Extract invoice data and automatically file processed invoices
Invoices from different providers arrive in a OneDrive or SharePoint folder. A Power Automate flow detects each new or modified file, uses AI Builder to extract fields such as supplier, invoice number, date and amount, adds the required data to an Excel table, then renames the source file with a reference and moves it to a processed folder.
Where the action happens
Power Automate / OneDrive or SharePoint / AI Builder / Excel
Product or feature needed
AI Builder invoice processing inside an automated cloud flow
AI required
Yes
Extra credits
Yes - AI Builder/Copilot Credit capacity can be consumed; monitor environment capacity and configure pay-as-you-go or sufficient prepaid capacity where required
Technical skill
Medium
Risk level
Medium - extraction errors, duplicate triggers or exhausted AI capacity can affect financial records and processing
Licence dependency
Power Automate rights are required. AI Builder/Copilot Credit and any premium connector requirements depend on the exact configuration and current tenant licensing.
Credit dependency
The invoice extraction AI action consumes AI capacity. Organisations should monitor AI Builder/Copilot Credit capacity because exhausted capacity can stop the AI step; the non-AI file and Excel actions have separate Power Platform request/service limits.
UK availability
available
What it can do
- Uses the prebuilt AI Builder invoice model to extract common invoice fields such as vendor, invoice ID, invoice date, totals and other structured values.
- Can write extracted values into an Excel table or another approved system of record as part of the flow.
- Can construct a standard filename from extracted values or an internal reference, then rename and move the invoice to a processed folder.
- Creates a repeatable end-to-end process from document arrival through extraction, recording and filing.
Limitations
- AI extraction is probabilistic and confidence varies by supplier layout and scan quality; important financial values should be validated before payment or posting.
- A 'file modified' trigger can run more than once if the flow itself subsequently changes the file, so the design needs a processed-state or folder pattern that avoids loops and duplicates.
- AI Builder/Copilot Credit exhaustion can cause the AI step to fail even when the Power Automate flow and user are otherwise licensed.
- Excel is practical for modest workloads but can become fragile with concurrent writes or larger transaction volumes; Dataverse, SharePoint Lists or a finance system may be more suitable at scale.
Prerequisites
- A Power Automate environment with access to the relevant OneDrive or SharePoint folder and Excel table.
- Access to AI Builder invoice processing and sufficient AI Builder/Copilot Credit capacity under the organisation's current licensing arrangements.
- A defined set of invoice fields, filename convention and processed-folder structure.
- Exception handling for low-confidence extraction, missing fields, duplicate invoices and failed moves.
Workarounds
Add a confidence threshold and route invoices with low-confidence or missing critical fields to a human-review queue before the spreadsheet is updated or the invoice is marked processed.
No premium licensing needed. External dependency: Power Automate logic / approval or review process.
If the volume grows beyond what Excel handles reliably, write extracted invoice data to a SharePoint List, Dataverse or the finance system instead, subject to connector and licensing requirements.
Needs premium licensing. External dependency: SharePoint, Dataverse or finance-system connector.
Notes
A sensible production pattern is: trigger on arrival → extract invoice values → validate critical fields/confidence → add or update the record → generate the standard filename/reference → move to a processed folder → log failures separately. Keep payment approval and other consequential finance decisions with people or existing finance controls.
Watch-outs for Add AI to flows
- AI actions can fail or stop if the environment has insufficient credits or capacity.
- Premium connectors may add a separate licensing dependency.
- Use AI to create drafts, flags, summaries or extracted data rather than making regulated AO decisions autonomously.
- Forecast consumption and set monitoring before a pilot becomes business-as-usual.
Sources
- https://learn.microsoft.com/en-us/ai-builder/flow-invoice-processing
- Use the invoice processing prebuilt model in Power Automate checked 2026-09-16
- Invoice processing prebuilt AI model checked 2026-09-16
- OneDrive for Business connector - Move or rename a file using path checked 2026-09-16
Last checked: 2026-09-16
Community feedback
Has your organisation tried this? Confirm whether it's possible, whether it's useful, and share any tips.
0
Confirmed possible
0
Reported not possible
0
Found it useful
0
Not useful
Loading feedback…