Adding a 2026 perspective: vision-capable LLMs (Claude Haiku, GPT-4V) have made automated structured extraction from receipt PDFs trivially cheap. For small businesses building personal receipt-matching tools, the bottleneck has shifted entirely to the Xero side - programmatic access to unreconciled statement lines for one's own org.
The current workaround is to manually export the Bank Reconciliation Detail report from the UI then parse the xlsx. Functional, but adds a manual click per tranche and prevents unattended/scheduled runs.
If full Finance API access couldn't be opened due to bank-feed contracts, a scope-limited alternative would still unlock significant automation: Standard OAuth only, single-tenant, the connecting org's own data only, read-only. The data is already mine via the UI - this would just make it accessible programmatically for the org I own.
Use case: receipt-matching pipeline for small org not outsourcing bookkeeping.
Michael Garrett
supported this idea
·
Adding a 2026 perspective: vision-capable LLMs (Claude Haiku, GPT-4V) have made automated structured extraction from receipt PDFs trivially cheap. For small businesses building personal receipt-matching tools, the bottleneck has shifted entirely to the Xero side - programmatic access to unreconciled statement lines for one's own org.
The current workaround is to manually export the Bank Reconciliation Detail report from the UI then parse the xlsx. Functional, but adds a manual click per tranche and prevents unattended/scheduled runs.
If full Finance API access couldn't be opened due to bank-feed contracts, a scope-limited alternative would still unlock significant automation: Standard OAuth only, single-tenant, the connecting org's own data only, read-only. The data is already mine via the UI - this would just make it accessible programmatically for the org I own.
Use case: receipt-matching pipeline for small org not outsourcing bookkeeping.