Case study
2026WebAI accounting practice buddy
Accounting practices lose hours every week to work that isn't really accounting. We built an AI buddy that sits inside the practice's daily workflow, reads every incoming document, enters and reconciles it, checks its own work and chases clients for what's missing.
- Client
- Australian accounting & property practices
- Year
- 2026
- Services
- AI engineering · Workflow automation
- Platform
- Web app, email intake, OCR
The problem
Chasing invoices, typing bank lines into the ledger, matching receipts. The most valuable hours in a practice were going to work that isn't really accounting.
One document, start to finish
From inbox to reconciled books
One document, start to finish
From inbox to reconciled booksClients just email it.
Invoices go to a monitored inbox, scanned statements are uploaded, and historical transactions come in as a bulk CSV. No retyping.
- Invoice INV-2041 · PDFcaptured
- Bank statement · March · scanOCR
- Transactions Q1 · 312 rowsCSV
OCR turns paper into data.
Scanned statements and invoices become structured, categorised line items.
- SupplierCoastline Plumbing
- ABN51 824 753 556
- Date12 Mar 2026
- Total inc. GST$4,180.00
Entries post and reconcile.
Transactions are entered and matched against the bank automatically, category by category.
- Coastline Plumbing · $4,180.00Matched
- Office rent · $6,500.00Matched
- Strata levy · $920.50Matched
The AI reviews its own work.
Every automated entry is reviewed. Anything unusual is flagged for a human before it reaches the ledger.
- Harbour Supplies · $1,284.00Differs by $42
- Fuel · $86.40Receipt missing
- Client payment · $12,400Matched
Then it emails the client.
Missing documents are requested in clear, professional language, and an outcome summary goes out after every run.
“We’re missing one receipt for March: $86.40 fuel on 7 March. Could you send a photo?”
●Under the hood
Capture, review, follow up
Email intake
Clients email invoices to a monitored inbox and entries are created automatically.
solves → Manual data capture from every invoice.
↺ Client replies flow straight back into the books
What we did
Designed around how practices really work
We started from the practice's daily workflow, not from the technology, so staff barely had to change how they work.
Workflow research
Sat with practice staff to map where the hours were going.
Product design
Review screens built for accountants, not engineers.
AI & OCR
Document reading, categorisation and the AI review step.
Cloud & security
Role-based access and a full activity history on AWS.
Discover
Mapped intake, entry, reconciliation and client communication across the practice.
- workflow mapping
- pain points
Design
Designed the intake flow, the review queue and client-facing messages.
- review UX
- message tone
Build
Built email intake, the OCR pipeline, CSV import and the reconciliation engine.
- OCR
- email-to-entry
- CSV import
Secure
Added account-level access control and logging aligned with practice structures.
- role-based access
- audit trail
Launch and learn
Rolled out to daily work and tuned the AI review on real documents.
- rollout
- AI review tuning
Built with
The stack behind it
Accountants stay in control
The AI flags. A human decides on every exception.
Secure by design
Account-level access control and a full activity history.
Fits the daily routine
Clients keep emailing. The system does the rest.
Scroll to switch the system on
Hours lost to retyping.
- ✕Data entry from invoices and bank statements ate up hours every week.
- ✕Follow-ups for missing documents relied on manual emails and memory.
- ✕No fast way to bring large historical batches into the system.
- ✕No consistent review step to catch errors before month-end.
Books that mostly run themselves.
- ✓Invoice and statement entry runs automatically every day.
- ✓Clients get professional follow-ups without anyone chasing them.
- ✓Bulk CSV and OCR uploads replace line-by-line entry.
- ✓The AI review step catches exceptions before they reach the ledger.
The product
Every run, summarised and followed up
Impact
Time back for real accounting work
Staff spend their hours on advisory and judgment work instead of data entry.
Clients get clear, timely requests instead of last-minute chasing.
Errors are caught before month-end, not during it.
Could this work for you?
Built for any team buried in documents
Accounting firms
Bookkeeping, BAS and month-end reconciliation.
Property managers
Invoices, levies and owner statements.
Bookkeeping services
High-volume client document intake.
In-house finance teams
Supplier invoices and expense receipts.
Next case studyAI agent trust layerA layer between AI agents and the CRM that checks every action before it reaches a customer: is the data true, is the action allowed, is it within budget, and can it be undone.Read next