The Accounting Podcast

Attention: This is a machine-generated transcript. As such, there may be spelling, grammar, and accuracy errors throughout. Thank you for your understanding!

Ashley Francis: [00:00:04] To Typekit is only, only the beginning and all of the complexity in the world, all of the just millions and millions of pieces of data that have been created because the Internet was created. We're going to have some a barrier in between us and all of that data that's going to quiet down the noise, simplify it for us. And also, as an added bonus, do stuff, do stuff we don't want to do because it's repetitive and boring.

Blake Oliver: [00:00:40] Hello and welcome to the Cloud Accounting Podcast. I'm Blake Oliver,

David Leary: [00:00:43] And I'm David Leary-

Blake Oliver: [00:00:44] And we are coming to you live from the recording studio with our special guest this week, Ashley Francis, CPA. Hey, Ashley.

Ashley Francis: [00:00:53] Hey, how are you guys doing?

Blake Oliver: [00:00:56] Doing great. Thanks for taking time out of your Saturday morning to to join us and talk about the holiday.

David Leary: [00:01:01] Weekend, a three day weekend. That's right. Let's record the podcast because we we can't skip a week. We have to record. I get.

Ashley Francis: [00:01:07] It. Yeah, I get it. I, I didn't realize that sort of thing until I was like, I'm going to make a newsletter. And then it's like, Oh, wait, that's a weekly thing. Oh, okay.

Blake Oliver: [00:01:16] Right, right. And then you start to think, Oh, did I really want to do this after, like doing 12 of them, you realize it's going to be hundreds of them. Yeah, but I mean, you just have such a wealth of knowledge. Ashley and I'm excited to have you on the show because you are a tax practitioner who is experimenting with AI. You've been on this ChatGPT thing since January and have been putting out some awesome content on your Twitter account which is at Seattle. Underscore tax everyone go follow Ashley on Twitter. So yeah, I'm excited to talk to you with David about all the stuff that's going on and you know what we can do with it right now and what we can expect in the future.

David Leary: [00:01:59] And Ashley had a tweet storm this weekend, right, or this week all about the new Microsoft build conference and tweet storm this weekend. Right. Or this week all about the new Microsoft build conference. And I couldn't keep up. So I said, Blake, just have her on the show so we can ask her the questions. It'll be easier.

Ashley Francis: [00:02:16] I felt so bad, I was like, okay, guys, this is just next few days. I'm just everything amazing. I see you're going to hear about it and there's so much that happened.

Blake Oliver: [00:02:28] So the thing that I saw and I haven't really been paying attention to this, but I did see the announcement about copilot, Microsoft, copilot, and there was this cool, splashy video that I want to play for everybody. It's just like a minute and a half long. It doesn't have any narration, so we'll have to like live narrate it. Okay. So feel free to join me in explaining for our podcast listeners what is what we are seeing on the screen. So here we go. Dramatic music.

Ashley Francis: [00:02:59] It was very dramatic. Like it made your heart beat a little bit, you know.

Blake Oliver: [00:03:06] Like introducing Windows Copilot integrated into all of windows. We've got this like chat bar on the side, somebody typing, How.

Ashley Francis: [00:03:15] Can I adjust my system? Do you see that? Do you know how to get into settings? Because I have to like flail around every time, right?

Blake Oliver: [00:03:24] So apparently now we'll never have to go into settings again and we can just ask it. It just did a.

Ashley Francis: [00:03:29] Dark theme right there just now. I didn't know there was, but yeah.

Blake Oliver: [00:03:35] We're dropping a PDF into the chat tool and it is summarizing the pdf.

Ashley Francis: [00:03:41] Brex with all of your apps.

Blake Oliver: [00:03:43] So somehow you're going to be able to like start a Spotify playlist from the chat. Yeah, Chill vibes. Stay in your flow and get creative. Now we're doing some graphic design here and we are asking it to send the graphic design to our team in Microsoft teams and it's doing it without us having to open the app.

Ashley Francis: [00:04:07] And everybody's very excited about it.

Blake Oliver: [00:04:09] Everyone uses a lot of emojis in these videos.

Ashley Francis: [00:04:11] They sure do. Yeah, they're.

Blake Oliver: [00:04:13] Really good at it. I don't know how people type emojis so fast. Like I always have to like open up a little search and I don't know, I need to figure that out anyway.

David Leary: [00:04:21] So if I quickly just summarize this quick. Four weeks ago, five weeks ago, six weeks ago, whenever it was, we talked about copilot. Microsoft showed all this stuff into Office 365. Now copilot is going to be just part of Windows.

Ashley Francis: [00:04:36] Yeah. So as they kept talking during the conference was like, wait a second, copilot is just what they're calling, um, the functionality that's going to be the functionality that's going to be in all of their products. So and obviously all of this functionality is going to work differently for each product. So like Windows, what we saw was it was pulling up that Spotify playlist. It was able to access all of your documents that live in your file Explorer. It was, you know, it's able to bounce around between your different apps. And so that's kind of how you work with your Windows environment in Office 365. It's like I saw this presentation where this woman was in teams chat called up the copilot, asked it to pull all of the emails related to a specific topic. Summarize it for her. Go grab the documents around that project and then create a response to an email based on what all the stuff that she saw there. Wow. And it did. And that was all she had. It was all in teams chat. Never had to leave teams chat.

Blake Oliver: [00:06:05] So it's pulling the emails from Outlook.

Ashley Francis: [00:06:06] Imagine, like with our clients and our client, like the amount of data we receive for a client and how. Like our clients receive so many disparate pieces of data that all deal with different things. And if a client asks us a question, we have to go down the rabbit hole of which folder is it in, what you know, what are all of the things attached to it? Do we capture everything? We're now copilot is just going to do that. What am I going to do with my time? Don't understand this. This is so.

David Leary: [00:06:43] Yeah, because a lot of it's that wasted time, right? You're hunting, hunting, hunting, hunting, hunting, hunting. You find it all and you got to open them up. And then a lot of times, it's like, you know, there's an email where the specific thing was said. You need to go see it or grab it, and that takes you. That's the one that you always can never find. The important one you're clicking, just search for it.

Ashley Francis: [00:07:00] And like 800 emails come up, you're like, was it was it May? Was it January? I know we talked about this. And the other thing is, too, is that there is that moment where you have to remember everything that happened around that particular conversation, synthesize everything you're reading in order to answer a quick question. And that makes those quick questions mean. What quick question is out there really, especially with the complexity of what we're dealing with now. I don't think there's any quick questions other than what is my middle name, right? It's everything is getting so much more complicated. We get so much, so much data. So the the idea that our like our clients perception of what a quick question is, is now going to actually be a quick question for us. That's so exciting.

Blake Oliver: [00:07:52] Yeah. I'm thinking back to my time as a manager in public and my biggest challenge was probably just keeping tabs on all of my clients. And so I had to take really good notes Every time I had a meeting, every time an email came in, I'd have I had like client notes and in one note and I had a folder or a page for each client and I'd try to keep track of them that way. But this could this could do that for me, which would just save so much time because I didn't always make good notes, right? Like, to be honest, I didn't always write everything down. So. Great. Yeah.

Ashley Francis: [00:08:24] Well with the new. So there's. There's the new teams coming out. Because one thing I hear consistently about teams is how much everyone hates it. And from what I understand, that's a function of the fact that it was never built to support as many people as it had to support during during the pandemic. Like suddenly everyone shifted to using teams and like, oh goodness, okay, let's let's get these folks supported. And so it wasn't really meant to do that. They had to kind of merge to, to software. Platforms are. See, this is why you take me to a developer conference. I'm only half helpful, but, um. And so there's a new team coming out that is in preview I think right now that fixes those problems of being slow and clunky. And then also when you are in a teams meeting, you can record the transcript, you can do that already, but you can have copilot in your transcript and talk to it about what's just happened. Or have it tell you how folks are feeling based on their what they're saying or what's what's the temperature of the of the meeting or the one the example I loved was if you show up late to a meeting, you can have it tell you everything that you missed. And if somebody mentioned you so soon, your transcripts are actually going to be useful because you can have them summarize all of the like data points pulled out instead of, um, instead of just. Like, That's great. I have a transcript. I'll never look at that again.

David Leary: [00:10:13] Really, one is the opposite. Like, instead of me getting a summary of the meeting, I'm going to give my 2 or 3 things I want to say in the meeting and just have chat, attend the meeting for me. And when they call on me it just responds. And then you don't ever attend the meeting. Just I don't hate the meetings.

Ashley Francis: [00:10:28] I hate that.

Blake Oliver: [00:10:28] This is good news because I saw a report from Microsoft that the number one killer of productivity in companies is drumroll. No surprise. Actually, it's just people spending an entire day a week in meetings, close to eight hours in an entire weekday each week in online meetings. Inefficient meetings are the number one workplace distraction that hurts productivity, followed closely by having too many meetings. That was from a Microsoft survey of 31,000 workers across the globe, nearly two and three people, regardless of whether they are working remotely in person or on a hybrid schedule, say they struggle with having the time and energy to do their job because of meetings and email bloat. So if I can summarize meetings for us so we don't have to attend them, if it can summarize email threads so we don't have to read them, then that's going to that's going to really help us as knowledge workers.

David Leary: [00:11:20] And they also looked at the data at that same article. They looked at the actual data they had of trillions of minutes. They've been tracking of people doing different things. But my takeaway from this article, Blake, is like and actually you could tell about this too, in general, like everybody's really hard on themselves, right? And like we just beat ourselves up and there's all these books like Deep Work and all this stuff, and like, there's this pressure of like, Oh man, I have to optimize my whole schedule so I can get 40 hours of deep work in a week. Because if I don't, I'm useless or you beat yourself up. And so it's really this an issue at all? Like, should we just really reframe like, hey, and the reality of the working world, if you get four hours of deep work a week, that's a home run. And that's just the that's just the environment we are like instead of creating this undue pressure because even articles like this, it just creates more pressure. Like you're, you're, you're not worthy, right? You're not doing enough deep work. Right.

Ashley Francis: [00:12:14] I think it's really interesting that we have this idea that the human brain can operate for 40 hours a week at its optimal level, the whole 40 hour a week workweek thing. And I might be incorrect on this, but I believe stemmed from limiting the number of hours that factory workers were required to work because they were working in like just 12 to 14 hours a day. And so 40 hours a week. Right. Was was the like concession for factory workers, physical labor. But the human brain isn't isn't capable of doing really smart things for 40 hours a week. You're going to get a good 3 to 4 hours of brilliance out of it a day. And then the rest of it is just kind of churning along. So kind of one of the exciting like, I really hope when this technology comes out that our first impulse isn't to do more work, like, Oh, look, it's done everything for us, so therefore we're going to add 100% more of of what we do. I'm hoping that that we can start discussing. Okay, this does a lot of work Now. We're going to kind of respect the human brain cycle and give folks the space that they need to do the deep work and then the space that they need to do creative work and, you know, do people stuff.

Blake Oliver: [00:13:51] Most of the work we do these days, like right now is not deep work. It's not creative work. It doesn't take a lot of brain power. It's just a lot of clicking and moving stuff around and collecting documents and replying to emails. It's really it's stuff that other people could do, but we don't have those other people to do it well.

Ashley Francis: [00:14:08] And it does take up like brain power, though, right? It takes up our valuable brain power to go out and hunt for documents, read through them all again, make sure we have everything, synthesize it, and then create output that takes up valuable brainpower, that could be doing things that are actually value add.

Blake Oliver: [00:14:29] So you're out there, you're doing stuff with ChatGPT. You know, we've got other folks putting out prompts and showing us what we can do. But, you know, the number of accountants, the percentage of accountants that are actually taking advantage of this has got to be really, really small right now, right? I mean, so it's going to take building this into products like Microsoft and in our practice management software, our tax software, to actually get those big productivity boosts that we're looking for.

David Leary: [00:14:56] Quickbooks, it's coming there.

Blake Oliver: [00:14:57] You know, So then I think, okay, well, knowing the pace of change in the world of accounting and tax software, especially in tax, especially in audit, it's going to take a while. Right. And that's kind of disappointing. But actually, I was talking to you a couple of weeks ago, and you were talking about how with Microsoft's Suite, there are products in there where we could actually start building this in our practices today.

Ashley Francis: [00:15:24] Yeah. Oh, that's very exciting. Yes. So there's actually a couple of options and one of them is and it's very exciting because they did all of these announcements at the at the conference. And one whole section of it was the power platform. And the power platform is kind of a funny name. And it's kind of a dorky name, I'm going to be honest with you. Like, you're like, what is a power platform? It sounds very 90s, right? We've got the power. You got the power. But it's it's these series of this is a suite of tools that. Yeah. Pulls in the the low code. So if you're familiar with Zapier, you have like a power you have power automate, which is in like like all Microsoft products, it's way more robust than a Zapier. So yes, it will move move things from point A to point B, but it will then allow you to completely transform what's what's in that point A to point B, So in our cases, we have so many repetitive tasks that can be automated. We just need to learn how to we just need to build processes, right? We're so bad at building processes for ourselves. So if we build processes that for these things that we do repetitively, we can automate it.

Ashley Francis: [00:16:59] And one of the things that I like, one of the number one things that I tell people is that think about the thing that takes the most time out of your practice. One of those things is engagement letters. Imagine if you could automate creating engagement letters. How much better would your life be? Probably all the better, right? Someone was saying it took. I don't know. It was like 100 hours for their admin to do all of their engagement letters. Oh yeah. I'm pretty sure their admin had other things to do during that time period. So you can absolutely automate engagement letters using power automate. And at the conference they announced that copilot is going to be is in power automate so it's in preview right now and Big Four power automate had this described to design feature which is basically tell it what you want and it will try to figure out what you what you're what you're getting at and now they have that. Plus once you're once your flow is designed, you can have like a chat, GPT sort of interface, kind of add things to it and explain to you what your flow is supposed to do.

Blake Oliver: [00:18:11] So it's power automate something that if I just if I have a subscription to Microsoft 365, I get that as part of it.

Ashley Francis: [00:18:17] Yeah. If you want the premium connectors, it's it's $15 a month. And then if you want the the robotic process automation, that's $45 a month.

Blake Oliver: [00:18:28] But so for my whole org or for like per user.

Ashley Francis: [00:18:33] Per user, but not everybody has to use it, right? Like in my mean I'm, I'm the only one who puts these sorts of things into my, my goodie basket when I'm adding these to my because I'm the only one who uses it. So you can you can choose it on a person by person basis and $15 a month for like. All of the connectors you get access to, plus the fact that it will modify your data in between and bring in data from other sources.

David Leary: [00:19:00] So Ashley, where's it at on the consistency? And this is the problem I'm having with with ChatGPT type tools is I'll do something five times in a row and all of a sudden, like the sixth time, it's so I'm like, it's Bizarro World where it got that from. So I'm thinking like, if you have to do 100 engagement letters, how, how does every letter like. Come out the same and you know what I mean? Where it's not like all of a sudden it just changed the tone and it created these new sentences and it's like, why did you just impulsively do that? Because I know it's just rolling a dice and predicting what to do first. But how do you control that, that variability that you see so much right now with this stuff?

Ashley Francis: [00:19:36] That's actually the great thing. So like when I talk about this stuff, I like to separate out using AI, generative AI versus when would you use an automation? Because the automation, you want to automate things that you want it to be the same every single time because the variability and randomness is built into the GPT four engine. So when we talk about ChatGPT, it's sitting on top of the GPT four engine. And one of the things that they really like about it is that it is random and has a level of entropy. In fact, one of the presenters was kind of lamenting that the GPT four engine lost some of its entropy when like during the training was like, Oh no, no, we're fine with that. We're fine with it losing like entropy. So when you think about like generative AI, you're thinking instead of it being like a robot assembly line, you're thinking about it in the sense of an intern, so someone to help you. But if you want an assembly line, you build an automation in power, automate instead.

David Leary: [00:20:53] Yeah, because I feel like the intern analogy right now, it feels like I have a new intern every two days and it's like when really an intern you're going to work with them and then you're going to say, then tomorrow they're going to kind of remember the way you ask them to do it the day before. And they might still make new mistakes, but it gets tighter and tighter and tighter. And that's that that's what I'm looking for, is how does how do these these models start storing our own personal data? Like, how do we how do we hook it up to our data on the back end? And I just feel like Microsoft's hinting at this like, oh, there's maybe $200 a month you can buy, you can get some level with Azure, some level of ChatGPT type product or copilot that connects to your data on your back end. So you can almost train it, right? You could feel safe putting a credit card number in so you can book your flight, right? That type of stuff.

Ashley Francis: [00:21:38] Yeah, they did more than hint at it at the build conference. So they basically mapped out how copilot is connected to your data. It sits like it sits in your data, has access to your data, can do what you need it to do with your data. And this is a question that came up several times. Copilot is the only thing that has access to your data. Microsoft isn't using your data. Nobody is getting trained on this data. Like their whole thing is we're not we're not using your data for anything. Your data is your data and we're not touching it. This copilot is your copilot working on your data. So like, give me an example of something that like a solution you were trying to to solve that your, your intern just completely messed it up after two days.

David Leary: [00:22:39] Well, I mean, this is more ChatGPT. I had it. I trained it so it could stick plane flights into my calendar. I could give it or just basically paste in the dirty email from Southwest and it would pump out X files from my calendar. And I trained it because I like to have the little airplane emoji on the email, train it all that. Then two days later it did not do it anymore. Yeah.

Ashley Francis: [00:23:01] Yeah, that's.

David Leary: [00:23:02] That sounds about it doesn't store anything. It doesn't store any of my it doesn't actually learn from me. And I think that's the that's the missing piece here.

Ashley Francis: [00:23:11] Yeah. So the thing I learned that mean and I would say that I've spent quite a bit of time watching videos and reading articles. And the thing, the new thing I learned about how now and maybe I already know this, but how ChatGPT works is when you start every new prompt. The model makes the decision about which rabbit hole to go down. It has, let's say, three different options that it can choose and it judges which option is going to be the best option based on the rabbit hole. It goes down. You could get any number of responses, but after a while it it every time you put in a prompt, it takes everything that it's learned before and goes down that same rabbit hole. Right. It doesn't choose new rabbit holes. It's just that same rabbit hole every time. And every time you put in, like, I'm going to travel here, I'm going to travel there. It's going down that same rabbit hole. But it's starting to forget things because it only has so much capacity.

David Leary: [00:24:16] Yeah, because even if I came to keep the same chat open, right, I don't close if you keep the same and I keep using that one, you're right. It doesn't it doesn't remember from 48 hours ago kind of it's.

Ashley Francis: [00:24:26] Going to start dropping to like how, how many tokens did you have in that chat? Do you know tokens. Oh, no. Okay. So tokens. I know all of this language is so strange. It is so white. Like why couldn't they call it a thing that we that that makes sense? Okay, so tokens are basically what I think of as the price that ChatGPT pays for turning through your data. And when we think about it in terms of like characters, like language, because these are language models, so the number of words, the number of spaces, commas, everything. If you're a big comma person that's going into your token limit, the you only get a finite amount of tokens and a finite amount of words into the model before it starts forgetting everything it told you before. So think it's like 3000 to 3500 words and it starts dropping off the end of the conversation and like losing its its track.

Blake Oliver: [00:25:33] And this is why prompt engineering is so important right now, because we have to put everything we want it to do in the prompt at the beginning. And we need to start new prompts, you know, before like David, you can't use the same chat over and over and over again for this.

Ashley Francis: [00:25:49] Yeah, you have to use a new one.

Blake Oliver: [00:25:50] You have to make a new you basically have to create the perfect prompt for it to make this file and just do a new new chat every time.

David Leary: [00:25:58] Yeah, because I can have different chats based on different buckets of, of the actions or the results I want in those chats. But they kind of go stale now is obviously it takes money and power and compute resources to to work. Now is this something actually that right now I'm paying the $20 a month plan? Is this something that it's just like, oh, one day people have bigger plans where it can remember for a longer amount. Is this a money resource thing or is it just like a technology? Like it just can't do it?

Ashley Francis: [00:26:24] Oh, it can do it. It can. Like you're like in the chat. Gpt GPT four engine has a a larger token. Um, can't think of the word right now, but like you technically could have a larger token library to send through it. But ChatGPT has it limited to this 4000, 3500. 4000? That's good to know. Um, artificially. So if you want more tokens, you go over to the chat GPT playground, which is about the time when a conversation folks's eyes glaze over and they kind of are like, Yeah, I'm fine. I'll just start a new I'll just start a new chat.

David Leary: [00:27:03] So does your and copilot is this going to these limitations that I'm seeing right now, is this something that Microsoft is going to use? My data should start going away.

Ashley Francis: [00:27:14] Yeah. Yeah. So it's your oh, this is a whole thing. This is the whole thing. Um, okay, so your data having access to your data means that ChatGPT doesn't need to remember it anymore, right? When we think about ChatGPT, it's super helpful to think about it in the terms of like mental memory. It's using its mental memory to remember all of these things. And I don't know if you guys have gone to a conference and you're like, I'm going to learn everything on day one and on day three, you're like, I don't remember my name. I don't remember where I'm parked. So ChatGPT is kind of the same thing where it has this mental memory, this ability to store up to a certain amount of data and like send it through its model. But when we have to add details to it because it's general, right? That's the G, the G is general, and then the P and the T are something else that I can never remember. So when we have to add details to it because it doesn't know us, it never met us, even though we spend maybe months and months and months quality time with A every time we open up a new chat, it doesn't know who we are. So we have to. Put in a ton of information for it to give us what we want. That's not going to be a problem in copilot because the information is already there. It's already mapped for us. Microsoft has their semantic indexing, which if you ask me what that was, I'd say I just know the two words and it looked really cool and I can give you a video on it so you can watch it. But basically it's like going into our data and being like, This is related to that and this way is related to that and this way.

Ashley Francis: [00:29:00] And so when we're working with Copilot, it's not having to remember everything. It's just going out to the maps that it made and pulling that data. And then the other thing, the other cool thing that is going to be the icing on the cake with this is bringing in plug ins, which is another weird word that we're going to have to learn. But plug ins are like, okay, so let's say you had a new staff person, right? And you're like, Here's your desk. We're going to tell you nothing and we're going to give you no tools. Good luck. You know everything that you learned in college. And we're going to come over and we're going to shout questions at you occasionally. Sound good. That's kind of how a lot of people treat ChatGPT and like an intern is like, I'm going to answer whatever you say because I'm I really want to answer your question, not because it's scary, but because it's a robot and that's what it does. So with plug ins, it's kind of the same thing as when we give our new staff a computer or we give them like for me, I use checkpoints, so we give them checkpoint or we give them a calculator because ChatGPT the math it does. And you know, everyone's like, ChatGPT failed the exam on 3.5 and like, yeah, it, it can't do math. Chatgpt four can, but it's all once again mental math. So giving it a calculator is like is like giving your new staff a calculator. You're not asking them to do all the mental math.

Blake Oliver: [00:30:38] And I just logged in a couple of days ago when I got back from my trip to ChatGPT and I saw I have plugins now. So I connected. I connected WolframAlpha. I connected Zapier. I connected Kayak. I haven't really played with it yet, but like I just tried. I just tried asking it to book a flight to Madrid for me and now it's asking me all these questions like the same thing a personal assistant would ask. Like from where are you flying? When do you need to get there? Do you have a preferred departure time? Do you want one way or do you want return? And it's going to go out to Kayak's database now and it's going to find me the flight that meets those conditions. Yeah. And like that's that's going to be mind blowing when that rolls out to more people, when more people start being able to do that, especially with with just in windows. Right. I need to book a flight. Just do it right there in the sidebar. Yeah.

Ashley Francis: [00:31:31] So that that Spotify thing you saw, that was a plugin, that was them bringing Spotify. And do you know who else has a plugin?

Blake Oliver: [00:31:38] Thomson Reuters. Really?

Ashley Francis: [00:31:40] Thomson Reuters has a plugin.

Blake Oliver: [00:31:42] What does it connect to?

Ashley Francis: [00:31:45] So I was super excited to see that plugin. I was like, Oh my goodness, I will never have to go out to Checkpoint again. Look at me. I will just do all of my text memos in word and have it. Make it look nice. It connects to West LA. It connects to all of the lawyer stuff right now.

Blake Oliver: [00:32:04] Oh, of course. The attorneys get it first.

Ashley Francis: [00:32:06] They get it first. But the fact that Thomson Reuters did it that fast and the it means that we will and maybe 3 or 4 years get our own plugin.

Blake Oliver: [00:32:18] Hopefully, hopefully not 3 or 4 years sooner, hopefully 3 or 4 months. That would be yeah, that would be nice.

Ashley Francis: [00:32:24] So the thing that like the way to think about plugins is that let's say we want to do everything, everything in just one space. Let's say that we're like, I never want to leave teams chat again. This is my safe space plug ins. Bring in that functionality to where we are so they're connected by APIs and the the plugin developers get to choose, you know, as an API for anyone listening because I keep saying that word and I forget that just two months ago when someone said that word to me was like the thank you, should I be offended? So an API is basically a software has a lot of doors into it that software providers permit certain actions to happen. And so if a software has an API, that means that that particular software is going to let you do something with that data or with a functionality, not all of it. Software providers get to choose what they let you have access to, but that's what these plug ins are. They're like basically a road between your robot who is sitting on your data and all of the functionality of your outside vendors and bringing it in into your space and you'll never have to leave your house again. Oh, wait, no, I don't think that's actually what happens. But, um.

Blake Oliver: [00:33:50] Well, if I can just chat, chat with the AI and ask it to get my groceries and pick up my laundry and and then maybe go outside.

Ashley Francis: [00:34:00] For you and get some sunshine. Yeah.

Blake Oliver: [00:34:03] Solve my vitamin D deficiency.

Ashley Francis: [00:34:05] Absolutely.

Blake Oliver: [00:34:06] Yeah. Order! Order some vitamins.

David Leary: [00:34:09] Do you want to jump into the earnings reports? Because Intuit, they all, you know, Sage and Intuit, they both loosely talk about AI, but I think Intuit's got some interesting marches there on and some of their, you know, you get into the conference calls, that's where all the good meat is.

Blake Oliver: [00:34:23] Well, because they're talking I know they're talking about using AI in TurboTax, right? To automate more and more of the the flow of that. And I sorry, I just triggered Ashley with that term.

Ashley Francis: [00:34:32] You you did. Oh, my goodness. You guys, can I say one thing that I saw that like I, I just knew that TurboTax is going to have it in there by the end of the year. I just knew it. I feel it in my pinky toes. Yeah.

Blake Oliver: [00:34:47] Yeah, go ahead.

Ashley Francis: [00:34:47] I was in I was in a session on a Q&A for AI, and a lot of folks are like, What about the destruction of the world? What about, you know, our data getting stolen? What about this? And what about that? And I was like, hey, you showed a you showed your forms recognizer and it it looked like you were you did something really cool there. Can it read tax documents? This man got so excited. He was like, yes, it can. And it's it's not just doing OCR now, it's GPT has eyeballs and it's reading text documents and extracting data that you just need to tell it what to go and get. It's it can do it for unstructured documents. Yeah.

Blake Oliver: [00:35:34] That's so that's great great news. Right?

Ashley Francis: [00:35:37] I had a whole summer plan to use AI builder to like, train a builder on how to read 1099 to impress my friends. But guess I'll just use that instead.

Blake Oliver: [00:35:48] So I guess that that leads us to an interesting question, which is, is it worth even taking the time to try and build these bridge apps that will use GPT to do this stuff or do we just wait until the AI builds it for us? Right. And that's why I wanted to talk to you about I want to talk to you about the Microsoft stuff because with power automate with power apps, right? We could plug into GPT and we could start automating all this stuff, but we're going to have to do a bunch of work to do it. But it seems like. This AI stuff is advancing so quickly that it might connect all of these rails for us in not too long. And then, I mean, you know, then we get into the whole philosophical argument of like, well, what will work be when AI is doing most of the work for us? And what will, you know, will the end of the will, will it be the end of humanity? But, you know, I think that takes us a little out of scope.

Ashley Francis: [00:36:44] Bu yeah, you wouldn't raise to the end there. Yeah.

Blake Oliver: [00:36:48] Well, once you get an AI that can that can reprogram itself, right? Which is I think where we're headed, right, I will improve itself. Is that, you know, can it do that. It can write code.

Ashley Francis: [00:37:00] Um, the interesting thing was in the because like I said before, remember when I said it goes down the rabbit hole? Right. It goes down the rabbit hole. And if it goes down a rabbit hole, it's going to just continue going down that rabbit hole. It doesn't have a way to go back. And so what? I wrote down his name a whole bunch of times today and I can't remember. He was like a co-developer and a co-creator of OpenAI. What he said was that the they're not there yet for the ability to for GPT four to go back and correct itself. That's not a capability that it has. And it would be it's kind of a leap to get there. And there's some models trying to trying to overcome that, like the the auto GPT stuff that, you know, that was supposed to overcome GPT four seconds like going down these rabbit holes like it was supposed to correct for it, but it's still very broken. Yeah, the idea, but still very broken.

Blake Oliver: [00:38:05] It got it got a lot of attention on Twitter and people started doing it. But then like none of them, none of them really worked, right. It always sort of spiraled into not productive behavior. So yeah, so that's a good I think that's actually a good news, right? In some ways that.

Ashley Francis: [00:38:22] I think so. I think that it will give us time to redefine what what we think work is, because right now, like I said, work is defined by a very old idea that was a concession to working people too much. And the fact that like we're holding large amounts of knowledge workers to that same standard that was created for like industrial workers, which probably was still maybe too many hours a day on on the on the body. Like, I would love for us to have the time to redefine what work is. And you know, the funny thing. Okay, so is it okay if I respond to some of the comments that popped up?

Blake Oliver: [00:39:11] Yeah, We've got a bunch of folks who have joined us in the live stream. We've got Sarah, we've got Brian, Jennifer, Judy, thanks for joining us.

Ashley Francis: [00:39:18] So to answer your question about like, why would we want to jump in now versus later and and the whole the the idea of ROI comes up as well. And so I wrote a I wrote a blog post in my my kitchen table automations newsletter, which once again has like, this is such a good idea to start. I'm just going to do it. And now I'm like, Oh, I'm going to do this forever. Okay, great. And with the Post was about was asking ChatGPT. It was this fictional like new staff person who wanted to ask ChatGPT how to save the company money. And depending on the adjectives that you use, you get different responses, right? So it was really fascinating because ChatGPT was able to generate all of these really great ideas to save the company money. And then in the last the last prompt was how do like, what ideas can I bring to my boss as a new staff to save the company money? That makes me look good. And it came up with a whole bunch of different responses, but that this young person, this new staff could do to improve, like the company's bottom line or the company's profitability. So I think when we think about ROI, we're thinking about it in like the really big, like heavy lifting sense of, oh my goodness, I'm going to have to become an AI engineer and learn all of these things.

Ashley Francis: [00:40:52] But there are so many tiny things we can do right now to incorporate this into our practices. And once again, going back to my the whole like there's going to be a difference between what AI is going to do and there's going to be things that you need to automate. So if the question is, you know, where should I focus first, I would say learn the basics of how to talk to the robot, because prompting is so much like prompting is so much different than a Google search query. Like we're having to relearn how to ask questions, ask clear questions rather than just a Google search query, which like pulls up everything under the sun. So like learning how to prompt is going to be an important skill because we are soon going to be prompting everything. But you know, once you get that under your belt, if you're not interested in creating like these bots, like I've been talking about that, break things down, that's fine. You don't need to, but you could also use it to help you learn new things. You can use it to help you identify areas of improvement. You can use it for so many different applications.

David Leary: [00:42:21] Every app in your tech stack that. You're currently using in your firm and with your clients will have a chat. Gpt, AI component prompt of some type. They'll be doing some piece of the workflow in that app through AI. So this is inevitable. Like mark my words, it is coming. We already I mean, we've already talked about some high level ones. These apps that send emails are all kind of adding it. We talked about Canopy and Carbon Client Hub is like the task list that will create the task you need. Like it's it's here. So like if you're thinking about like, why should I jump in? You're right, you could just sit on the sideline to because it's, it's just going to show up in all your apps you're using. And hopefully if the apps do it right, you shouldn't have need this expertise of prompts the app should be able to build a layer in between right to where you have to do less where it is just that magic light bulb button, you hit it and then it does all the work for you, right? But the real thing, I think the learning and that's why experimenting is important is understanding the limitations. Because if you just, Hey, I hit the magic light bulb and it it did something for me like you're you need to get to that point where you're like, okay, I know I have to check it, right? Like, I don't even know where I saw this. I think it was I'm blanking out on who tweeted. It was on Twitter like somebody talked. It's like, really? You're turning into like a traffic cop for AI, right? You're like, okay, that's okay. Let it through. That one's not okay. Like, go back, try again. And that's a that's the skill set you need to have, right? Is that interpretation skill set?

Blake Oliver: [00:43:51] So before before we go any further, Ashley, I want to give you a chance to talk about your newsletter. You have a newsletter. You started on Substack kitchen table.

Ashley Francis: [00:44:02] I did? Oh, my goodness. Yeah. So I was this was this is this is a shout out to Jason Stats. He had a daily daily show about just starting stuff, starting creating content. And I'm like, I talk all the time on Twitter. Why don't I just make a newsletter so I could talk all the time in a newsletter and it's free. And the whole idea around this newsletter is that a lot of the technology is kind of big and scary and we're hearing a ton about it and we don't know if it's applicable to us. So let's make it simple. Let's make it easy. Let's make it not scary. I'm not talking about anything like super wild. I'm not like the first section is like ChatGPT. The first section is ChatGPT. And then talk about like a Microsoft app, because if you pay for Microsoft 365, you get probably between 100 and 3 million apps that you've never used before. Every time I go in there, I feel like I'm like, Oh, I've never seen or heard of this app before. What does it do? So I talk about like a Microsoft app that you can use.

Ashley Francis: [00:45:21] And then the the last section is about using power automate, but not in like the scary you're a technical person and you you love to use Excel macros. It's more like you want to use it because it's going to make your life easier. What do you need to know? And so yeah, it's, it's the whole point of it is just to make it really tech friendly and or really, really friendly for accountants and tax professionals and even financial planners because I feel like we get bombarded with all of this like, crazy tech stuff and it doesn't need to be scary. I mean, we have learned really hard stuff in our profession. We have probably the worst software to learn and then learn the workarounds. And then we have like the technical stuff, like the the tax and accounting technical stuff. We have the capability of learning these things. It's just the way that it comes to us doesn't make any sense in our vernacular. So that's the goal of the newsletter is just to help folks get comfortable.

Blake Oliver: [00:46:36] You've also got some courses I did.

Ashley Francis: [00:46:38] This is another thing that was like, I don't think anybody's going to sign up for this. I it's a it's a I didn't think so. I didn't think so. So this one of the things I kept seeing over and over again was like, don't have time to go down this rabbit hole. I don't have time to learn about prompting. And I'm like, Oh, no, we don't have time not to learn about prompting because there are ways to. Well, for one thing, I don't know if anybody knows this, but nobody actually knows how to prompt chat too effectively. There's like studies that come out about, okay, so we found a new way to get a better response from the robot, like they made this thing and it didn't come with an instruction manual. So the idea of this class is like it is basically, you don't have to go down a rabbit hole. I'm going to lead you down this nice well-manicured path that is going to take you less than five minutes a day. It's just 30 days of tiny prompts take you less than five minutes a day, and by the end of it, you will feel very comfortable with just the basics of of a good, solid prompt and and like how you can different ways you can use ChatGPT to make your life better easier. Like that's what we're here for, right? This. Yeah. If it wasn't making our lives better, we wouldn't use it. Well, I like.

Blake Oliver: [00:48:09] The five minutes a day thing. That's. That's really good.

Ashley Francis: [00:48:12] Everyone's so busy.

Blake Oliver: [00:48:14] My key takeaway is, just from talking to everybody who's doing anything with AI is just. It's almost like everything in life just start learning about it and it doesn't have to be overwhelming. It can just be a little bit every day, just, you know, sign up, start using it, and and you'll be ahead of 98% of the profession, right? That is not doing anything. That is just sitting back and and waiting.

Ashley Francis: [00:48:40] I agree. Yeah.

Blake Oliver: [00:48:41] And it doesn't mean that you have to like, go and like like no, nobody that I know of has actually like built any of this into their standard processes in their firm yet. Right. This is not like like deeply. Like, nobody's doing that yet. But if we're learning about it, then we can, you know, when these tools become available, we can start doing that and we'll be ahead of the herd.

Ashley Francis: [00:49:04] Yeah. I also think that we're going to be seeing a lot of like AI flavored flavored software coming out like my, my, my software has now this flavor of AI and my software has 100% more AI in every bite. And unless you know or unless you're comfortable with what sort of AI is actually impressive and what sort of AI is like. Okay, well, thanks for putting that in there, but your software isn't enhanced by that AI, so I'm not really going to get excited about that. I think that's like just kind of getting in and learning about it is, is a way to figure out, okay, so should I be impressed or is this just something that they added because they wanted to put on their.

Blake Oliver: [00:49:58] The marketing, the marketing feature, you know, not the actual productivity.

Ashley Francis: [00:50:02] Feature, organic free range AI that has never that has never been outside of its like home and cage free, everything like that. Yeah. Uh, the, the thing I was surprised about with this class, it's so that first class, the June class, it is asynchronous, right? The idea is like you're going to get a video from me at the beginning of the week, but that's the only time you see my face because I'm not the star of this show. You're going to get every day. You're just going to get a little tiny exercise. Easy to do less than five minutes that will help kind of hold your hand and like playing with AI. And then if you're over overzealous than ambitious, then you love extra credit. There will be weekly exercises that will like take you a little bit more down the rabbit hole, but not not into the. I mean, you're not going to be building a bot like a auto, auto bot. Auto bot from this year. We're not automating.

Blake Oliver: [00:51:06] Our jobs yet.

Ashley Francis: [00:51:08] We're not? Yeah, this is not the class for that. So that class was so surprised it filled up. Um, that's great. I wanted to limit it to 100 people so that because I want people to get in the comments and share what they learned. Because I think it is like once you get into using ChatGPT properly, it is incredible. Like you're like, why does this exist? How is it possible? Kill it with fire, things like that, like all of the responses. And so I want people to share and and help others, like come up with more ideas. Right. So I was so surprised it filled up. So I have a July class now if if folks want to sign up and uh, yeah, what's the amazing I don't think.

Blake Oliver: [00:51:57] We mentioned the URL. Where should people go for both your newsletter and the class? Is there one place they can go? Wish Well, you've got kitchen table automations. Dot com.

Ashley Francis: [00:52:07] Kitchen table. Yes. Kitchen table automations, dot substack.com for the newsletter.

Blake Oliver: [00:52:13] Got it.

Ashley Francis: [00:52:14] And then kitchen table automations dot thinkific. I think i.com. I keep calling it thinkific and that is not the name. It's thinkific.

David Leary: [00:52:28] We'll put the link in the show notes. Okay. Yeah. That shouldn't be a problem.

Ashley Francis: [00:52:31] I'm I'm, I'm so excited though for folks to just start getting in and, and playing around with it because I think once they get in there and they start seeing what it can do. If you think about our profession and how many barriers we have to actually providing client service, we have so many things in the way that get between us and our clients, like bad Data. We have too many emails, we have all of our data siloed. We have, you know, everything is coming in from different places. The ChatGPT is only, only the beginning. And it's going to it's going to like all of the complexity in the world, all of the just millions and millions of pieces of data that have been created because the Internet was created. We're going to have some a barrier in between us and all of that data that's going to quiet down the noise, simplify it for us. And also, as an added bonus, do stuff, do stuff we don't want to do because it's repetitive and boring.

Blake Oliver: [00:53:46] And here's an example. While David was talking and I wasn't listening to him, I asked ChatGPT to book me a flight on Kayak or find me some flights on Kayak. It asked me all these questions. I said, I gave it a tough one. I said, I need to go to Madrid on Monday. You know, find me some flights. It asked me seven questions, I replied. And then it gave me four different options. With the price, with the return. How many stops? All that. Like just right then and there. That easy. Just using the plug ins like. Yeah.

Ashley Francis: [00:54:20] And if you go to the search right now so here's here's Microsoft wrecked my day because my class was like we're going to do it's week one. We're just learning the framework for a good prompt, which I know sounds like, Oh, that's that's going to take a whole week. Yes, because a good prompt has certain aspects to it. Week two was going to be adding in flavor and ways to enhance your prompt to make it even better. Week three we were going to go over to Bing AI and show what you can do with like pulling data from web sources. And week four was like is, you know, actually using it in your real life. Right? But at the at the conference they were like and now we bing AI is going to be the search feature for ChatGPT four and soon it will come down to ChatGPT 3.5. And was like, No, you didn't. You didn't just do that to me. Like, What are you doing? It's all changing so fast.

Blake Oliver: [00:55:31] That's, that's the thing that's nuts about all this. Like, I feel like, you know, when David and I started this show, what was it, you know, five, six years ago? David You know, the developments in cloud based accounting were coming in hot and fast, but it took years. And now we're we're seeing the same pace in months.

Ashley Francis: [00:55:53] It's it's not comfortable. It's very uncomfortable. Like someone who's trying to keep ahead of it. It's uncomfortable. It's hard.

Blake Oliver: [00:55:59] I tuned out for a couple of weeks. I come back and it's all different, you know, Like that's that's going to be the big challenge for us as professionals is just, yeah, dealing with this pace of change. So thank you, Ashley, for being out there and putting this information out there and helping us understand what's going on.

Ashley Francis: [00:56:15] Yeah, I think there's the nice thing is, is like as professionals, we really only have to worry about specific things, right? Our number one thing we have to worry about is security and then everything else. At least that's how I feel about it, like everything else falls out of that. So all of these like, auto parts and like gpts, that'll drive your car and gpts that'll do your laundry as a professional, Like, first look at security and then I look at everything else. Does it make my life better? Does it solve a problem I have? We have so many problems that we can solve while we're waiting for other people to solve other problems that it feels like we could probably just play around and come up with cool things that make our lives easier. And if someone then comes along and makes a similar cool thing, that's great.

Blake Oliver: [00:57:11] And then we know how to use it, we know how to.

Ashley Francis: [00:57:14] Use it and whether we should be impressed or not, Right? That's the other thing. Should we be impressed or not?

Blake Oliver: [00:57:19] Well, actually, it's been so great talking to you, David. I think you have something you want to say.

David Leary: [00:57:23] I was going to say we're coming up on the hour, but we're going to have to go long break because there's just news we got to talk about. We'll just have to go long. We'll just have.

Blake Oliver: [00:57:29] To do a double episode this week. So let's you and me figure out when we'll do that. We'll do a news episode. It was great talking to you, Ashley.

Ashley Francis: [00:57:39] I'm so sorry, David, for taking up all your time and you're excited on talking about.

Blake Oliver: [00:57:45] We've done this before where we have a guest on and we just like, we can't stop talking about the thing that you're here to talk about. And I'm so glad we got to delve deep into it with you. People should follow Ashley at Seattle. Underscore TurboTax. You can follow me. I'm @BlakeTOliver. How about you, David?

David Leary: [00:58:04] Just on all the socials @DavidLeary. Two of the things to plug. So we are going to be doing a session at Engage on Monday, June 5th at 7 a.m. So if you're going to guess, it's a CPA. Cma, right?

Blake Oliver: [00:58:19] It's a CPA hyphen, cma.com now.

David Leary: [00:58:26] So we're doing a session with ShareFile. The session is called Elevate and Modernize your Client experience presented by ShareFile. And then if I'm understanding correctly, actually you're speaking at Engage 2023 as well.

Ashley Francis: [00:58:37] I am. I guess we're the Monday crew I'm going to be speaking at. I think it's one 130. I should have I should have this. I'm just going to show up when I show up. But it's called navigating the Navigating Unchartered Opportunities is going to be a session on AI and personal financial planning.

Blake Oliver: [00:58:57] Can we swap with you because we're on at 7 a.m. So.

Ashley Francis: [00:59:01] I'd be I love being awake for hours by that point. I don't know your topic very well, but I'm sure that I can muddle through.

Blake Oliver: [00:59:09] Well, I look forward to meeting you in person at Engage. And we got to. We got to figure that out. Yeah, for sure. And I think I was.

Ashley Francis: [00:59:17] Going to come by your booth for your millionth episode so then I could borrow your equipment to sometimes my one of my trainings.

Blake Oliver: [00:59:27] You should. Sometimes it feels like a million episodes, but it's a million downloads. It's downloads. Yeah, but that's so pretty cool.

Ashley Francis: [00:59:34] That's.

David Leary: [00:59:34] Yeah. Monday evening, if you're engaged. 2023, swing by. We are going to do a little some cake and a toast, hopefully celebrating our millionth download. We're in the booth and track us down. Cool.

Blake Oliver: [00:59:48] Thanks everyone who joined us live today. As always, you can tune in, subscribe to Cloudaccountingpodcast.com on YouTube and you'll get notified when we go live. And we love to to chat with you all. Ashley we'll see you pretty soon.

[01:00:03] Yeah. Thanks, guys.