When a bank transaction lands in your accounting software, a machine usually gets the first look. A person gets the last one. That split is what we mean by human-in-the-loop bookkeeping, and it’s easier to trust when you can see exactly where the handoffs happen. Here they are, along with what we won’t do with your data and the questions worth asking us or anyone else.
Why we’re telling you this
Most bookkeeping companies now say they use AI, and most stop there. That isn’t enough for something as personal as your business’s numbers. You should be able to ask what a tool touches, who checks it, and what happens when it’s wrong, and get a real answer.
We also have a bias to admit. BooXkeeping is a franchise of local owner-operators, each one a Chief BooXkeeping Officer, backed by a national team, and we’ve supported 100+ franchise brands. People do the work here. But we’d be foolish to ignore tools that make those people faster and more accurate, and you’d be right to distrust us if we pretended AI does nothing. So this is the honest middle.
Where AI works in our process
Picture a hypothetical two-unit sandwich franchise. Each week its bank feed brings in card-processor deposits, a delivery-app payout, vendor payments, payroll, and one payment to the franchisor. Here’s what the software handles first.
Bank feeds and first-pass categorization come first. Transactions flow into QuickBooks Online or Xero (we’re certified partners with both), and each one gets a suggested category. The bread supplier, the utility, and the card fees look the same every week, so they sort themselves after the first few corrections. When a person fixes one, they can turn the fix into a rule so the same question doesn’t come up twice.
Receipts and bills get read for vendor, date, and amount, then attached to the transaction so the paper trail exists. The IRS’s page on recordkeeping is a fair reference for which documents are worth keeping.
Matching ties the $1,684.20 delivery-app payout to the sales report behind it. The report shows $2,210.00 in sales, so the software can surface the $525.80 that went to commissions and adjustments, the kind of gap that used to mean a calculator and a long evening. And anomaly flags catch things like the same produce vendor charging twice on one day before anyone posts it.
Each of these saves real time, and none of them is a decision. The software proposes. It doesn’t approve.
Where a person always steps in
Every batch of suggestions gets reviewed by a person before it becomes your numbers. They look at the exceptions first: new vendors, odd amounts, and anything the software flagged or wasn’t sure about. Refunds, split deposits, and payments that cover several invoices get the same treatment. Software offers a guess, and a person decides. Then come the parts software simply can’t do.
A person decides which month things belong in. Say our sandwich shop receives a $4,800 catering deposit on the 29th for an event on the 12th of next month. The software sees cash coming in. A person sees a deposit for work not yet done and asks how you and your CPA want it reported.
A person asks you questions. When a $612 charge doesn’t look like the business, they don’t guess. They ask, in plain language, what it was for. That conversation is bookkeeping too, and no tool can have it for you.
For franchise owners, a person also handles the franchisor. The bank statement shows one $2,310 payment. The franchisor’s invoice, in this made-up example, splits it into $1,905 of royalty and $405 for the brand fund, and your franchise agreement may require them to be reported separately. A person compares the bank, the invoice, and the agreement, then makes the split. Structures vary by system, so check your own agreement.
And a person closes the month. They reconcile each account to the bank statement, read the reports for sense, and sign off. Sign-off means a named human has looked at the result and stands behind it, so if a number is off, you know who to call.
“AI is my friend, not my enemy”
Our CEO, Max Emma, put it this way in a public interview: “AI is my friend, not my enemy.” We share that view, and we’d add what a good friendship looks like. You lean on a friend’s strengths and stay honest about their weak spots. AI is tireless, fast, and good at patterns. It’s also confident when it’s wrong, blind to your intentions, and unable to pick up the phone and ask what a charge was. A tool like that belongs on the team, and it doesn’t belong in charge. Our guide to AI and bookkeeping goes through the strengths and the failures in more detail.
What we won’t do with your data
Your ledger is full of things that are nobody’s business but yours: bank details, customer names, payroll, what you pay yourself. So the rules we hold ourselves to are short.
We won’t use a public chatbot as a shortcut for your numbers. Pasting a client’s bank export into a consumer chat tool to save ten minutes is exactly the kind of convenience that becomes a problem later. We won’t let software make final changes to your records without a person reviewing them. And we won’t get vague when you ask which tools touch your data. Specifics, like which vendors process it, belong in writing, and you should feel free to ask us for them before you commit.
What this means for you as a client
In practice, it comes down to three things. The sorting stops being the bottleneck, so more of the time goes to the parts that need judgment. There’s a name behind the work, a local Chief BooXkeeping Officer you can call, with a national team behind them. And you’re never asked to trust a black box: if a report looks odd, you can ask how a number got there and get an answer that involves a person, not a shrug. Accounting firms that resell our work under their own brand get the same split of software and people.
The same split holds if you’re catching up after months of neglect: software chews through the volume, and a person untangles the parts that need judgment. Our catch-up bookkeeping guide lays out the order of work. How your software is configured matters as much as who reviews it, and our QuickBooks Online setup guide shows what a sound setup looks like.
Questions to ask us, or anyone else
Bring these to any provider, including us.
- Which parts of my account does AI touch, and which are done only by a person?
- Who reviews the work, and how do I reach them?
- What can the software change without anyone approving it?
- Where does my data go, and can I see that in writing?
- If an error slips through, how will I find out, and how soon?
Our guide on how to choose a bookkeeper sets these beside the rest of the decision, like experience, software, and how you’ll communicate.
Frequently asked questions
Does BooXkeeping use AI on my account?
Yes, in the repetitive parts: suggesting categories, reading receipts, matching deposits, and flagging oddities. A person reviews the output and signs off on the month. If you want specifics for your own account, ask, and expect a straight answer.
Will AI replace my bookkeeper?
Not the parts that matter most to you. Software takes on volume, while judgment, cutoffs, and the conversation with you stay with a person. That’s the arrangement we’d want as owners, too.
What if I don’t want AI used on my account at all?
Say so. It’s a fair preference and worth talking through. Just know that some of this is built into QuickBooks Online and Xero themselves, such as bank feed suggestions, so “no AI at all” takes care to define. Decide together what you’re comfortable with.
Where to go from here
Take the questions above to whoever handles your bookkeeping today, and see what you hear. Specific, unhurried answers are what you’re looking for.
If you’d like to hear how human-led, tech-forward would work for your business, book a free consultation and bring those questions along. It’s a conversation with a real person about your setup, and our service is month-to-month if you decide to go further.
When a bank transaction lands in your accounting software, a machine usually gets the first look. A person gets the last one. That split is what we mean by human-in-the-loop bookkeeping, and it’s easier to trust when you can see exactly where the handoffs happen. Here they are, along with what we won’t do with your data and the questions worth asking us or anyone else.
Why we’re telling you this
Most bookkeeping companies now say they use AI, and most stop there. That isn’t enough for something as personal as your business’s numbers. You should be able to ask what a tool touches, who checks it, and what happens when it’s wrong, and get a real answer.
We also have a bias to admit. BooXkeeping is a franchise of local owner-operators, each one a Chief BooXkeeping Officer, backed by a national team, and we’ve supported 100+ franchise brands. People do the work here. But we’d be foolish to ignore tools that make those people faster and more accurate, and you’d be right to distrust us if we pretended AI does nothing. So this is the honest middle.
Where AI works in our process
Picture a hypothetical two-unit sandwich franchise. Each week its bank feed brings in card-processor deposits, a delivery-app payout, vendor payments, payroll, and one payment to the franchisor. Here’s what the software handles first.
Bank feeds and first-pass categorization come first. Transactions flow into QuickBooks Online or Xero (we’re certified partners with both), and each one gets a suggested category. The bread supplier, the utility, and the card fees look the same every week, so they sort themselves after the first few corrections. When a person fixes one, they can turn the fix into a rule so the same question doesn’t come up twice.
Receipts and bills get read for vendor, date, and amount, then attached to the transaction so the paper trail exists. The IRS’s page on recordkeeping is a fair reference for which documents are worth keeping.
Matching ties the $1,684.20 delivery-app payout to the sales report behind it. The report shows $2,210.00 in sales, so the software can surface the $525.80 that went to commissions and adjustments, the kind of gap that used to mean a calculator and a long evening. And anomaly flags catch things like the same produce vendor charging twice on one day before anyone posts it.
Each of these saves real time, and none of them is a decision. The software proposes. It doesn’t approve.
Where a person always steps in
Every batch of suggestions gets reviewed by a person before it becomes your numbers. They look at the exceptions first: new vendors, odd amounts, and anything the software flagged or wasn’t sure about. Refunds, split deposits, and payments that cover several invoices get the same treatment. Software offers a guess, and a person decides. Then come the parts software simply can’t do.
A person decides which month things belong in. Say our sandwich shop receives a $4,800 catering deposit on the 29th for an event on the 12th of next month. The software sees cash coming in. A person sees a deposit for work not yet done and asks how you and your CPA want it reported.
A person asks you questions. When a $612 charge doesn’t look like the business, they don’t guess. They ask, in plain language, what it was for. That conversation is bookkeeping too, and no tool can have it for you.
For franchise owners, a person also handles the franchisor. The bank statement shows one $2,310 payment. The franchisor’s invoice, in this made-up example, splits it into $1,905 of royalty and $405 for the brand fund, and your franchise agreement may require them to be reported separately. A person compares the bank, the invoice, and the agreement, then makes the split. Structures vary by system, so check your own agreement.
And a person closes the month. They reconcile each account to the bank statement, read the reports for sense, and sign off. Sign-off means a named human has looked at the result and stands behind it, so if a number is off, you know who to call.
“AI is my friend, not my enemy”
Our CEO, Max Emma, put it this way in a public interview: “AI is my friend, not my enemy.” We share that view, and we’d add what a good friendship looks like. You lean on a friend’s strengths and stay honest about their weak spots. AI is tireless, fast, and good at patterns. It’s also confident when it’s wrong, blind to your intentions, and unable to pick up the phone and ask what a charge was. A tool like that belongs on the team, and it doesn’t belong in charge. Our guide to AI and bookkeeping goes through the strengths and the failures in more detail.
What we won’t do with your data
Your ledger is full of things that are nobody’s business but yours: bank details, customer names, payroll, what you pay yourself. So the rules we hold ourselves to are short.
We won’t use a public chatbot as a shortcut for your numbers. Pasting a client’s bank export into a consumer chat tool to save ten minutes is exactly the kind of convenience that becomes a problem later. We won’t let software make final changes to your records without a person reviewing them. And we won’t get vague when you ask which tools touch your data. Specifics, like which vendors process it, belong in writing, and you should feel free to ask us for them before you commit.
What this means for you as a client
In practice, it comes down to three things. The sorting stops being the bottleneck, so more of the time goes to the parts that need judgment. There’s a name behind the work, a local Chief BooXkeeping Officer you can call, with a national team behind them. And you’re never asked to trust a black box: if a report looks odd, you can ask how a number got there and get an answer that involves a person, not a shrug. Accounting firms that resell our work under their own brand get the same split of software and people.
The same split holds if you’re catching up after months of neglect: software chews through the volume, and a person untangles the parts that need judgment. Our catch-up bookkeeping guide lays out the order of work. How your software is configured matters as much as who reviews it, and our QuickBooks Online setup guide shows what a sound setup looks like.
Questions to ask us, or anyone else
Bring these to any provider, including us.
- Which parts of my account does AI touch, and which are done only by a person?
- Who reviews the work, and how do I reach them?
- What can the software change without anyone approving it?
- Where does my data go, and can I see that in writing?
- If an error slips through, how will I find out, and how soon?
Our guide on how to choose a bookkeeper sets these beside the rest of the decision, like experience, software, and how you’ll communicate.
Frequently asked questions
Does BooXkeeping use AI on my account?
Yes, in the repetitive parts: suggesting categories, reading receipts, matching deposits, and flagging oddities. A person reviews the output and signs off on the month. If you want specifics for your own account, ask, and expect a straight answer.
Will AI replace my bookkeeper?
Not the parts that matter most to you. Software takes on volume, while judgment, cutoffs, and the conversation with you stay with a person. That’s the arrangement we’d want as owners, too.
What if I don’t want AI used on my account at all?
Say so. It’s a fair preference and worth talking through. Just know that some of this is built into QuickBooks Online and Xero themselves, such as bank feed suggestions, so “no AI at all” takes care to define. Decide together what you’re comfortable with.
Where to go from here
Take the questions above to whoever handles your bookkeeping today, and see what you hear. Specific, unhurried answers are what you’re looking for.
If you’d like to hear how human-led, tech-forward would work for your business, book a free consultation and bring those questions along. It’s a conversation with a real person about your setup, and our service is month-to-month if you decide to go further.