Show Notes
For the month of September, Scott has been running an AI bookkeeping agent in parallel with his existing accounting system and team. Nothing has been cut over. At month end, he's comparing both sets of financials.
The setup, in five steps:
Bank feed. BankSync.io (~$7/month) pulls daily transactions from connected accounts via Plaid.
Daily ingest. At 8am, Stu — Scott's bookkeeping agent — grabs the prior day's transactions, filters out anything already processed, and ingests what's new.
The queue. New charges wait to be categorized. Nothing posts automatically.
Human confirmation. Stu proposes a category and the rule it believes applies. Scott confirms or corrects. Only then does it post to the ledger.
Portability backup. The ledger copies to Dropbox, so the books can move to whatever agent platform comes next.
The rule: The agent doesn't spend money and doesn't post anything unconfirmed.
A second test inside the test: Grok recently released a Grok Finance plugin with read-only bank access — also via Plaid. Scott is comparing it against BankSync to see if both produce the same result, since Grok Finance is free.
On the CPA question: The accounting team isn't going away. The CPA still oversees the work. The question is whether the low-level coding work can move, freeing people for higher-value work.
Currently running three companies through the system. More advanced land-investing accounting is being tested separately — Accounting for Land Investors students should watch their class LMS.
Before you try this: Vet any repo you connect to your financial data. Check the maintainer, the update history, and what the code does with your information.
Got a business question? Ask Scott here: scotttodd.net/ask
📜 Full Transcript (Click to expand)
Scott Todd (00:49.74)
Accounting and AI, my two favorite topics. That's what we're going to talk about on this particular video. Welcome to Fix My Business. I'm your host, Scott Todd. I built multiple seven figure businesses after leaving my corporate 500, Fortune 500, actually Fortune 300 VP job. And I want to help you build a business that you love to do and not be constrained by it. Now, AI has come a long way.
It's come a long way. And when we talk about these AI agents, which I've talked about on this channel previously, we've talked about all the AI agents and how I'm using them. And I continue to explore the ways that I'm going to use it within my business. Frankly, I spend a lot of money, probably you do too, on accounting software, accounting teams, and they hold their place. They really do. But some bookkeeping is just simple. Some bookkeeping can be complicated. I'm thinking about like accounting for land investors.
But if you have a business in which, you know, it's really a little bit more simpler simpler, where you have just transactions coming in, you can record them. I have a cool thing that I'm testing I want to share with you. So for about a month now, I've been testing out this GitHub repo called slash books. So slash books is written to replace QuickBooks, Zero, whoever it is.
And basically the way that it works is that each one of your companies, you can have multiple companies, they each get their own files within your computer. Okay, so it's not like we're going to use QuickBooks or we're not going to use we're not going to use X Zero. What we're going to do is we're just going to have this AI agent be able to write and read the transactions that's coming in. And the way that we set this up is really quite simple. You can give the
The GitHub repo to your AI agent. I'll include that in the descriptions. You just give your GitHub repo to your AI agent. You say, hey, I want to set this up for my company. And it's called slash books. Now, what it's going to do is it's going to set up the actual books for your business. And you can give it the start date that you want. And what I'm doing is I gave it a start date of September 1st. So I've been running this.
Scott Todd (03:10.514)
Now for the month of September, and I've been running it in parallel with my existing accounting system. So I don't want to just cut off the accounting system and the accounting team. What I want to do is I want to run this in parallel and so that I can really look to see what it looks like at the end of the month. How did it perform? And SlashBooks will go out and it will create all your company financials for you on request. It will do all of your closings.
And I'm going to walk you through in this video how I've gone about and set this up and really how it works. So the way that it works is really five steps. First of all, once it's loaded on your machine, the thing that we're going to do is we're going to have our agent download the files. Now we need some sort of a bank feed for this. So at the beginning, when I started this, I actually needed to use a service called banksync.io.
And it's cheap. I think I pay like seven dollars a month for it. And ultimately you connect it to your bank accounts. And all it's gonna do is it's gonna pull every single day all of your bank transactions from the accounts that you have set up. So it's gonna download those to your computer. And then what will happen is the AI agent every day at eight o'clock, my AI agent's name for for bookkeeping is Stu. Every morning at
At 8 a.m., Stu fires up, wakes up, he grabs the transactions from the previous day, they filter them, they look at what's already been processed so that we're not getting double entries. And then what they're doing is they're ingesting what has not been loaded into the ledger yet. And when that happens, it actually goes into a queue where the new charges wait to be categorized. They're not posted, they're just in this queue.
And then what happens is I actually have this set up to where it's running on both pieces on two different two different storage devices. So let me explain. The first storage device is on Grokbot's main computer. So Grokbot or Muse, whatever you're using, maybe are you going to use OpenClaw or you're going to use Hermes. It doesn't matter. They have a computer there that they all operate on.
Scott Todd (05:32.666)
So in my case, I have Stu running on Grokbot. He has his own computer. It's going to take that and it's going to store that information on its computer. Then I also have it set up to where there's a backup ledger on Dropbox. So Dropbox basically allows me the portability that I can now take this to any other AI agent that comes along. And if you know anything about AI agents, guess what?
Tomorrow there's going to be a new one. In fact, I think it's today that I'm recording this that they're talking about how OpenAI is going to release their agents. They're going to take OpenClaw and make it whatever it's going to be. So we're always going to have these AI agents that we want. And so the whole thing here is that we want to build in our portability. For me, all of my files are stored in a Dropbox folder. They're stored there so that I get portability.
And then what happens is all of my agents, wherever platform they're on in the future, I have access to move them where I want to be. So ultimately what's happening is that fifth step is that the books get copied over there. Now what happens is every day at 8 a.m., Stu will go out, he'll grab that day's transactions, and he will give me a list of the transactions from the day. Now, AI
And Stu are learning how these things get coded. So it's building a rules about this. Hey, when I see this kind of revenue come in, I'm going to charge it to revenue. I see this expense going out. This must be payroll. I'm going to charge it over here. So the system is always learning, writing its rules, and it will show me of the eight or ten transactions that I have today. It will show me and say, hey, here's your transactions. And by the way,
Here's the rule that I believe that it is, and it's just going to confirm it. It's no different than seriously, it's no different than my CPA, who in the beginning they have to know what it's for, and then later on they begin to learn it and they start to code it on their own. So it's the same concept here without necessarily having the human in there. Now, I confirm the categories. Every morning at 8 a.m., sometime throughout the day, I'll go in and say, Yeah, you got it right, or no, change that one.
Scott Todd (07:52.681)
It will change it and then it goes ahead and post it to the ledger. The AI agent is told with strict rules, you're not spending money and we we're not posting anything unless it's been confirmed. And you know what? If we find out that something's not right, no worries. Now I mentioned that on September first when I started this, the way that I had to get the bank transactions into Grokbot is through this service called Bank Sync.
And bank sync has been rock solid throughout basically the the month it's here. But in the last week, Grok has also announced a plugin called Grok Finance that allows read only access to the bank connections that you set up. So they both use plaid. So it's no different. Your security level remains the same. Plaid is a platform that kind of links to your bank and to the whatever sources you're trying to sync that information to.
So at the end of the day, today we're now running another test where we're comparing Grok Finance to Bank Sync, because as I mentioned, Bank Sync is $7 a month for me. but Grok Finance is free. So we'll see if we can even reduce that further. But what happens is every day we're pulling these transactions, weekdays, of course, pulling the transactions in. It's it's suggesting the categories that it needs to write to. It's gonna write to those.
And at the end of the at the end of this little test that we're using between Bank Sync and Grok Finance, we're gonna go ahead and see if they basically produce the same outcome. And if so, we'll just kind of go with the the free Grok finance. Now, there are some other rules. For example, some charges still show up as pending on the bank account. Well, pending I pending transactions, they do not get coded until they actually hit. And what will happen is we will continue to kind of
measure this throughout the week so that we can kind of run that test. But at the end of the month we're gonna pull the financials based on what our CPAs pulled versus what our Grok bot pulled. We're gonna see how close they are together. And that's the thing is that can we leverage AI for this? Absolutely. But see I still want a human in the loop. But this is this is kind of the what's going to open up for for
Scott Todd (10:15.849)
for businesses is how can we leverage AI, not necessarily to replace somebody. And you might say, you're talking about replacing your CPA. Well, no, the CPA still has a role because they need to oversee the work, but can we reduce the low level bookkeeping work that people can do? And can we replace that with higher value stuff? So my accounting team's not going away. But if we can stop them from having to do some of the mundane coding work,
then life will get a lot, lot easier. Now I mentioned that that slash books can have multiple companies and it can. It can it can run multiple companies. And for more advanced accounting that we talk about like in accounting for land investors, well we're also starting to to do the testing on that. So for those of you that have taken accounting for land investors, hey keep watching your class LMS. We'll also send you a note.
Because we actually have some ideas there that of some skills that we can build in to help you make your accounting easier. But you can run multiple companies on the on the system. So I've got multiple companies that we're testing here. And right now we have three companies up and running through slash books, and it performs well. It performs well. So we'll see what happens at the month end.
We'll see how that goes through and produces our financial statements. And I can't wait to see how they compare to what our CPAs have produced. But this is something that you can look at doing. And again, if you want the prompt that we used, it's going to be either in the show notes or in the description below. But yeah, this is how we're using AI with accounting. My two favorite topics. If you have a question, be sure to head over to scott.net forward slash ask.
And I look forward to seeing you in our next episode. Take care.
Accounting and AI, my two favorite topics. That's what we're going to talk about on this particular video. Welcome to Fix My Business. I'm your host, Scott Todd. I built multiple seven figure businesses after leaving my corporate 500, Fortune 500, actually Fortune 300 VP job. And I want to help you build a business that you love to do and not be constrained by it. Now, AI has come a long way.
It's come a long way. And when we talk about these AI agents, which I've talked about on this channel previously, we've talked about all the AI agents and how I'm using them. And I continue to explore the ways that I'm going to use it within my business. Frankly, I spend a lot of money, probably you do too, on accounting software, accounting teams, and they hold their place. They really do. But some bookkeeping is just simple. Some bookkeeping can be complicated. I'm thinking about like accounting for land investors.
But if you have a business in which, you know, it's really a little bit more simpler simpler, where you have just transactions coming in, you can record them. I have a cool thing that I'm testing I want to share with you. So for about a month now, I've been testing out this GitHub repo called slash books. So slash books is written to replace QuickBooks, Zero, whoever it is.
And basically the way that it works is that each one of your companies, you can have multiple companies, they each get their own files within your computer. Okay, so it's not like we're going to use QuickBooks or we're not going to use we're not going to use X Zero. What we're going to do is we're just going to have this AI agent be able to write and read the transactions that's coming in. And the way that we set this up is really quite simple. You can give the
The GitHub repo to your AI agent. I'll include that in the descriptions. You just give your GitHub repo to your AI agent. You say, hey, I want to set this up for my company. And it's called slash books. Now, what it's going to do is it's going to set up the actual books for your business. And you can give it the start date that you want. And what I'm doing is I gave it a start date of September 1st. So I've been running this.
Scott Todd (03:10.514)
Now for the month of September, and I've been running it in parallel with my existing accounting system. So I don't want to just cut off the accounting system and the accounting team. What I want to do is I want to run this in parallel and so that I can really look to see what it looks like at the end of the month. How did it perform? And SlashBooks will go out and it will create all your company financials for you on request. It will do all of your closings.
And I'm going to walk you through in this video how I've gone about and set this up and really how it works. So the way that it works is really five steps. First of all, once it's loaded on your machine, the thing that we're going to do is we're going to have our agent download the files. Now we need some sort of a bank feed for this. So at the beginning, when I started this, I actually needed to use a service called banksync.io.
And it's cheap. I think I pay like seven dollars a month for it. And ultimately you connect it to your bank accounts. And all it's gonna do is it's gonna pull every single day all of your bank transactions from the accounts that you have set up. So it's gonna download those to your computer. And then what will happen is the AI agent every day at eight o'clock, my AI agent's name for for bookkeeping is Stu. Every morning at
At 8 a.m., Stu fires up, wakes up, he grabs the transactions from the previous day, they filter them, they look at what's already been processed so that we're not getting double entries. And then what they're doing is they're ingesting what has not been loaded into the ledger yet. And when that happens, it actually goes into a queue where the new charges wait to be categorized. They're not posted, they're just in this queue.
And then what happens is I actually have this set up to where it's running on both pieces on two different two different storage devices. So let me explain. The first storage device is on Grokbot's main computer. So Grokbot or Muse, whatever you're using, maybe are you going to use OpenClaw or you're going to use Hermes. It doesn't matter. They have a computer there that they all operate on.
Scott Todd (05:32.666)
So in my case, I have Stu running on Grokbot. He has his own computer. It's going to take that and it's going to store that information on its computer. Then I also have it set up to where there's a backup ledger on Dropbox. So Dropbox basically allows me the portability that I can now take this to any other AI agent that comes along. And if you know anything about AI agents, guess what?
Tomorrow there's going to be a new one. In fact, I think it's today that I'm recording this that they're talking about how OpenAI is going to release their agents. They're going to take OpenClaw and make it whatever it's going to be. So we're always going to have these AI agents that we want. And so the whole thing here is that we want to build in our portability. For me, all of my files are stored in a Dropbox folder. They're stored there so that I get portability.
And then what happens is all of my agents, wherever platform they're on in the future, I have access to move them where I want to be. So ultimately what's happening is that fifth step is that the books get copied over there. Now what happens is every day at 8 a.m., Stu will go out, he'll grab that day's transactions, and he will give me a list of the transactions from the day. Now, AI
And Stu are learning how these things get coded. So it's building a rules about this. Hey, when I see this kind of revenue come in, I'm going to charge it to revenue. I see this expense going out. This must be payroll. I'm going to charge it over here. So the system is always learning, writing its rules, and it will show me of the eight or ten transactions that I have today. It will show me and say, hey, here's your transactions. And by the way,
Here's the rule that I believe that it is, and it's just going to confirm it. It's no different than seriously, it's no different than my CPA, who in the beginning they have to know what it's for, and then later on they begin to learn it and they start to code it on their own. So it's the same concept here without necessarily having the human in there. Now, I confirm the categories. Every morning at 8 a.m., sometime throughout the day, I'll go in and say, Yeah, you got it right, or no, change that one.
Scott Todd (07:52.681)
It will change it and then it goes ahead and post it to the ledger. The AI agent is told with strict rules, you're not spending money and we we're not posting anything unless it's been confirmed. And you know what? If we find out that something's not right, no worries. Now I mentioned that on September first when I started this, the way that I had to get the bank transactions into Grokbot is through this service called Bank Sync.
And bank sync has been rock solid throughout basically the the month it's here. But in the last week, Grok has also announced a plugin called Grok Finance that allows read only access to the bank connections that you set up. So they both use plaid. So it's no different. Your security level remains the same. Plaid is a platform that kind of links to your bank and to the whatever sources you're trying to sync that information to.
So at the end of the day, today we're now running another test where we're comparing Grok Finance to Bank Sync, because as I mentioned, Bank Sync is $7 a month for me. but Grok Finance is free. So we'll see if we can even reduce that further. But what happens is every day we're pulling these transactions, weekdays, of course, pulling the transactions in. It's it's suggesting the categories that it needs to write to. It's gonna write to those.
And at the end of the at the end of this little test that we're using between Bank Sync and Grok Finance, we're gonna go ahead and see if they basically produce the same outcome. And if so, we'll just kind of go with the the free Grok finance. Now, there are some other rules. For example, some charges still show up as pending on the bank account. Well, pending I pending transactions, they do not get coded until they actually hit. And what will happen is we will continue to kind of
measure this throughout the week so that we can kind of run that test. But at the end of the month we're gonna pull the financials based on what our CPAs pulled versus what our Grok bot pulled. We're gonna see how close they are together. And that's the thing is that can we leverage AI for this? Absolutely. But see I still want a human in the loop. But this is this is kind of the what's going to open up for for
Scott Todd (10:15.849)
for businesses is how can we leverage AI, not necessarily to replace somebody. And you might say, you're talking about replacing your CPA. Well, no, the CPA still has a role because they need to oversee the work, but can we reduce the low level bookkeeping work that people can do? And can we replace that with higher value stuff? So my accounting team's not going away. But if we can stop them from having to do some of the mundane coding work,
then life will get a lot, lot easier. Now I mentioned that that slash books can have multiple companies and it can. It can it can run multiple companies. And for more advanced accounting that we talk about like in accounting for land investors, well we're also starting to to do the testing on that. So for those of you that have taken accounting for land investors, hey keep watching your class LMS. We'll also send you a note.
Because we actually have some ideas there that of some skills that we can build in to help you make your accounting easier. But you can run multiple companies on the on the system. So I've got multiple companies that we're testing here. And right now we have three companies up and running through slash books, and it performs well. It performs well. So we'll see what happens at the month end.
We'll see how that goes through and produces our financial statements. And I can't wait to see how they compare to what our CPAs have produced. But this is something that you can look at doing. And again, if you want the prompt that we used, it's going to be either in the show notes or in the description below. But yeah, this is how we're using AI with accounting. My two favorite topics. If you have a question, be sure to head over to scott.net forward slash ask.
And I look forward to seeing you in our next episode. Take care.