The decision is usually made on headcount cost, then quietly re-made every month by whoever holds the records. What determines how well either arrangement works is where the ledger sits and how many copies of it exist.
The decision is rarely all or nothing
Most growing businesses end up with a split rather than a clean choice. Day-to-day entry sits with an office manager or a part-time bookkeeper, the year-end accounts and tax filings go to an external firm, and something in between, usually payroll or the monthly management pack, moves back and forth depending on who has capacity that quarter.
That split is sensible. A statutory filing is done a few times a year by someone who does it hundreds of times a year, and paying a firm for that is cheaper and safer than learning it internally. Recording a purchase invoice is done daily by someone who knows which job it belongs to, and sending it out of the building adds a delay for no benefit.
The question worth answering is not which side wins. It is which specific tasks sit on which side, and what has to travel between them each month for both sides to be working from the same numbers.
What the arrangement is really buying
An external firm sells three different things that often arrive in one quote. There is compliance work, meaning statutory accounts, tax computations and filings, priced per event and largely fixed in scope. There is processing work, meaning entering transactions, running reconciliations and preparing a monthly pack, priced by volume or by hour. And there is advice, meaning someone reading the numbers and telling the owner what they mean.
An in-house hire buys availability instead. Somebody who can answer a question the same afternoon, chase a customer whose payment is a fortnight late, and who accumulates knowledge of how the business works that nobody has to re-explain each January.
Businesses usually notice a problem in the middle band. Processing work is where volume grows fastest, where the fee scales with transaction count, and where the difference between the two arrangements is mostly about how information moves rather than who is skilled at what.
The cost that does not appear in either quote
Setting a monthly fee beside a salary leaves out the coordination work, and that is often the part that grows. Someone inside the business has to gather documents, answer queries about coding, check the pack when it comes back, and explain the entries that do not look like last month. That work sits with an owner or an operations manager who was never hired to do it, so it does not show up in the finance budget at all.
The size of it depends on how the records are held. When the firm works in its own copy of the books and the business works in another, the month contains a reconciliation between two versions of the same period, plus a lag while corrections are applied on one side and sent back to the other. When both sides post into the same ledger, the coordination is a query about a specific transaction rather than a reconciliation of two files.
The real question is not who does the bookkeeping. It is how many copies of the books exist while they are doing it.
One set of records, whoever is doing the posting
AOC Accounting holds payables, receivables and the general ledger in a single ledger, with the chart of accounts controlling posting so the same transaction cannot be recorded twice. An external bookkeeper posting into that ledger and an internal staff member raising invoices in it are working on the same rows, so there is no monthly step where two sets of records are brought back into agreement.
Reports come out of the ledger immediately rather than through an export, which changes what the business can see between visits from the firm. An owner who wants receivables ageing on the 12th does not have to ask for it and wait, and a correction posted late in the week appears in the next print of the report rather than in a rebuilt spreadsheet model.
User-defined journal types matter more in a split arrangement than in a single-owner one. Where the sequence, the printed document and the auto-reversal rule are part of the transaction type, a recurring accrual behaves the same way whether it is posted by the firm in month one or by a new internal hire in month seven. Up to 10 analysis dimensions per function mean the splits the business manages by, such as project, office or client group, are captured at entry, so a pack the firm prepares and a figure the owner pulls are drawn from the same tagged transactions.
What tends to pull work back in-house
Three situations move processing work inside, and none of them is about fees.
The first is timing. A business that needs to know its position weekly rather than monthly finds that an external cycle built around a monthly pack cannot deliver that, and the gap gets filled by an internal spreadsheet that gradually becomes the number people trust.
The second is operational coding. Once transactions have to be tagged by project, site or contract, coding them correctly requires knowing which job a delivery was for. That knowledge sits with the person who ordered it, and passing it to an outside processor by email costs more than entering it directly.
The third is volume with variation. High volume alone is fine to send out when transactions look alike. Volume with exceptions, such as part payments, credit notes and disputed invoices, generates queries, and each query is a small handover in both directions.
None of these argue against using a firm for compliance and advice. They tend to move the daily entry inside while the specialist work stays out.
What tends to pull processing work inside
- The business needs its position weekly, and a monthly external cycle cannot deliver it.
- Transactions have to be coded by project, site or contract, and that knowledge sits with the person who ordered the work.
- Volume comes with exceptions such as part payments, credit notes and disputed invoices, and each one is a handover in both directions.
What the system does not decide
Shared access to one ledger does not decide who should hold which task. That depends on the volume the business does, whether it can hire and keep someone, and what its adviser is genuinely better at, and it is a management judgment rather than a software setting.
It also does not remove the need for controls between the two sides. Deciding who can create a supplier record, who releases a payment run and what an external party may post without approval is policy work. A system can enforce those rules once they exist and record who did what, but it will not write them.
A chart of accounts that was never designed for how the business now runs produces the same unclear reports whoever maintains it, and that cleanup is a working session with people who understand operations rather than a migration task. And for a business whose accounting is genuinely simple, a firm working in an entry tool remains a reasonable answer, and the case for a change should rest on complexity the current setup cannot hold.
The question to ask before either arrangement starts
Ask a prospective firm, or the person being hired, one thing: where will the ledger live, who else can post into it at the same time, and what has to be sent back and forth each month for both sides to agree.
An answer describing a shared ledger with named permissions means the monthly coordination is queries about individual transactions. An answer describing a file that is sent over, worked on and returned means the business is paying for a reconciliation every month, on top of the processing, and the cost of that grows with transaction volume rather than staying flat. That answer also tells the business what it will have to unpick later if it changes the arrangement, because records held in someone else’s copy of the books are harder to take back than records held in a ledger the business owns.
One ledger, whoever is posting into it
A short walkthrough of shared-ledger access, user-defined journal types and analysis dimensions, using your own split between internal staff and your firm.




