Bookkeeping automation
Bookkeeping automation for accountants and bookkeepers: code it once, review it after.
Bookkeeping is the same decisions made every month for every client: this supplier goes to this account, this merchant is fuel, this transfer is rent. Pileform learns those decisions per client, from your rules and from each correction you make, applies them to the next pile, and leaves you a review queue instead of a typing job.
Bookkeeping automation for accountants and bookkeepers means coding each client's repeat suppliers once and reviewing afterwards. Pileform applies your categorisation rules first, then that client's counterparty memory, and only then a suggestion against that client's chart of accounts. Each correction feeds that client's rules for the next period, and nothing reaches the ledger until you confirm it.
The repeat work is the work.
A bookkeeper’s month is mostly repetition. The same forty suppliers per client, coded the same way as last month; the same card merchants on the statement; the same rent transfer. Typing those in again is where the hours go, and where one client’s habit gets applied to another client by mistake. The practice setup around it, with roles, per-client ledgers and a review queue per company, is described under bookkeeping software for accountants; this page is about the coding that repeats.
Pileform keeps what it learns inside each client company. Categorisation rules by description, SKU or supplier run first and adjust from your corrections. Counterparty memory holds each supplier’s aliases, VAT number and default debit and credit codes, and its confidence rises as you confirm. Twenty clients under one login are twenty separate rule sets, never one blended guess. What the rules cannot settle is suggested against that client’s chart of accounts and flagged when unsure. The pile itself is read by invoice automation and bank statement automation, the rate per line is decided by Cyprus VAT software, and confirmed entries post into Xero, QuickBooks or the client’s own ledger.
How the monthly cycle runs.
The client’s pile, as sent
A PDF, a ZIP of photos, a folder of scans, the bank statement. One drop per client per period, with no sorting first.
Coded from that client’s history
Rules first, then counterparty memory, then a suggestion against that client’s chart of accounts. Each line keeps its own VAT rate and its description as printed.
Reviewed where it matters
Confident rows and flagged rows sit in one queue, to confirm in bulk or one at a time. Each correction you make feeds that client’s rules for the next period.
Posted and kept
Confirmed entries post to the client’s ledger without duplicates. The per-supplier Excel workbook, with the source embedded, stays in the practice’s archive for the retention period.
What is remembered, client by client.
Learning is scoped to one company at a time. Nothing one client teaches it is applied to another.
- Categorisation rules
- Description-pattern, SKU-pattern and supplier-default rules, evaluated before any AI step and adjusted from your corrections
- Counterparty memory
- Each supplier’s and customer’s aliases, VAT number and default debit and credit codes, with a confidence score that rises as you confirm
- Chart of accounts
- Per company, seeded from a Cyprus default template or synced from Xero or QuickBooks, with deleted accounts kept for the audit history
- Supplier registry
- One canonical record per supplier, VAT-number scanning, and two records for the same party merged into one
- Posting history
- What suggested each entry, its confidence and the action taken, kept per company for the reviewer and the auditor
Who it is built for.
Bookkeepers with a monthly client list
Many small companies, each coded its own way, closed on the same few days each month.
Accounting practices
Staff upload and review, the owner posts, and each client’s learning stays with that client.
Practices taking on a client mid-year
Opening balances brought in, the chart of accounts synced from the client’s ledger, and the rules built from the first corrections.
Bookkeeping automation, answered.
From two sources: the rules you write and the corrections you make. Each confirmed line teaches the counterparty memory that supplier’s default codes, and the categorisation rules adjust from your edits, all within that client company.
No. Chart of accounts, counterparties, categorisation rules and posting configuration are all per company. Twenty clients under one login are twenty independent rule sets.
Yes. Description, SKU and supplier-default rules run before any AI step. An AI suggestion is used only above a confidence floor, and anything below it waits for you.
No. Every proposed entry waits in the review queue. Confidence scores let you work through it faster, confirming in bulk, but only a person with the owner role posts to the ledger.
No. The destination is chosen per company: Xero, QuickBooks Online or Desktop, Business Central, BTMS or Esoft, or the CSV for anything else.
Run one client’s month.
A free account comes with 30 free processing pages that do not expire and needs no card. Correct a few lines, then send the next month’s pile and read what it remembered.