534 The Recent AI Updates Accounting Firms Need To See

Friends, we got a special surprise today. Chad Davis is probably the world's leading accounting firm AI expert right now. Through the work that he's doing running a pretty big community of accounting firm AI champions and the work that he's doing in my own community. Talking with tons of real folks in accounting firms every month now about like, what's actually working, like, what are people finding success with? Had the opportunity to have a little sit and chat with Chad the other day. We got the cameras out. I said, chad, tell me the, the stuff that's happened in the last, I don't know, 30 to 60 days that people most need to know about. Been a while since we've had chat on the podcast. It was good. Connecting again. That's what you're going to get today. A bunch of cool recent AI stuff, but not like the up here like too big brained AI stuff like the things that people are actually using and the, and the new announcements that are most relevant to accounting firms. Hope you enjoy. Without further ado, here's my sit down with Chad Davis.

Foreign.

Chad Davis, been been a while since.

You'Ve been on the pod. It's great to have you back.

I miss our podcast.

Jason, listen, the numbers since you were last on, they've gone. No, no, no, no, no, no other way. They've gone off the charts. So for folks meeting Chad Davis for the first time, what's the 60 second Chad intro?

I am an admirer of Mr. Jason stats from afar. But most importantly, thank you for putting out all your videos. We don't need to talk about me for a second. I think I speak on behalf of all accountants when I always say every day these videos come out and you look at the title and you're like, do I want to listen to this? And most of the time the answer is yes.

Most of the time. Chad, I, I did a podcast about like presenting well on video and like framing and all that stuff and Chad, I think you slacked me and you said. I actually watched that whole one. That was great. Thanks. Which just is like loaded with. I normally don't watch anything you do, but this one in particular was, this.

One was really good. Okay, fine. My name's Chad Davis. I'm from Canada. For the last eight years, I have lived in an RV with my wife and two kids. I own an accounting firm in Canada called Live CA with my business partner Josh. And on the side I hang out with a bunch of accountants online in a community called Automation Town, which is, which was named after our old podcast podcast and all we do is talk about the newest and the most frustrating things about AI and automation in the accounting industry. And we reference your videos a lot. So thank you.

Yeah, automationtown, if you have like the, the AI champion in your firm. This is how I usually frame Automation Town is like it is heaven for them. It's a whole bunch of other kind of AI champion types. So Automation Town definitely check that out.

Chad.

Also we're wrapping up an interconnect in Vegas and Chad just ran like a two hour like the biggest AI mega session here at the Conferen. Chad also runs a monthly month in review meetup in my online community. And so like, I feel like Chad is actually more plugged into what are the announcements that are vibing with accounts that they're excited about the most. And that's the goal for today. Like last 30 days catching people up. What are the headlines that maybe they've missed that as you've talked with folks, accountants have been most excited about.

Yeah, it's fun. I think there's a lot of stuff that happens every single day. Whether you're looking at the news cycle or what your friends are talking about or, or LinkedIn or Twitter. Like, it's really hard to decide what is good and what's worth your time. And unfortunately I have no filter, so I'll take everything in and I'm sure there's. There's some, something happening up here because of that, but I'm, I write it all down. So as part of prepping up for like monthly summaries inside of Automation Town and monthly summaries inside of realize is like, really, it's almost cathartic to write it down because you, you start to think about like, is this something worth my time? And most of the time it's not. Or it's overly technical, like to API, to code, to whatever. Because that's where a lot of the advancements are happening. And all we really want is should we pay attention to this and if we should, is it worth making a change in our firm? I think people are exhausted.

That's a crazy valuable service to accountants is filtering the insane amount of news for like specifically accountants and what's relevant to the leader, what's relevant to the team. You were telling me before this about a tool that both of I, both of us love, but like a free lifetime access sort of version of it.

Now there is one. I mean, that's the issue. It's like you're spending a monthly dollar amount for this transcription tool and most Companies, most, sorry, most laptops and computers have a dictation function. So why would you pay for it? Right.

So that is. I'm talking to my computer and it's turning what I'm saying into text.

Exactly. So typically, those are not performing as well as they could. So a whole bunch of companies have popped up. And you can now have that text be processed through AI to be able to have it fix the grammar, insert paragraphs, have context about what app you're actually on. So it will write for that app. If you're on WhatsApp versus email, it might actually format it differently. So the tool that just kind of popped up is called Voice, Inc. And what it does is it lets you pay 25, 35 bucks a lifetime. So, so you pay it once and now you have it, and they have a free trial. And I went through the free trial and eventually switched off of the tool that I was paying, you know, 10 bucks a month.

What were you using? Who are the leaders here right now?

So I have a Mac, and Macs typically have more access, have higher access to these types of tools. So I was using Whisper Flow to start with, but then I switched over to Super Whisper. And there's different elements where some are only taking transcriptions on your local machine and not using anything online. And then things like Whisper Flow or.

Online only, like, Whisper Sync is the other big one.

That's right.

This is like, if you haven't played with this yet, it's still a big trend. It kicked off maybe a year ago, but. And dictation's always been a thing, speech to text. Like, computers have always been able to do a version of this. It's just like, it got really good all of a sudden, and it's kind of profound. Like, I've talked to a number of people where their biggest computer use habit that's changed in the last year is they type way less and they just speak to it way more.

I find that when people are in ChatGPT or Gemini or Claude or Copilot or whatever LLM they're using, they're exhausted because they have to write so much. And it's just to prompt the AI to create a better prompt. So why not just throw as much human text as you can versus computer text, as much text as you can, into the. Into the thing. Let it figure out what you're trying to say without you having to type really proper.

Yeah. So there's almost like this intermediary step where I speak into the thing and it records it and it processes the audio file, but you can instruct it to process the audio file different ways. Like you alluded to this earlier, you can have a specific prompt for Here is how I want the tone of my emails to be. And it will take kind of your rambling statement that like the way we talk actually doesn't always translate very well to written text, but it can contextually, based on your instructions, make the output like, here's how I want my emails to be because it sees that your email app is active. Or here's how I want it to do it. If I'm using ChatGPT like Jeff is, just give it the whole version of whatever it is that I said. What does your setup for that look like in different apps? Do you tailor that?

Very much so. I used to have everything sort of localized and just run it. But now ever since switching to voice sync, it has a new sort of setting where it has really great paragraph recognition. So I'm using that a lot more. So that's my setup. I just, I change the hotkey to the function key the bottom left on the keyboard, press it once, it works and it's good. So I also find too that when I'm on calls with people and we have to hop on an AI together because I do that a lot during sales calls because we're trying to be very, you know, pro.

Sure.

AI. Okay. I didn't know we're using them. I start, I'm like here, just wait a second. And then I talk while I'm on the call and then that shows and inevitably that is the focus of the call. Like wait, I need to sell accounting.

You gotta show me the thing.

What did you just do? So that's, that's kind of fun setup too.

I completely had this experience. I made a video, I was demoing something in Claude. It was like early days cloud artifacts. It was a very, very nerdy, very like specific AI thing. And I just made it a quick little 10 minute loom. But I instructed it with this speech to text tool. I think it was Super Whisper. Super Whisper is what I'm still using. And so I, I tell it what I want and it generates all the, puts all that text in there. 90% of the questions about the video had nothing to do with I was what I was talking about. They're like, how are you talking to your computer? And you show me how to do that. That's amazing.

Yeah, that's good.

So the headline here is we've got a cool lifetime access version of this thing.

Yeah, And I think in a world where all of us are paying so much money per month for subscriptions, it's a breath of fresh air to say, you know what, this is going to become commonplace. The makers of this tool realize that this feature is going to be built into a lot of the tools that are used anyway, so you might as well grab it while you can.

Yeah, like ChatGPT has a very good built in version of this. As I, as I said that, I'm like, I don't think I actually use this for ChatGPT. I think I use ChatGPT's built in one usually.

Mm.

You touched on something. If I can take us down a totally unrelated rabbit hole, let's do it. The stacking up of the monthly subscriptions, like our tech is getting so good, combined with tech companies wanting to like personify the role that their new AI is doing, like, oh, this is now, this is gym. This isn't just an incremental improvement. Like this is like an employee replacement. All of this stuff, it's just like driving up the cost of how much people are spending on software. Do you think the future of accounting firms is just like an outrageous software bill with a little, little baby human layer around it?

I mean, it's really hard to speculate because every week new things pop out. But the way I kind of look at it, there's two paths diverging. There's the mid market and up path, which is going to try to replace a human. They're going to say, you know, for 60 grand, this is something that will give you the power of 10 people. Yeah. And that will land with some people. And unfortunately the people making those decisions will believe that and probably cause more harm than good. And then there's going to be other tools where they're reasonably priced and they will be a stretch for people. But that stretch won't be so bad because it'll enhance the people's ability to do current work and give them more capacity. Hopefully they can enjoy their job a little bit more. I think that's the winner for me is the enablement, the smaller price increases. And then when they have essentially some experience with that $200 a month, $1,000 a month type of tool, they're going to recognize its value because they don't have to hire another person, but they can keep the team the same size and we'll probably see a lot more of that.

Did you say in your intro that you run a firm? Have we called this out?

Jason doesn't listen. He's like a one track mind.

Chad runs a 50 or 60 person firm.

Has your software bill as a percentage of revenue gone up since chat GPT came out?

No, it's gone down.

Really?

Yeah.

How?

Tell us how.

We have a very, very low percentage of tech spend. Like it's like between 1.8 and 3% some what, it's insanely low.

And I've heard average is five.

Yeah, four to five is what is what I'm seeing in community and with other, with other people that share this number. The reason we have it so low. And this includes, yeah, this includes all the, all the spend for the firm.

I'm not like client ledgers. Not.

No, no, not client ledgers. So if we take the like what I spend on, I'm the big spender in our firm. One person. So I've got, you know, the $200 Claude, the $200 open AI, the Ultra, like all the things that you can have so that you have full access and unfettered access and not limited access.

Some things so that you can impress clients on sales calls specifically for that.

Or it's just to get into this fancy room that you have just to be able to talk to you about it. But inevitably it's really nice to be able to try new features when they come out. And I don't think you need everybody on your firm to do that. You need to have some people that are power users like the AI champions you talked about to assess. And I think that assessment kind of falls on me because I'm looking at it every day. But none of the rest of the team needs to do that. So because we're a Google firm, everybody's using Gemini and we have about. So it's a 60 person firm. There's only 10 people that have the ChatGPT teams and they are typically in the management level and they're people that are using projects for really sensitive and important things. And that's good enough, right? It works.

So if you follow my main YouTube channel, you know that about six months ago I ditched my traditional bank for Mercury. Mercury, now sponsor of this episode, actually made the move before they sponsored any of our stuff. Mercury, over 200,000 companies now trust them for banking and managing their finances. It's everything that you wish traditional banks were like good access management, not having to do two factor authentication Olympics to get into your client's banking stuff, which is by the way like a major security risk to your firm. When the strategy to get the info we need is to tell the client to Forward you the code that came to them in a message that says never share this code with anyone. And we, we feel dumb about it, right? We're like, yeah, no, really sorry for the convenience inconvenience. Don't be sorry. Tell them there is a better way. Mercury. It gives you firm level access where then you can give the members of your team granular access to whatever they need, even down to like specific bank accounts. Much better from a security standpoint, but also like from a workflow standpoint is going to make your life, oh my gosh, so much better. Now Mercury. They will literally pay you money to put your clients on Mercury. Thousand dollars referral bonus per eligible client. Terms and conditions apply. Of course you get Mercury Pro for your firm. It's worth checking out. It's really good. Mercury is a fintech company, not a bank. Terms apply for bonus eligibility. See Show Notes for details. Totally worth checking out. I really like Mercury. Get all that down in the show notes. In a world where there's more just AI over hyper Y than ever before and you see a new landing page every day that's like, whoa, this sounds exciting. But then you can't actually tell if there's anything real going on there. When all that's going on, it's nice to know what things actually work, what are real, what firms are like actively using and having success with. I'll give you one tool, Stanford Tax buddy. They were kind of first aboard like the AI Tax workflow hype train, but they're now, I don't know, three or four years in. I know probably well over a hundred accounting firms that love Stanford Tax is a really, really good solution for better organizing your 1040 work. It gives you a single place where you can pile all those documents in. It creates this like AI checklist based on the prior year tax work of exactly what items we need from the client. Then I can just chuck all those documents into one place regardless of how they come in and it'll automatically check those items off the checklist and then give me like an actual like work paper in the cloud as an alternative to an Adobe Acrobat PDF. Whether your clients want to work with another portal or not. There's a lot of people that are just now using Stanford Tax internally because it saves you a pile of time on the just the managing of the.

Documents and all that stuff.

It's great company, they've been around an accounting solution that's actually really, really good. Worth checking out if you do 1040 work. Learn more about Stanford Tax Check out the link down in the show notes.

Let me take this rabbit hole into another rabbit hole. Well, what is, as a Google firm that's closer to Gemini probably than most people are, and the whole, like, is this better than that? Like, I think everybody agrees Copilot is way off. But where. What's your sense of where Gemini vs ChatGPT is? Now?

If I had to use numbers, there's. There's almost no debate right now that the usability and value that people get from ChatGPT and OpenAI's ecosystem is probably at the highest end. So if you worked backwards and just said, hey, the ChatGPT is a 10, you're probably going to say, Claude's an 8.5. And then you're going to go down and you're going to say, well, I've been liking Copilot ever since they started entering GPT5 inside of it and using that. So maybe if it was just for chat and working through things like five and then with Gemini, it fits sort of in like the 6 range.

That's all really?

I think so.

Oh, wow.

I think it's really good technically, incredibly good for technical things, but for writing and for. I'm only talking about images and things, but for writing and the feature parody, it just, it misses out. It doesn't have like, the instructions and the knowledge the projects have. Gems aren't there yet. A lot of people will look at NotebookLM and say, okay, well, that's how I throw stuff in and have conversations and create funny podcasts with. But that's not really like daily use. So it's good enough to take somebody from nothing to something. And if you think about the majority of the people in your firm, they're not AI nerds. They're not spending all their time in the tech. They're doing the work. So for them to be able to say, you know what, I would like to try to help get help with the spreadsheet, or I would like to upload this sensitive client file that I have all the stuff within work for a payroll allocation or something like, I'm okay with that. I can try that. But if you did the exact same thing in ChatGPT, you're just gonna have a way better experience. So I'm okay with the majority of people having the six because it does create good emails if you train it enough. It still loves EM dashes like everybody else, but it's not enough pain to put Everybody over to ChatGPT. And I think, like, your advice of, hey, just if you can just put everybody into the team's plan. You're gonna love it. It's great advice. But for some clients, I think this maybe goes to the spend a little bit. Like, we save a lot of money and I don't think there's a big enough delta for the amount of people that are using it to justify it at 60, but I think for 10, 20, 30 maybe, it probably is.

So do you think the incremental value a power user gets from ChatGPT versus, say a Gemini, do you think the delta for them is bigger than for a more beginner user? Yeah, I have this conversation often with people around Copilot where they're like, well, we've, we have access to this and so we just use this. And I struggle with, is that a justification for them to have the minimum, cheapest tool, or is the opposite actually true that these people's time are so incredibly valuable and accountants are so expensive that they are worth more than to use anything but the best, the very best tool? Like, how do you feel about that?

I think I agree with you completely, but my actions don't follow along with the belief and I think it does, it does come down to price a little bit. So for us, I still think, even though we're two years into tra, like educating our team on how to use things, the upkeep, what's the word? The uptake on daily AI use and working through things is still pretty average. People are still testing it, they've got their work to do, but they're not power users. So will giving them that tool make them power users and will that make them better at their jobs? I don't think so just yet, but I think for the power users, that's why it makes sense for us to give them what it is. So I think there's. It's okay to live in both worlds. Honestly, I think if you have the money, you should. We've just been very fiscally responsible, hired a new controller. Hi, Brian. And it's been great. Hi, my Brian.

Love you, Brian.

Thanks, Brian. Yeah, it's nice to have someone looking at the costs.

Okay. 1, 1 3rd rabbit hole and then we're back on trial. Care do you find in practice as folks are like getting up to speed? You said the uptake is like, it's, it's nothing crazy. Like folks are getting comfortable with it. Is that a huge spectrum? Like, have you had people that went wild and are like doing amazing things and people on the other end of the spectrum or kind of across the board. Is it middling?

Good question. I think, like anything in life, there are going to be outliers. And for us, the outliers. I can think of a specific bookkeeper, a specific manager who didn't have access to the ChatGPT stuff, and another person on another team. Those three people showed a lot of interest, and that interest flared into their work, and they became power users, and they're just like. They're still flourishing. So I think where you're going with this, which is right, is like, why not just give them the full access to stuff and maybe they'll, like, turn it. I think it's because not everybody is like that. And like, with human nature, there's going to be people that really shouldn't think that the be all, end all is your job. And the reason why we have our firm is because we want people to have as much of a meaningful life outside of work as they do inside. So we're sort of embracing that and not forcing them to do things. And it sounds like it's a money play, but really it is a let's kind of move up together play. And I know that's like the most political way of skirting around the money issue, but it's. It's the reality, I think. I've. I've had this conversation with almost everybody on the team, which is like, if we should. Sorry. If we should give the team members the ChatGPT teams plan, we should, and, like, let me know when that's needed. And it just. It never really pops up.

Yeah, it gets into a hard, tricky thing where you're, like, picking favorites and it's. There's some selectivity there, and that's hard, too.

Yep.

Okay, so Voice Inc. That's what we're talking about. The Voice Inc. Free lifetime or not free, but once lifetime, you buy it. Once you have it forever. You actually said lifetimes when you first said it, which got me thinking, like, similar to months. Like, I'm assuming I will only have one, but who knows where AI will end? So that was a big one.

That was good.

And we'll link each of these down below. What else has been big in the last 30 days?

Yeah, we should probably go quicker on those ones. You know what the next one is? Fishbowl. And the inventory management. Yeah, the inventory management. And normally when you think about AI advancements, you don't think about inventory.

I try to think of inventory as little as possible.

Me too.

Yeah.

So I got a demo of their new AI and what they've tried to do is say, okay, inventory isn't so mechanistic and boring anymore. There's actually a lot of pieces that go into demand forecasting, into offering alternatives to customers, you know, buying from different suppliers if there's issues. And like no one's really taken that I know of. I'm not a big manufacturing person, but they've said, hey, based on all of these inputs, let me architect an incredible alert system and demand forecasting system and allow you to be presented with options and recommendations instead of you having to go do that yourself. And I feel like that's a perfect use case for AI where you have multiple sources of different types of data, a brain and instructions on how this stuff kind of interacts with each other and then output, which gives people actionable, useful tips on how to run their company. So good on Fishbowl. I think it's good to be determined if it's as good as they say it is. I, I hope it is. But seeing that type of innovation is another reminder that apps and their innovations are where we're going to see our biggest gains. Not necessarily us vibe coding and building our own things.

Yeah. So agent kit, like visually it's really cool. It's kind of this visual agent, like multi step workflow builder. How realistic is it though that a normie can use that?

Not very realistic.

A day after it was released, we held sort of like a live session inside the community and we just like all the nerds came. It was incredible. They brought it up. A bunch of them had already tried it and just listening to how they had to architect set up and use it like it's so far out of the reach of a normal person. And then even if you did get to it, there's elements that are not as good as other agent builders or tools and it's not as progressed as others. So they're like, you know what, we'll just kind of wait and see where this gets developed and go from there. But it is kind of cool to see people like, like ramp using it they're using on their policy agent.

Yeah, they've got a cool kind of like you give it the, you give it to your company spend policy and it'll automatically approve or deny expenses and even show the user in the expense policy. Here's why it got denied, which is this cool tech thing, but also saves the human from having to be the jerk.

Oh, it's such a good. This isn't me saying no. This is their main. Yeah, saying no. Yeah, immediately.

So I, I, when I first saw.

The agent kit stuff that like the.

Visuals are all really cool and it's like cool. This like modular sort of AI agent.

Builder sort of thing. And I think we all want that. If that's not the most accessible version of that, where would you send somebody that's like, I'd love to build like a multi step AI agent thing.

The two most accessible places right now are things that you've mentioned before. Zapier Agents and Lindy.

Lindy. Okay.

I think those two are the easiest right now. About 60 or 90 days ago, Lindy released something called Lindy Build and they're agents. And even though the pricing is a little bit high, if you get it right, this is one of those situations where if you had to spend a hundred dollars on something that is always accurate and always on, I think this is where like you can get comfortable with the extra spend. But the first time you do it, it's a little uncomfortable. So the thing you have to watch out with Lindy is that as you're building, you use these credits up and if you're making a lot of mistakes, you run out of credits. It could cost you a lot to build this agent, but it's very drag and drop and very accessible and really easy to use. With Zapier, they're leading the way with like accessible agents. So type in what you want, it builds it. You know, a lot of people are doing email agents and things like. But essentially because you already have all your authentications into your tools inside of.

Zapier, you're already connected to all your apps.

It's so easy to just say do this, do this, do this. But I also believe there's this over romanticizing of happening with agents where like, oh, I'll just build an agent to do that. When really what they need is a workflow or an automation that is linear, that just does something. So I only ever build agents when a workflow or an automation, like an AI workflow or an automation, can't actually do what I want it to do.

Yeah, because it's kind of the less durable option. So the, so the theoretical excitement around like a modular agent is kind of the intelligence in a box notion of it. So you brought up the example of people building email agents. The idea that you can have an email agent that, that triggers every time you get a new email. And then you ask the AI model, is this email a newsletter? Is this email a new lead? Is this meal? Email an existing client and build like kind of these logical. Then okay, if it's this, then do this with it next. And we've had like workflow abilities like this in the past. But the power of a large language model is it can do some things that those more deterministic workflows could not like. You know, I think that's probably why people are starting with emails because it's like a body of text and AI can kind of make sense of it and make decisions based on it.

Yeah.

Do you have any of that sort of stuff that you that is like running right now that helps you every day?

I have over like 400 automations in Zapier and make that do all kinds of stuff for me in the firm. But they're not agents.

Yeah.

And I think it's because I fundamentally don't have a need for it because a lot of the, even the sentiment analysis or hey, an email comes in and check it and do this thing. You can do that in a zap, you can do that in a make automation without needing it to be an agent. So that's where a lot of those things. I still have old automations running because they work. I don't want to touch it. The short answer is every agent that I've built that is there for production has always let me down in one way or another. So I have smaller, more reliable workflows that work. And my favorite automation I have right now is completely manual, which is the dumbest thing to say. But after a meeting transcript I have other stuff that will take a transcript and then turn it into a bunch of stuff. You did a lot of really good prompts for that. If you haven't seen that, definitely check out that link in the comments below. Can I say that link in the shout outs there was. Yeah.

And.

That'S fine to like do like repeatable stuff. But after a sales meeting ends, I take the transcript, not the summary and I drop it in a project inside of ChatGPT and it has some instructions and some examples that I like first like summary emails. But I get it to probe exactly like what's missing so that I have like the perfect summary email that goes out. So I drop it in. I just dropped some screenshots from the spreadsheet that I have even though I know I could create an automation to that spreadsheet because it's in Google sheets. Drop some screenshots of some data points that are really important and maybe just something to add a little bit of color about the call and like how I'm feeling about it. That's outside of the transcript Press enter, comes back with a bunch of questions again. Here's a good example of why you would use the $200 a month ChatGPT plan. I set the thinking option to heavy and then it runs for five or eight minutes and then gives me almost a perfect summary email of everything I need about services included, excluded issues, pricing, all that stuff. And I'm okay, that, that's a little bit manual because I want to have that chat interaction. I don't want it to be so mechanistic that then I'm like, well, this isn't what I wanted.

Yeah, you said it was a manual. You're, you're chucking stuff into a ChatGPT conversation.

Yeah, project.

We're actually working on a YouTube video of how to use AI for putting together your proposals, using transcripts, stuff like this. You've talked about this before. Can you go to more detail on that? Like, does this go as far as it knows the services you'll offer and like map the conversation to a collection of services that you would build into a proposal?

Cool. I mean this is also a really good, good call out for Anchor because they're, they're one of the only ones that has an API that allows you to create and send proposals.

Okay, right.

Other ones don't that I know of. So I think that's really powerful from an automation perspective. But for me, if we go back to that conversation of, well, I just don't trust it. I think you're going to have that trust issue with proposals. So for me, what we did was we don't laugh at me, but we created this Google sheet that essentially has 250 data points for every entity in a sales call. Some of it's applicable, some of it's.

Not, but it's all potential drivers of.

Yeah, yeah, exactly. And from there the word I was trying to look at is like, it's all indexed, so every, every field has an id.

Okay.

And then we've used Google Scripts, Google Apps Script to take all of that very structured data, press a button inside the menu, and then it will create the proposal perfectly every single time. Because we don't necessarily need to trust an agent for that. So then now I have this beautiful proposal that's in Google Docs because it works perfectly with Google Apps Script. It has all the tables and the pricing and the services and the preamble and the terms and conditions and everything from just a template Google Doc that it takes from. And then that is the PDF proposal. And then my summary email that essentially goes into even more detail than some of like the stuff you put in a proposal is what I send along with the proposal. And that used to take two hours for every lead. Dude, proposals are a lot of work. It's so much stressful. So because it did that, I can get my summary email from just the screenshots doing it, waiting the eight minutes and go through a couple iterations, 20 minutes ready to go. The proposal before it would take us just an hour to get the proposal the right way that we needed it. Because there's so much many entities and so many things. Now it's a button press. It takes roughly a minute and a half to process.

How much tinkering do you do with the. Out with the output? Like, I would still have the. I would have the same concerns of like, oh, what is it?

So we've been tinkering since I can count, I swear, eight months. So as soon as we started that, eight months later, we're still tinkering a little bit. Because every time you see something you want to improve on, you're like, do I improve the script or do I just improve the template? And you're like, you're trying to figure out where you want to make the improvement. And that's the hard part.

And that's how you end up with such a long list of potential inputs. Yeah, okay, that's super cool. So there's like a. There's a basic version of this that we can get to with some prompting. There's. This has kind of more infrastructure involved. I imagine that's something you talk about inside of automation town.

Yeah, this is the kind of exact kind of stuff. I think what's kind of nice is that people will come up with their own issues. They'll throw it in like, hey, I don't actually know what the. I actually don't know what the next step is here. What do I do? Yeah, and this is why I kind of how I kind of learned about the. The anchor API stuff. Did an interview with them, posted it inside of automation towers. Like, oh, I had no idea they had an API. And it's like one of those things where you do have to contact them to get access. They have to approve your use case. Such as, like everybody can get it, but if you can get access, like they are very pro app, ecos app infrastructure. They're very pro collaboration and integrations.

You know what's the worst, man? When you hire somebody and they start working for you and you're like, I thought this person knew stuff. The number of times I've been hoodwinked by somebody that interviews well and they have umpteen years of experience and you pull them in and you're like, you don't what? You don't even know your stuff, man. This isn't going to work. Let me tell you what one of this episode sponsors Cloud Accountant Staffing is doing about that. So Cloud Accountant Staffing help accounting firms find talent around the world, be it the Philippines, South Africa, Latin America. And they have an online portal where you can go out and see folks that they have recruited, schedule interviews with those people directly on their platform. And one of the things now that every single one of those people have is a collection of assessments. I'm looking at one right here. Assessments for business tax, for individual tax, for critical thinking, for understanding instructions, their personality type, and so many of these things like individual tax, this person I'm looking at, they're in the 93rd percentile. And so cloud account staffing, they have a huge volume of folks that they have go through these assessments and they hand pick the highest performers and they serve them up for you to make your wildest dreams come true. Cloud Account Staffing's Backstory they actually ran an accounting firm, hired some offshore folks, had an amazing experience, and then just started helping other accounting firms do the same same. Interested?

I thought you might be.

Check out the link down to Cloud Account and Staffing in the show notes. If your firm still runs on spreadsheets, sticky notes, or a half a dozen different apps to manage work, there is a better way. Enter this episode Sponsor Financial Sense it is the all in one platform for accounting firms to manage everything. Workflows, client requests, capacity, billing and team collabs all in one place. Automate recurring work, track deadlines and get visibility into what your team's working on. Built specifically for firm owners who want less chaos and more control. That's actually an important detail if you're the type of firm owner that wants more chaos, or if you want less control. This one's actually not for you. Now this December 12th to 13th, Financial Sense is hosting WorkflowCon, a free virtual conference built around the theme Clear the chaos, find your calm and meet. I'll be joining for a live Q and A on the subject of smarter client comms, but but comms of all kind using AI alongside a bunch of other firm leaders. It's pretty like Rockstar. List of speakers they got. Register for Workflow Con via the link below. Start your free 14 day trial of financial Sense using the link below with coupon code jason10 grab all that stuff down in the show notes we should.

Take on the Atlas browser. So the new browser from OpenAI that's kind of like, it'll do some agentic stuff. You can talk alongside the web page. Like, is that interesting yet?

I got up on stage and gave a little bit of a lecture on why you shouldn't be using stuff like that. Okay. I don't know how that actually went down and how people took that. Yeah, but you also did a video on Atlas browser, didn't you? So you probably have some, some opinions. For me personally, I, I recognize that if really smart people are saying be cautious because the ability for prompt injection inside of images and in text is quite high. And naturally if there's just something that knows everything about everything you've ever said in ChatGPT and can then go do stuff and then can be influenced, isn't that the worst version of the Internet that you can think of?

Yeah. Doesn't this just the next decade though of struggling with this issue? How can I trust this thing to do the thing? How similar is this to trusting other humans to carry out your will? Like employing other people and them not doing what you expect.

And maybe because as humans we think we understand other humans, they work, but we have no clue how these computers.

Are really thinking, so don't use it right now, don't use right now. I think to me it doesn't rise to the level of usefulness yet to where you have to have that really hard kind of consideration.

Babysit it.

Yeah.

And that like that's kind of where.

I'm at right now with agent mode and chat. GBT is the biggest risk in my mind with injection. And so injection is like instructions, malicious instructions being fed to the AI somehow by stuff on the website that you can't see that you won't see there. It's a problem on the larger web if I'm giving it a specific job to do inside of an app that I already trust doesn't feel like a huge issue to me. But regardless, in a read write environment, I still have a hard time walking away from the computer at all. But that's a read write environment. Like I've heard the example of like people having access to a paid research library, like researchers like, and it can go out and pull all this stuff and they cut the agent loose inside of this paid research library that they've logged into and all I can do is just fetch helpful information. Like awesome. That seems like a great agent use case for me. Yep. Doing offline stuff with agent mode. Like, agent mode is way better with spreadsheets. Give it a spreadsheet, have it do a thing. Like, that's amazing. Like, that's just a file that it's doing stuff with and giving it back to you. But, yeah, this just feels like the tip of the iceberg of can I really trust this AI to go out and do XYZ and the horror stories that are going to be in the news.

And, yeah, I think you will start to see stories and it'll. It'll. It'll scare a lot of people. And I think with all tech people that have not been exposed to the risks are going to be hurt. Yeah, probably the fastest. And it's no different than. It's a weird example, but, you know, people are still getting scammed over the phone and this is just. I'm not saying this is a scam, but, like, it has the potential to be. So I'm glad you brought up the local stuff. I'm glad. Glad you brought up the, like, the guardrails, where you only have access to a certain thing and you can't go out and do other things. That's great. I think the more we can lean into those, the better. Which kind of brings me to the next app. Unless you want to go back to the OpenAI.

Only other thing we didn't cover on OpenAI is apps inside of ChatGPT.

Yeah.

So they make this big announcement and it's always like a collection of apps like Travelocity and Instacart and booking dot com. Plan my European vacation.

Like, who's ever going to trust an.

AI to plan your European vacation? Some people do, but basically it's just like you can at an app. And if the app's connected to ChatGPT, you can see into some of the data in there and it's on their new MCP protocol. I understand. Does this matter? It doesn't feel like a thing that the accounting vertical is going to be supporting in ChatGPT. So should we care?

I don't. Not until our tools are connected.

Yeah.

Actually, I brought this up at the session here yesterday as well. And it's like, can you imagine a world where your instructions, the same time you're looking for a recipe for carrot cake, the next one you're like, oh, by the way, can you put the expense for this carrot cake into QuickBooks and add QuickBooks? And like, that's not far off. But it's not.

Now, those are totally the examples they give to, like, it's my friend's birthday and they're in the anime Find me a recipe for xyz. Here, here are the, here are the ingredients that I have. Actually, I, I, we're having people over and I have a very limited, like set of cocktail supplies and I'm like, what can I do with this? That was great. Yeah, that actually was amazing at that.

But it didn't taste very good. But it was probably amazing.

It wasn't going to taste good either way.

No, no, no, no, that's a good point. So if we move on to the next tool, it has everything to do with giving accountants control in an environment where the agents aren't going out into the wild Internet. And that's an app called Artifact.

Artifact.

Get artifact.com. so the idea with Artifact is they released this thing called Omni Workflows. What I really like about tools like this is that they're not just hiding behind the words like, oh, we are AI powered. They actually say, hey, if this task inside of your accounting ledger or your business is important, we're going to give you the infrastructure to create an agent. It's going to have guardrails just like ChatGPT's thing we just talked about. And you know, for a couple grand a month you can have a couple agents running for a 20 person firm. And if it works, it works. You got us for support, but like we're here for the accounting industry and we want to make things work for you. Like, I feel like that's a pretty accessible option for people that aren't hiring people full time or on the side to build all this stuff in your own environments, but in an environment that's presumably more secure than OpenAI's browser or agent, that kind of stuff.

Yeah. So it's just for accounting firms.

I actually don't know if it's just for accounting firms, but presumably it is. Let me just bring it up for a sec. Yeah, it says the most advanced AI for accounting and finance. I mean, I wish people wouldn't just talk in absolutes that much, but they have. A lot of.

You hated it as soon as it came out here.

Yeah, like there was, there was another thing I was looking at. I was like, yeah, whatever, it, I didn't hate it as soon as it came out, but it was, it was, it was worth looking into. So like they've got, hey, create your own client health dashboard and you know, event by event. How is this actually working in your accounting? Is it like looking for anomalies and alerting you and your financial reporting, like it's, it's, it's giving you all kinds of, of insights but it gives you the capability to create it yourself too.

Oh, okay.

Like so that's, that's the real win here. Which is, which is essentially you know, orchestrating your own agents and.

Customer reconciliations, getting client data collection, kind of like a content snare kind of thing. Yeah.

So we're at Connect. They just did a mainstage talk talking about kind of the whole plan of all this cool AI stuff that we're going to do. And it's simultaneously exciting but not real yet. Like what are. If you had to hand pick maybe the couple things that you think feel most real, that people should be excited about, what would they be?

I think you nailed it with like.

I could see the disappointment in your face as soon as I said that. That feel real.

Here's the thing. You know these, these billion dollar companies that are moving a lot of data and a lot of things have to look at things from security and data infrastructure. Just have a lot of work to do and because of that I feel.

Like I'm about to be let down.

You said it yourself, it's like is not here yet. Right. So we're getting teased with a lot of the stuff on stage around what's coming and that's things like in their upper mid tier market, this Intuit Enterprise suite. Right. The idea of multi entity accounting being able to look at margins and give it price lists and look at inventory and you know, create the most optimal profitable purchase and sales price, like that's all really cool.

They're talking about doing all that with AI.

Yeah.

Okay.

Right. And, and you can do it right inside of QuickBooks.

Okay.

Right. But it's not here yet. So it's really hard. It's really nice to get excited. I think there's a, there's a great possibility. I was talking to somebody earlier, I said look, if they do what they say they're going to do, this is really good for the accounting world. But we recognize that things happen in stages and not everything is complete when it's launched. And a lot of people write off tech that they've tried even though it is getting better every couple weeks. So I think with anything and this is in particular like try it. There are agents now that exist and there will be a couple newer agents that are released soon, especially around payroll just went live and there's more accounting agent insights that are pretty interesting and you can give those a try. But just remember like this is the first version and things will always get better. And don't write it off if it's not perfect in the beginning, because if you do, you might miss out on something six months down the road. That happens. And just because you saw a little insights button and you clicked it a few times four months ago and it didn't do well, doesn't mean that you shouldn't click it in six months.

Yeah, yeah, that's hard. You have a bad, like a negative first experience and you kind of write it off. Is there anything there that ought to be changing our behavior now? Even though it's not real? Is there anything that ought to be impacting the tech decisions we make and all that?

I think so, because right now we're starting to see that whole build versus buy conversation change a little bit more towards buy. Like most people really want to buy software. They don't want to be out there building it. And the stuff that they are building, it's for edge cases or it's for very specific things. It's either overpriced for for someone else to build it with you, or it is incredibly expensive. You just spend a lot of time on it. So I think the behavior that's changing is that people are starting to trust buying tools more now than trying to code something themselves.

Yeah, they could see they were working on it.

Yeah, I think so. So, you know, I think a few years ago the conversation was, hey, can we hire somebody to build those apps?

Yeah.

Now it's more like, hey, can I hire somebody to consolidate our information such that new tools and infrastructure can use that to be able to have more effective use cases for AI.

Yeah.

Who's doing a really good job of that is dark horse.

Oh yeah.

They've got a. They've got a guy who.

Dark horse to contextualize. How do you describe dark horse?

I would say the brainchild of a madman who wants to upheave the entire industry and give CPAs autonomy and business models that just never existed before.

That clears it up.

Yep, absolutely. So essentially a CPA can become a dark horse principle and then the dark horse brand is used. They will give you sales leads, they might even give you a team. And then like you're part of an infrastructure without needing to start your own firm. But you are like starting with, you know, five steps ahead.

It's like a supported. The supported practice. Yeah.

Yeah. So with them, they needed data infrastructure and they needed centralization and they needed like to be able to support each one of the principles. So their one of their biggest first moves was to say, how can I centralize this all in a database? How can I create the context to be able to have all the AI work profitably or work accurately? So that's what they did and that's what they're working on. It's kind of cool.

Do you ever miss our podcast Automation Town? You're avoiding eye contact.

It's because. Because it makes me sad. I do, I do like that. Yeah, I miss it. Why?

Yes, I think it was a lot of work.

It was. You know, a lot of people don't recognize that. We paid a lot of money too.

To our editor since the professional sound designer. Like it was a whole thing. It was so good. Why?

Why do you ask?

It's just because we just don't. I don't know, we don't. We don't get to do this as much anymore. You know, I think we should. Okay.

You want to come to Spain?

No, that's been Chad Davis. Chad. So here's how I think of Chad.

Right. Correct me if I'm wrong. You got an AI champion in your firm or the person that's super curious about this stuff wants to discover more. Chuck them in automation time. They're going to be surrounded by a ton of like minded nerds.

That's Chad's community.

My community. Realize. Chad hosts a monthly session in there giving you exactly this, like riffing on the biggest thing of the last 30 days. Realize is more a community for firm leaders.

I would say I also host another hour session and realize where we kind of go deep on some automation builds.

That's right.

So a lot of members of the community have some cool stuff that they're doing and they bring those like in a show and tell types of environment. So that's another session.

Cool.

Chad, thank you for doing this.

Thanks, Jason.

A heads up for my podcast friends. We put together the biggest, scariest project my team has ever worked on. A big old run of live events around the country headed up until year end to. To ensure that your 2026 is just the best year ever. Let me run you through the cities real quick. We're coming to Austin, Oklahoma City, Phoenix, Seattle, Portland, San Francisco, Los Angeles, Atlanta, Nashville, Chicago, DC, Philadelphia, NY City and Boston with some super cool half day workshops that roll into dinners. You get CPE for it. It is an event that is built for your entire team to get the whole squad feeling inspired so that if you are a firm leader, you're not feeling like you got to put everything on your back.

Headed into 2026.

Gonna be a super fun experience.

Experience.

It's going to be a surprise guest at every city. We're going to record an installment of the podcast live at each of these events. You're going to be part of the podcast itself. What? And, man, I would love to see you there. To learn more, head over to jasononfirms.com events and I hope to see you there.