Power Automate is Microsoft's automation platform, and a lot of finance and operations teams want it to pull tables out of PDFs and drop them into Excel on a schedule. It can do that, but there is no single "PDF to Excel" button. You assemble the workflow from a few separate pieces, and the right pieces depend on whether your PDF has real text or is a scan. Here is what actually works in 2026, what each route costs, and when a workflow is overkill.
Can Power Automate convert PDF to Excel?
Yes, but not with one action. Power Automate reads a PDF and writes the data to an Excel file by combining a document-processing step (AI Builder, Azure AI Document Intelligence, or a third-party connector) with a step that adds rows to a table in an Excel file stored on OneDrive or SharePoint. The cloud version has no built-in plain "convert PDF to Excel" action, so every working flow stitches together two or three actions. Power Automate Desktop is closer, with an "Extract tables from PDF" action, but it only handles PDFs that already contain text.
How do you extract data from PDF to Excel in Power Automate?
The most common cloud path uses AI Builder document processing. The steps look like this:
- Build an AI Builder document-processing model and train it on about five sample PDFs that share the same layout.
- Tag the fields and tables you want, then publish the model.
- Create a flow with a trigger (a file lands in a folder, or an email arrives).
- Add the "Extract information from documents" action and point it at your trained model.
- Add an Excel "Add a row into a table" action, mapping each extracted field to a column.
For repeating tables you loop the row-add action over each detected row. The catch is the training requirement: AI Builder wants consistent layouts, so a folder of mixed document designs means more than one model.
Power Automate Desktop or the cloud version for PDF to Excel?
Power Automate Desktop runs on your machine and has an "Extract tables from PDF" action that pulls tabular content straight into a variable you can write to Excel. It is the quickest route for a text-based PDF and needs no model training. The cloud version is the choice when you want the flow to run unattended on a server, triggered by an incoming email or a SharePoint upload, with no one logged in. Desktop is better for ad hoc and local work; the cloud is better for scheduled, hands-off pipelines.
How to extract a table from a PDF in Power Automate
If you would rather not train a model, two no-model approaches exist. Azure AI Document Intelligence has a prebuilt Layout model that returns every table in a PDF as structured JSON; you call it from a flow and parse the JSON into Excel rows. Alternatively, third-party connectors such as Encodian, Plumsail, Adobe PDF Services, or PDF4me expose a direct "convert PDF to Excel" action that recognizes tables for you. These connectors are premium and usually metered per document, but they remove the model-training step entirely.
Does it work on scanned PDFs?
Only if optical character recognition is in the loop. A scanned PDF is an image, so the plain text-extraction actions return nothing. AI Builder and Azure Document Intelligence both run OCR automatically, and the premium connectors offer a high-quality mode for image-based files. Power Automate Desktop's basic table action does not OCR, so it fails on scans. If most of your documents are scans, lean on the AI routes or a converter built around OCR rather than the desktop table action.
Is Power Automate free for converting PDF to Excel?
Not for this. AI Builder document processing and the premium PDF connectors are paid features that sit outside a standard Microsoft 365 license, billed through AI Builder credits or per-action connector pricing. Azure Document Intelligence is its own metered Azure service. The build also costs time: training models, mapping fields, and testing the flow against real documents is a project, not a five-minute task. That investment pays off when you process the same layout hundreds of times a month. It rarely pays off for a one-off file or a handful of mixed documents.
What goes wrong when you build the flow?
A few problems trip up most first attempts. Layout drift is the big one: an AI Builder model trained on last quarter's statement layout starts returning blank or shifted fields when the vendor tweaks the design, so you retrain or split into separate models. Tables that span several pages can fragment, with the connector treating each page break as a new table. Pages holding two or three tables side by side often merge or scramble columns. And the premium licensing surprises people: the trigger and Excel steps may sit inside your plan, but the extraction action that does the real work usually does not, so test the cost on a small run before you schedule it across thousands of documents.
When is a direct PDF to Excel converter the better choice?
If you are not running a high-volume, repeating pipeline, a workflow is more plumbing than the job needs. For one-off files, mixed layouts, or a quick batch, it is faster to drop the file into a PDF to Excel converter and download the spreadsheet, with no model to train. Our tool reads table structure and runs OCR on scanned PDFs automatically, and you can convert many PDFs in a batch in one pass. If you genuinely need scheduled, unattended automation across one steady layout, that is where Power Automate earns its setup, and our guide to automating PDF to Excel conversion covers the broader options. Either way, the goal is the same: get clean rows out of a PDF and into a spreadsheet you can work with.
Power Automate is not the only document workflow worth wiring up. If the source data arrives by email instead of a folder, a tool that turns incoming emails and their attachments into structured data can feed the same spreadsheets. And teams automating supplier invoices end to end often pair extraction with accounts payable automation so approved bills flow straight to payment. Pick the layer that matches how repetitive the work really is.