Purchasing Guides / Client Bookkeeping Transaction Coding

Client Bookkeeping Transaction Coding

A guide to buying an AI workflow that suggests bookkeeping codes for client transactions and drafts questions about unclear items.

Yearly running cost

Assuming 1,440 client months a year

By hand
$86.4K / year
2,160 hours of work
Hosted service
$185.8K+ / year
This workflow
$164 / year
Machine usage only; setup, hosting and review are extra.
Should it write codes into your books?

Review suggested codes from uploaded records, or post approved codes to your ledger.

One client month includes up to 400 bank and card transactions.

Your bookkeepers approve the codes, answer for the books and send client questions. It never reconciles, closes a period, pays or messages a client on its own.

What is included

Read a client’s bank and card exports, suggest an account, class and vendor for each transaction with the reason, send uncertain items to a review list and draft the questions for the client.

Start with each client’s chart of accounts, your firm’s coding rules and a few months of transactions your bookkeepers already coded. The connected option reads uncategorised transactions from your ledger and posts the codes a bookkeeper approves.

How reliable should the coding be?

Set targets for correct coding, fewer manual corrections and spotting items that need client input.

How these standards are measured

Test on held-out client months your bookkeepers already coded, across several clients and industries, including new vendors, transfers between the client’s own accounts, owner payments, refunds and split transactions. Measure accepted codes separately from the share coded without review. A code without a stated reason counts as a failure.

Targets for your selected standard
What is checkedTarget
Suggested codes acceptedTransactions where the suggested account, class and vendor match the bookkeeper’s labelled coding, divided by all transactions the workflow coded without sending them to review.≥95%
Coded without reviewTransactions given a suggested code without being sent to the review list, divided by all transactions in the test client months. Reported together with suggested codes accepted, so a high share cannot hide wrong codes.≥75%
Client questions foundLabelled transactions that need information from the client, such as an unexplained transfer or a payment with no matching vendor, that appear on the client question list, divided by all such labelled transactions.≥90%
Codes with a reasonEvery suggested code must state the rule, past coded entry or transaction description it relies on.≥100%

Suggested codes are prepared work for a bookkeeper to approve. Your firm still owns the books, the reconciliation and the client relationship.

How quickly should a client’s month be coded?

Choose how soon after the transactions arrive the suggested codes, review list and client questions are ready. Bookkeeper review time is separate.

Timing details

Time from receiving a client’s month of transactions to the coded list, including rules, model calls, queueing and retries. Unusually busy months may need agreed limits. Confirm client count and hardware with your provider.

The target applies to at least 95% of agreed test runs, with 4 in progress at a time.

How much do you want to spend per client month?

Set the AI processing budget for one client's monthly transactions.

Cost details

Includes model calls, retries and a shared hosting allocation. Ledger subscriptions, bank feeds, your bookkeepers’ review time and the calling agent are separate. Re-coding a month after the client answers is another run.

Reference machine cost: $0.31 – 0.61 per client month at 120 client months a month. The selected cap is a target to test, not a replacement for this estimate.

Where do you want it to run?

Run it in your cloud or on your own server. Choose whether transaction descriptions may go to an approved AI service.

Data and access details

Runs in a cloud account your firm controls, with access by client team and access logs.

Only the transaction lines and chart of accounts needed for a suggestion go to the selected external model. Agree retention and confirm this fits your engagement letters and security plan.

Keep client ledgers, bank exports and coding history inside your environment. Send only the transaction lines and chart of accounts needed for a suggestion to the approved model, under agreed retention terms. Private model only keeps client data on your hardware; the connected option still reaches your own ledger.

How do you want to use it?

Choose where you want to use it. You can select more than one.

Anything else your provider should know?

Optional. Your choices are included automatically.

Common questions

Do I need this if our ledger already suggests categories?

Not necessarily. If your ledger’s own rules or a hosted AI bookkeeping service already code well enough and your clients allow it, use them. Buy a workflow when client data must stay in your environment, you need a private model or your firm’s rules need checks a service does not offer.

Does it reconcile accounts or close the month?

No. It suggests codes and lists questions. Reconciliation, adjustments and closing stay with your bookkeepers and your existing tools.

What happens with a transaction it has never seen?

It goes to the review list or the client questions with the reason, instead of a guessed code.

Can we run it without sending client transactions to an external AI service?

Yes. Choose a private model and supply your own hardware. Measure coding quality and running cost during the pilot.

Does it send questions to clients by itself?

No. It drafts the questions. A bookkeeper edits and sends them through your usual channel.

Download RenX

Get the app.

Install, sign up, and start on your free plan with welcome credit included. No credit card required.

On a platform not listed? Leave your email and we'll notify you when a build is available.