though, to how powerful it is to have backwards compatibility in your, in your language run times. You know, we, we historically get very frustrated by the slowness with which certain languages move sometimes. And you're like, Oh, can't you guys just break it and be done with it? And then stuff like this happens.

[00:10:36] Ben: You're like, ah, thanks for not breaking it.

[00:10:40] Adam: Seriously.

[00:10:42] Tim: Like ColdFusion has deprecated things like a decade ago and it still works. It still runs. They just deprecated it. It's just a flag that says, yeah, it's deprecated, but like, it actually still runs.

[00:10:55] Adam: I don't think there's anything wrong with that as long as it's not, you know, holding them back.

[00:10:59] Ben: Yeah.

[00:11:00] Adam: You know, if, if you have to say, you know, hey, this feature that we deprecated 10 years ago, we're going to lose it because we're upgrading to a new version of, you know, some Java library internally, and that's gonna.

[00:11:11] Adam: Make the old feature not work anymore. Okay, cool. It's been deprecated for 10 years, whatever.but like

[00:11:18] Tim: But instead what they do is they do a new Java. And then all of a sudden your date formatting that had capital M's forever for months now have turned into seconds and now all your dates are screwed because of an undocumented change. That's what they do.

[00:11:36] Ben: Well, remember IE, Internet Explorer, used to embed two different rendering engines, didn't it? There was like the quirks mode rendering engine and the modern rendering engine. And it would just switch between the two depending on your doc type declaration, I think, I think that's what it

[00:11:52] Adam: doctype or I think that there was like there were ways to like force it into quarks mode You could like put just like this weird

[00:11:59] Ben: There was a, there was a meta tag too. You could do like meta tag, i. e. edge compatible or something too.

[00:12:07] Adam: early days of

[00:12:08] Ben: Yeah.

[00:12:09] Tim: Yeah. Words IE and modern do not a portmanteau make.

[00:12:16] Adam: Yeah, pretty one pretty easy to to package up there for me. Just code rot issue has been solved well, actually, I feel like Instead of solving it. Yeah. Instead of solving it, we just kind of kicked the can a little further. You know, we're, we're able to continue on the latest runtimes and that's, that's good enough for now.

[00:12:31] Tim: Mm hmm.

[00:12:32] Adam: That's it for me. How about you, Ben? What do you got going on?

[00:12:35] Ben's Triumph

[00:12:35] Ben: I'll go with a double triumph. I have a soft triumph. Double. Making up for Carol.I have a soft triumph and a technical triumph. A soft triumph is over the weekend. And I. Finally uninstalled Twitter and Facebook and LinkedIn from my phone. And this is an X for my phone. And this was, this is no shade against the platforms per se.

[00:12:55] Ben: it was just, I, I was spending too much time just mindlessly scrolling, like not even scrolling and looking for anything in particular. Just, it was a distraction and I was having trouble fighting it. especially with regard to Facebook Reels. Facebook Reels is their short form video content. And I kept trying to hide it from my Facebook feed and it would disappear.

[00:13:17] Ben: And then like 15 minutes later, it would come back and, and, and my lizard brain would take over and I would just click through these videos and I didn't want to, and I felt bad about it and I just couldn't stop myself.and I just know that that was not a good use of my time. So I finally uninstalled it and feeling pretty good about that.

[00:13:35] Ben: I'm still using the platforms. I'm just checking them on my desktop, occasionally

[00:13:39] Tim: Okay. I was about to ask, how are you going to know what I cooked over the weekend?

[00:13:42] Ben: testicles. It's not a hard one.

[00:13:46] Tim: Touché, touché.

[00:13:47] Ben: So that's my soft triumph. My, my technical triumph is that. I, I started playing around with SQLite. SQLite's interesting technology. It's been around since like 2000. So it's been around almost 20, yeah,

[00:14:01] Tim: Mm hmm.

[00:14:03] Ben: it's been around forever. And, it's super popular. I, I, people estimate that it's like the most widely used database in the world because it's an embedded database.

[00:14:11] Ben: It can just be shipped with anything. Cause it's, it doesn't even need a separate process. It's literally just. Driver code that has kind of this in memory representation of the database that sometimes flushes to disk. And, but I've just never had to use it and it seems to be having a moment. And, and by that, I mean, a lot of the technical podcasts that I listen to have recently for, you know, maybe just randomly been talking about SQLite and how great a database it is and how much people are loving it.

[00:14:39] Ben: And like people are getting it to run in the browser and they're getting it to run in WASM and they're getting it to just be everywhere. So. It felt like something I wanted to, to play around with. And, there's a company that creates a JDBC Java database connectivity driver for SQLite. So I was able to just create a data source in Code Fusion that connects to a SQLIte database, which is really just a file that gets safe to dis and I don't know, I, I don't think it has any particularly valid use in, in like a cold fusion based web application, but.

[00:15:14] Ben: It felt fun to do and, and, it was nice to try something kind of just for the sake of trying it, not actually trying to solve a particular problem. So that's me.

[00:15:24] Adam: Yeah. I mean, I feel like I wish I could remember, I think I've used SQLite in something and it's so, it's bizarre how easily it just drops in.

[00:15:35] Ben: I, and I think Safari as a browser, I think they've supported it for a really long time.

[00:15:41] Adam: Yeah. So there's a, that's what it was. There's a. A, an HTML five feature or something was like a, they didn't call it

[00:15:48] Ben: IndexDB.

[00:15:49] Adam: What? Maybe it was index db, but there was Yeah. Some, some sort of like in browser, in-memory database.and everybody was just implementing it with SQLite and they were like, Nope, we're taking it away because it's not SQLite, the standard.

[00:16:01] Adam: It's, you know, it was intended to be like a new thing. And I, I don't understand the politics of it, but that's, that's kind of what happened.

[00:16:07] Ben: Yeah. But it, it, it seems like, it seems like it would be worth me noodling on a little bit only because from everything that I've heard, because there is so little overhead to creating databases after they're just physical files, it can completely change the way you think about data access. So one of the examples that I heard on a podcast recently was that instead of having a multi tenant database where you have a user ID column in every table to differentiate this owner from that owner, you can literally just say this file is for this user's version of the database, and this file is for this other user's version of the database.

[00:16:44] Ben: And there's no more multi tenancy at the database level because everyone has their own database. Which I think is kind of, there, there was, I guess there is a database. There was a database called CouchDB and I, and I've never really seen it get talked about. I don't, maybe it didn't get a lot of traction, but I think that was kind of one of the things that they did was they would just create these on demand databases per user, and then you'd be able to sync them to the database.

[00:17:10] Ben: Individually to devices and you can start to do, what do they call it, local first or offline first. That development where you're, you're intending for the user to have a local database and then it would sync up to a remote PouchDB or a CouchDB using this library called PouchDB. Anyway, it seems like SQLite kind of enables that type of thing.

[00:17:30] Ben: I mean, I don't know how the syncing would work, but it, it just seems like. It's one of those things where you, you, you question everything about data persistence, and if you can question everything, what else can you question in how you think about application development?

[00:17:43] Tim: Sorry if you've said this, but so what's your use case

[00:17:45] Ben: I have no use case. I have no use

[00:17:49] Tim: I didn't hear it. Okay.

[00:17:51] Ben: than, I just thought it would be fun. And, you know, I think I'm sure it's part of it is like procrastination. I have other, maybe more highly valuable stuff that I could be doing, but I don't want to be doing. So I'm like, Oh, let me take a couple of days tangent on a SQLite for a bit.

[00:18:06] Ben: And then I feel like, I feel like I'm doing something, even though I'm not really doing anything, which is the best kind of productivity.

[00:18:13] Adam: I know that feeling far too well, right? Like.

[00:18:16] Tim: You're not playing World of Warcraft? Oh, what's going on? I

[00:18:20] Adam: with that because I'm at this point in my career and in my company, we're like. I don't have really any authority. We're a very flat, structured company, but like, we kind of get left on our own. You know, occasionally Steve will be like, okay, I need you guys to float these things to the top of the to do list.

[00:18:37] Adam: These things are time sensitive. I need them sooner than other stuff. But like, if there's nothing like that going on, it kind of just like leaves us to like, okay, these are the, these are the, the goals that we're running towards, but like anything that you want to do that gets us closer to that, right? And sometimes I feel like.

[00:18:57] Adam: I'm just, like, I'll find myself doing something and I'm, and then I have this moment where I go, am I moving in the right direction? Am I running toward one of those goals? Like, kind of, right? Like, sometimes, and often I think that I have those feelings when I'm working on something that is going to make our lives better as we move in that direction, right?

[00:19:16] Adam: Like, this, the, when I re rewrote that, mod, that collection of modules from, you know, JavaScript to TypeScript, And added tests and all that stuff. Like that didn't tangibly move us down the field, but it, it is going to make it easier and safer to move down the field. And, you know, there's been so many different things like that.

[00:19:37] Adam: It's just like, sometimes I feel like I'm, I'm filling potholes and, but for each pothole, I have to go invent a new type of pothole filler.

[00:19:47] Ben: Well, sorry, Tim, you were about to say something.

[00:19:50] Tim: was going to bleed into our topic. So you go ahead and I'll,

[00:19:54] Ben: I, I just wanted to add that. I think there is a, an implicit value. In play, meaning the, the, you know, the art of doing something just for the sake of the enjoyment. I think playing in the context of programming creates creativity in a very different kind of way, and I can't tell you how many times I've done some stupid little experiment only to two, three years later come into a problem.

[00:20:23] Ben: And I'm like, whoa. I know how to do this. And like, you know, reach back into the bag from three years ago. Yeah.

[00:20:32] Tim: Is that a hackers reference?

[00:20:33] Adam: No,

[00:20:34] Tim: Dress? Oh, okay.

[00:20:36] Ben: so yeah, I think, I think play for the sake of play has a knock on effect for

[00:20:43] Adam: you know, I, I a hundred percent agree. And. Honestly, I think maybe what makes me feel like a little bit sour in the gut when I'm in, in those moments doing those things and I have those bad feelings is because I'm not having fun, right? It feels like I'm potentially not moving us in the direction we need to move and also not having fun.

[00:21:02] Adam: So I'm like, what am I even doing?

[00:21:04] Ben: No, I understand what you're saying there.

[00:21:06]

[00:21:07] How to Sell AI

[00:21:07] Tim: I agree with you, Ben. It's like, and it's kind of what I want to talk about today. So, you know, we've talked about AI on this show for a while, but I haven't really been that I've been learning about it, but that was one of my goals for actually last year and this year as well, learning more about it. But then I, when I found what I've been talking about lately about this, AI system tied to a voice with a VoIP, I actually got really excited finally.

[00:21:33] Tim: And so I started playing with it. There was no, at the time there was like, I just really think this is cool, I'd like to play with this. And now it's turned into a product prototype. Which customers are excited about and I'm going to, so because I fired, because I fired my sales guy, he was supposed, he was supposed to go to a show next week in Nashville to a bunch of,it's called NAMIC.

[00:21:55] Tim: It's, the National Association of Mutual Insurance Companies, and this is their farm insurance edition, which we do have some farm insurance companies, that are our customers.And so he was just gonna, I saw his slide deck and I was like, this really is not that exciting. He just, you know, there's a term in sales where you just show up and throw up.

[00:22:18] Tim: You basically tell everything, here's what we do. Right. And people don't care. They're like, I don't really care what you do. I care what is going on with what I'm doing. How are you going to help me with that? And so I'm pivoting and I very quickly had to, come up with, you know, contact the organizers and say, Hey, the other guy's not with us anymore.

[00:22:40] Tim: I'm taking his place. I'm coming to Nashville and we're actually going to not talk about payments directly. We're going to talk about AI. And with the idea of I'm going to educate your farmers, I'm going to educate them about what AI is and what it's doing now, how much of this is hype, how much of this is legit, and how can it actually help you?

[00:23:02] Tim: And is it even worth the effort?

[00:23:04] Adam: That's going to probably be very well attended.

[00:23:07] Tim: I hope so. And, and I'm eating my own dog food. So today, because I had to do the slides so quickly, cause they're due tomorrow. I basically put the premise of my talk into chat GPT, asked it to create a slideshow for me, or, you know, a PowerPoint presentation, so it printed out all the bullet points for me and it gave me a description of each image.

[00:23:28] Tim: And then I went to Adobe Firefly

[00:23:31] Ben: Oh my goodness.

[00:23:32] Tim: and put in the image prompt from chat GPT and said, give me an image. I got an image from that and put that on the slide deck and the same tool I'm using to, you know, do the calls where they call, you know, insurance people who are, you know, Policies are expiring or need to make a payment.

[00:23:50] Tim: And, you know, with AI voice and all that, I have all the phone numbers that everyone who's attending the show, about 700 people, and I'm going to call them using AI

[00:24:01] Ben: Yeah.

[00:24:02] Tim: to tell them to come to my AI talk and they can ask me questions, they can ask me questions. Like I'm going to basically going to prompt it with all the stuff about our company and that stuff about, you know, what we're talking about and.

[00:24:14] Tim: And just jam the prompt, you know, full of information about us and send them the, I've got to, you know, I'm going to keep an eye on it. I want to see what kind of the reaction is. I'm not going to just send all 600 out at once. I'll probably do one or two, kind of watch them modify the prompt a little bit.

[00:24:32] Tim: And, yeah, so I'm all in on this AI, but the main thing is, cause it excites, it's exciting me. Right. I can't sell the stuff we do, you know, obviously has value because we make millions of dollars every year, but I'm just bored with what I've been doing the past decade, right? So this excites me, this gets me excited.

[00:24:52] Tim: I can sell this because it's new and interesting to me. And, you know, eventually this will be boring and I'll find something else, but. I think education rather than actually trying to sell people something is a lot more valuable.

[00:25:06] Adam: I think you, you're absolutely right. You find yourself in a very useful position, right? Like, just like you were talking about, you know, instead of showing up and throwing up, you know, they, what they're interested in is what is interesting and relevant to me today that wasn't yesterday, and what do I need to know about it?

[00:25:20] Adam: And that is AI, and how much of it is hype, and how can I make it benefit me? So, I mean, it sounds like you're right on the right path.

[00:25:28] Tim: So what I need some help from you guys. I need some, some input. It's like, I don't, I can technically explain how, how like large language models work and things like that. But how do I explain that to people who are not, cause I've been to this show before and insurance people, so they're not, they're very smart people, but it's like, they're not super technical, right?

[00:25:49] Tim: How do, how do I explain this topic

[00:25:51] Adam: Yeah, yeah, yeah. So, you know, those, they're like back when I was on Twitter, so, you know, like four years ago,

[00:25:57] Tim: X.

[00:25:58] Adam: well, it was called Twitter at the time. something that would go around from time to time is like, okay, put, open up your phone, type, you know, reply to this read or whatever, and then like put in these three words and then just use the, the, the first auto, like the center or left or whatever auto complete word suggestions.

[00:26:14] Adam: And let's just see what your phone says. It's just predictive text based on, and it's trained on all of the words that you've ever typed into your phone. And so, the way I think about ChatGPT is like, it's that same thing. You know, it's got sort of a prompt there, a couple of words to get it started, and then it's trained on everything.

[00:26:35] Adam: Like, you know, some very large subset of the internet. Like, all human knowledge or whatever. So, that's the way I think about it. It's really just predicting what is the most likely next word in the sequence. ChatGPT. com

[00:26:47] Tim: Yeah.

[00:26:47] Adam: But hopefully with enough context and enough priming to lead it in a good direction,

[00:26:53] Tim: Hmm.

[00:26:53] Ben: know what it makes me think of? I don't know if anyone has ever seen this, but. Every now and then I'll come across a YouTube video where it's a magician and they say something like, I'm, I'm, I want you to think of three mammals, and one of them is an elephant and I want you to think of three numbers, and they have to be below this and above that.

[00:27:12] Ben: And anyway, he goes through a couple of things and he is, I assume he is priming you to build up some mental model. And he goes, and the number you're thinking of is 17. And you're like, what? That's crazy. This is a YouTube video. How could you possibly know? And it's just that, you know, they're good at predicting how you're going to complete that thing.

[00:27:30] Ben: And AI is like that, except if you could do it in like 15 trillion parallel ways.

[00:27:37] Tim: Yeah.

[00:27:38] Ben: it's bonkers. It, it, it's like it, it frustrates me because, on one hand, it seems so ridiculously Redundant. Like it's just taking other people's information and regurgitating it. And yet at the same time feels weirdly creative and it's, I don't know how to hold those two things in my head.

[00:27:58] Tim: I, I'll say today in kind of eating my own dog food, right? So I'm going out there, I'm gonna sell them AI with the idea of I AI just to get them in the door. Really? I'm trying to sell them credit card and payment processing via, via, but it's like. I was having a really hard time, like coming up with a slide deck that initial sort of getting started bit was like, for some reason, being a human being, that seems to be difficult for us.

[00:28:30] Tim: Just, you know, our minds are at rest and we're not doing anything. We're like, I really need to do this. You're procrastinating. It's like, all right, this is due tomorrow. So I hop into chat GPT and I just, basically I write and explain like seven sentences to say, here's what I'm talking about.first thing it did, it said, give me a better extract of what this is, and it did.

[00:28:52] Tim: And so I sent that off to the organizer. So here's my extract. Oh, that sounds great. Fantastic. We need your slides. I'm like, okay. So I took that same extract that ChatGPT gave me, popped it in there, and it says, I need to have slides for a 15 minute presentation on this. and I, you know, a little bit, a blurb about my company, a blurb about me.

[00:29:13] Tim: And. I mean, it popped it out pretty, I probably only changed like three or four things. Honestly, it had a case, it had a case study in one of the slides. I'm like, I don't really have a case study. So took that out. It was like, other than that, I'm like, this is pretty darn good.

[00:29:31] Adam: So, I mean, are you, you, you mentioned early that you, you intend to kind of shed some light on how much of this is hype. Are you also planning on, you know, so, and actually something you said just recently that made me think of this is you're trying to sort of sell them on AI and how it could benefit them.

[00:29:50] Adam: But like, are you also planning on pointing out the, the deficiencies, like how it will just confidently give you wrong information?

[00:29:58] Tim: Yeah, for sure. Yeah, so that's sort of the, that's the cons of it that I'm going to sell. Like, you know, that many times AI will be very optimistic about what it can do. And so you have, you know, and that's where our expertise, right? You know, I'm not selling them tools to create AI. I'm saying, give us your payment processing and we'll build the tools, AI.

[00:30:18] Tim: And one of the things that we know is that you need to make sure you, Lower the expectation of your AI that it can't, you know, email people or text people when it can't, and you do other things and that, that, that's what we'll do.

[00:30:33] Adam: Just like AI with a clubbed foot,

[00:30:35] Tim: yeah, yeah. A little hobble it just a little bit. And then also, you know, having a disclaimer, you know, at some point of the call, it basically says, you know, this AI generated, please verify with your agent that, you know, all this is, is true.

[00:30:50] Tim: but kind of,

[00:30:51] Ben: know that credit cards have been using, I mean, I think they called it machine learning for a long time to do fraud detection. Is there,

[00:31:01] Tim: which we

[00:31:01] Ben: I was going to say, yeah, is there stuff in the insurance world that you have been using in like a less sexy approach that, that has been helping?

[00:31:09] Tim: No, well, I mean, we don't really deal with their insurance policies. So it's like, there's no fraud detection there, but like, you know, our, We do monitor if like the same IP is like sending constant payments over and over again, right? So like we had earlier the card testing, things like that. Is that really AI kind of, but I mean, it's not based off a large language model.

[00:31:32] Tim: It's just more of a pattern recognition kind of, kind of thing. so I, I mean, I'll talk about that as well. I I'll be, I'll be honest, the thing I'm struggling with the most is I don't, I don't want to assume that people there don't know anything about AI,

[00:31:48] Adam: Mm hmm.

[00:31:49] Tim: I don't want to condescend to them by over explaining it and I only have like 10 to 15 minutes and I'm going to include a demo at the end where I basically, Have the AI call me and we'll have a conversation with it and talk back and forth and do that sort of thing.

[00:32:04] Tim: So I need about probably about five minutes for that.so how do I talk intelligently without talking down to them? That's my biggest fear.

[00:32:14] Ben: can only tell you that. I'm a fairly technical person, and I don't really understand what this is. And to me, the, the fancy autocomplete feels like a pretty good explanation. I, you know, cause autocomplete I get, and that kind of feels a little bit magical in and of itself. And this just feels like autocomplete plus plus. So, I don't know. I mean, I don't know how much you would expect anyone there, if this is not their domain of expertise, to like really even understand. So, I don't think you'll be condescending to anyone.

[00:32:51] Adam: I think it'll be easy to get people that are not, already familiar with the basics up to speed, right? You know, you, you'll, if it's the autocomplete example or whatever, right? To get people up to that, like, you know, I, more than just, I've heard, I should look into this and, and they actually understand the, the general thesis of, like, it can generate words for you or it can, you know, whatever, do, do the text to speech for you sort of thing. And then you just go from there, right? So you just spend 30 seconds to a minute, like, okay, if you're not familiar, like, we're just going to take one minute and get you up to speed and then you'll, everybody will be on the same page and go from there.

[00:33:33] Tim: The other thing that I worry about, cause I've been to the show before. This is a very, so these farm mutuals, even just the mutuals in general, insurance companies are very family oriented kind of companies. They tend to be like, you know, a mom and pop, grandpa, they've been in insurance for two generations.

[00:33:50] Tim: And these are hard shows to sell to because particularly if it's in like Orlando because they'll show up with all the kids and grandkids and they will spend like 10 seconds on the floor to see what's going on and then they will just book it to Disney World. Fortunately this is in Nashville and Nashville is a great city but that's a drinking city and so they probably don't have the kids with them on this one.

[00:34:13] Tim: They're there to see Kenny Chesney and all that, you know, go to the bars and everything. so because they're so family oriented, I think one of the biggest objections I'm going to face is, you know, they're hiring people to answer the phones. I'm basically what I'm trying to do is sell them a service that will replace or slash supplement their customer service. Right. So they can answer questions and that sort of thing.

[00:34:38] How to Sell Potential Job Replacement

[00:34:38] Tim: So my question to you guys is, how do I position this so that it's, it doesn't come across as I'm taking, taking away people's jobs. Although maybe I am, I don't know. Right.

[00:34:55] Adam: jobs that will continue to exist are the jobs that require specialized skill and knowledge, right? the, the jobs that are going to be the first ones to go are going to be the ones that are easy to automate. Okay. And so, like you were just talking about, you know, helping them incorporate it into their businesses.

[00:35:12] Adam: What if, like, if the pitch is, what if there is never, a, a wait, like a, a hold time on the phone to call and get basic information, right? You can call and ask for details of your account or whatever and get those all back without ever having to talk to a person and never having to wait because the phone, the, the AI can answer 50, 000 phone calls at once.

[00:35:35] Adam: and, and then if they do need to talk to a person, then they can escalate past the AI to a person, right? So like, that's kind of like a, I think that that's where the, the sort of the secret sauce of AI stuff comes in. It's like people don't scale as well as ai, right? So the, the things that you can accomplish with AI and that, and where that overlaps with, things that, what did I say?

[00:35:59] Adam: the things that AI can do and the things that. that you need high scale for, right? So like answering the phones and getting basic information, all of the, if you're a customer service person on the phone, probably I would guess like 70 percent of the time that you spend talking on the phone is just like easy stuff, right?

[00:36:18] Adam: So they never have to take those calls. And then, mm hmm, mm

[00:36:24] Tim: And one thing I was thinking of positioning is like, you know, cause they really, I know these people, they, they tend to, they're very loyal to their phone. You know, I don't want to fire Susan. She's been there for 15. So don't, let her answer the phone from 9am and take a two hour lunch and then to 5pm. But after that, it's like, what if they call after 5pm?

[00:36:47] Tim: Well, your AI can answer the questions that she can and, and, you know, still keep her on. Don't worry. You know. Let her retire. And then, you know, 10 years from now when she's out, you know, retired, it's like at that point, maybe AI will be so advanced, you won't need to hire someone else. So

[00:37:04] Ben: Well, certainly,

[00:37:06] Tim: yeah. Susan. js.

[00:37:08] Ben: having to be on hold, you know, is a frustration that everyone can relate to. So, you have a real good foot in, in terms of being able to relate to the audience.

[00:37:21] Tim: Yeah. Cause I mean, yeah, the phone system can answer any, any number of calls at the same time with no issue. No, no on hold. Okay.

[00:37:31] Adam: Yeah. I mean, that'll be interesting to see, like. How it can interface with backend data. So if it's, if it's receiving calls, then you don't have a whole lot of room to put a prompt in, right? You can say like, this is what you're capable of doing. And hopefully like eventually it can, you know, it can say, okay, give me your account number.

[00:37:48] Adam: And then from that, I can then go and answer, okay, your next bill is due on this date, and this is the amount. And

[00:37:54] Ben's Analysis Idea

[00:37:54] Ben: Yo, ran into a thought and this is probably a, an AI too far. it could be something fun. Let me just throw it out there. It's a bad idea.

[00:38:03] Tim: no bad ideas.

[00:38:04] Ben: if part of what you had in the presentation was an analysis of all of those robocalls you made telling people to attend and you, it's, you know, I don't have a good idea here only to say, like, you could say, look, I don't There are 700 people were potentially attending.

[00:38:21] Ben: I was able to make a call to all 700 of you using, you know, 37 unique, voices tailored based on your gender and your age and your background. And, uh, we were able to do sentiment. Well, no, but, uh, but, but like, that's, you know, okay. So take. Yeah.

[00:38:40] Tim: Yeah,

[00:38:41] Ben: And, and I don't know if you could do like sentiment analysis, you know, I, are you using bland AI, I assume is what you're going to, yeah.

[00:38:48] Ben: So I don't know if there's any, it sounds like, I think from previous calls, you were saying that, they record all of the transcripts. So I w you know, you could, I don't know, run that thing through some sort of sentiment analysis. You could say, well, we found is that people, when we use this voice, on average, people had a better experience compared to when we use this voice and, and make it like, oh, wow, this isn't just, it's not that I'm just scaling up people, I'm scaling up people and I'm getting insights that I wouldn't normally be able to get.

[00:39:15] Adam: Yeah. You gotta be like, okay, so, is John Smith in the audience? So John, raise your hand, raise your hand. We really pissed you off with that phone call, didn't we? I'm so sorry.

[00:39:24] Tim: I love it. I actually, I've really been, I I'm going to do that. We're really, really liked that

[00:39:30] Ben: Oh, there you

[00:39:31] Tim: it's a real, they're going up, I mean, if they show up, right. Cause I go call all of them, assuming I get ahold of them and they go, they're going to show up to the show. And to my, to my talk, they've already had an experience with me.

[00:39:44] Tim: And so even if it's a good experience or a bad experience, I can kind of make that a story about, you know, and my fallback on, on like, you know, if there's a bad AI experience is there's just not enough data. It's the more data that I can give the AI about, you know, the customer and their experience, the better the call is going to be.

[00:40:07] Tim: And so, you know, if they're like, yeah, well, you called me and it pissed me off because you know, blah, blah, blah. I've noticed that some of them. Like have different names, same phone number and email address because they probably have had their secretary or some planner put their info in. Cause I mean, there was a John Smith and a Sally Jones.

[00:40:30] Tim: Email is in there. I'm pretty sure that's not his email address, but she's like signing him up for the show. He's like, I don't want to get the spam. So, you know, and I'm like, you know, the reason you had a bad experience is because you didn't have a good data. So the main thing is that we're going to do when we integrate you into our platform is we want to make sure that your data is clean as possible so that your people have the best experience.

[00:40:49] Tim: And so that's the selling point rather than a negative.

[00:40:51] Adam: Yeah, exactly. You take a negative and you turn it into a positive. Heck yeah.

[00:40:56] Tim: I love it, Ben. I'm going to give you credit for

[00:40:58] Ben: All right. Nice.

[00:40:59] Tim: Yeah,

[00:40:59] Adam: And then, also I was thinking like, you know, when you, when you get there and you're in your session and you can be like, you know, okay. I did, I sent you these, the, the phone calls. How many of you were gonna, you know, there's, if there's like, you know, is it, is it like, seminar style where like there's only one presentation at a time?

[00:41:17] Adam: Or is it like a, like a ColdFusion conference where there are maybe four or five different talks on at

[00:41:23] Tim: It is only, I'm the only one presenting. So that's, yeah, I got everyone's attention.

[00:41:27] Adam: Okay. Well, then my thing probably wasn't going to work. I was going to say like, you know, how many people were going to attend before you got the call and how many people are here because you got the call and I convinced you to come or because you, I

[00:41:37] Tim: Well, I mean, I mean, they could just not come, right. They could go do something else. There's a, there's a show floor too. So yeah.

[00:41:43] Adam: You can still do, like, how many people got the call and, and, you know, you're more interested because of it.

[00:41:50] Custom AI Voices

[00:41:50] Tim: The only thing I'm upset about is that, so I sit you guys on our, host channel, my own AI version of my voice. Did you guys

[00:41:59] Ben: Yeah, that was great.

[00:42:01] Tim: How'd it sound?

[00:42:01] Ben: thought it sounded great.

[00:42:03] Adam: Yeah, I mean, it, it, the likeness of you is definitely there. The, there's, like, some breathiness

[00:42:07] Ben: you were just

[00:42:08] Adam: little nuances that are, yeah. Yeah, it very, like, bookish. Right? It wasn't, it wasn't, it didn't have the personality, but it had

[00:42:15] Tim: Tim Cunningham, Tim Cunningham being a narrator rather than just me talking, right? But yeah, I mean, I listened to it. I'm like, yeah, that's my voice. That's a hundred percent my voice.

[00:42:23] Ben: hundred percent. I, I wasn't even sure what you were telling us at first, meaning like you posted this audio recording and I'm listening to it and I'm waiting for you to tell me something about the recording, and then I realize like, oh, this isn't Tim, this is a fake. I get it.

[00:42:41] Tim: So the only thing, so the company I'm working with, like they are so backed up. I can't, in order to have a custom voice, I can't even have that conversation with them until like June. Like June 4th, I have a call with them about getting enterprise status. Cause I would really love to call people with my voice, right.

[00:43:01] Adam: Oh, for, for the,

[00:43:03] Tim: for the invitation, yeah, for the presentation. Yeah, that would be awesome. But I just, yeah, it's not going to happen just because of the timing of it. But, maybe in the future, I think I could probably just do this podcast with AI, just, just my voice and we'll see. I think that'd be great.

[00:43:18] Ben: You know, it'd be kind of fun, and I'm sure there'd be licensing issues here. When you're on hold with a service, having the option to just chat with like a celebrity AI, like, you know, wait, time is 20 minutes. Would you like to talk to AI Samuel Jackson? Like, yeah, sure. Yeah.

[00:43:43] Tim: snakes on this monkey fighting plane.

[00:43:47] Adam: You know what, honestly, I want, every time I get put on hold, every single time, I wish that they had, implemented the, the, like, I'll call you back when it's your turn feature.

[00:43:56] Tim: You know, I never trust those.

[00:43:57] Ben: No, I agree. There's, there's definitely a huge trust gap

[00:44:00] Tim: Yeah, I'm like, they're gonna forget me. They don't want to call. Particularly if it's like, if it's not something where you're trying to buy something from them. If it's like a government thing or, it's like, yeah, you're not calling me back.

[00:44:13] Adam: I, I, the very first time I remember running into it was with Amazon customer service. I was trying to return something and, and they, I mean, I think if I remember correctly, like I signed up for it online. Like there was a, I forget if it was like you call and then they give you the option to be called back.

[00:44:29] Adam: I think that was probably one of the times I've done it, but then other times I've done it, like, you know, you're, you're filling out the form online and you're like, I'd like to be called, you know, when, when you can get to me. Okay, daddy. yeah, so, like, honestly, that, the thing that I like about that is that it's so respectful of my time.

[00:44:48] Adam: And, and you guys said you're scared of it. I have never used a system like that and not gotten a callback. Like, maybe, maybe I lost my place in line and they served 30 people more before they got to me, but I was off doing other stuff. And they, you know, grabbed my attention when, when it was my turn or when they were ready for me, which was fine with me.

[00:45:08] Tim: guess Ben and I have just been hurt more than you have.

[00:45:11] Adam: Speaking of, of you being old, the, I was thinking for your phone call thing, I don't know if you have the ability to do this, but maybe you like pick a cutoff age and if people are below a certain age, you text them instead of call them.

[00:45:23] Tim: Yeah. So everything I do on voice, I can do on text as well. So text is even quicker and cheaper because you're not doing the text speech. So yeah, definitely. Well, I appreciate you guys input. I'm looking forward to it. It'll be next week. So I don't know when we record it. Maybe we're flying out Wednesday and then coming back Thursday.

[00:45:46] Tim: Then my daughter's graduation is on Friday. Okay.

[00:45:48] Adam: So I have to record early next week.

[00:45:51] Tim: Yeah, well, if we do early, then I can't give you an update. But if we do later,

[00:45:55] Adam: Oh, that's right. Okay. So you, the thing is, is what you're flying to. Gotcha. That part went right past my brain.

[00:46:03] Tim: how it goes, but, yeah, Ben, I appreciate your, your input. I'll, definitely kind of give this some statistics on the outgoing calls to get them to show up and their sentiments.

[00:46:12] Ben: cool, man. I, I'm, you know, I'm excited to hear

[00:46:15] Adam: So your, your slides you already submitted, are they like final? You're not allowed to change them anymore.

[00:46:20] Tim: I can, I can change, I mean, I'm paying to do this, right? So they, it's pay to play. Right.

[00:46:27] Adam: it's probably just like, we're making sure you're not full of

[00:46:29] Tim: exactly. Yeah. They, they want. They wanna make sure that you, they had like all these rules. You got, have you spoken in public before? I'm like, yeah. A lot, a lot

[00:46:39] Adam: I have a suit just for conferences.

[00:46:41] Tim: Exactly. Exactly. It's, yeah.

[00:46:45] Ben: Very cool. Good luck, man. I'm, I'm excited for you.

[00:46:48] Adam: Yeah. So maybe, maybe two weeks we'll get a recap. That'd be awesome.

[00:46:51] Tim: And I, yeah, and I'm excited to be excited. It's, it's been a while since I've been this excited about something, so I'm glad to get back in the presentation game.

[00:46:59] Adam: Cool. Cool. Cool.

[00:47:00] Patreon

[00:47:00] Adam: All right. Well, then this episode of working code is brought to you by uninstalling code rot and firing Facebook from your phone and listeners like you.

[00:47:06] Ben: Heck yeah.

[00:47:08] Adam: If you're enjoying the show and you want to make sure that.

[00:47:10] Adam: If we can keep putting more of whatever this is out into the universe then you should consider supporting us on Patreon. Our patrons cover our recording, editing, and transcription costs and we could do this every week without them. Special thanks, of course, to our top patrons Monte and Giancarlo. You guys rock.

[00:47:25] Thanks For Listening!

[00:47:25] Adam: we are gonna go record the after show which is a perk for the people that do support us on Patreon.basically the outro plays and we just keep talking and Tim got a 3D printer he wants to talk about. Apparently the northern lights are moving south. We're going to talk about that. and who knows what else, we'll see.

[00:47:41] Adam: And you'll see if you support us on Patreon and you can get all of the back episodes of the after show as well. You just, it unlocks the entire feed for you. So you can go back and

[00:47:51] Tim: it's a whole new world.

[00:47:52] Adam: the whole back catalog of it.

[00:47:54] Tim: whole new world. I'm,

[00:47:57] Adam: sings a new song, a Disney song in every episode.

[00:48:00] Tim: Yeah, I do. I do.

[00:48:03] Adam: if you want to do that, you can head over to patreon.com/workingcodepod and help us out. It's just a few dollars a month and it helps keep the lights on around here. That's going to do it for us this week. We'll catch you next week. And until then,

[00:48:15] Tim: Hey guys, this is not a sales pitch. I am always Always excited to tell you your heart matters.

next episode: 180: The Laws of Software

prev episode: 178: Upgrading From Node 0.10