The Document Understanding feature was developed to improve the accuracy of AI Agents in one of their most common uses: automatically extracting data from attached documents (such as invoices, contracts, receipts, etc.).
This enhancement combines AI Agents’ intelligence with OCR (optical character recognition) technology, making the reading and extraction process much more accurate, especially for complex files or those with diverse formatting.
🚀 How It Works
At the moment, the feature is not enabled by default. To use it, you must request activation from your account manager or Pipefy support.
Once activated:
All AI Agents in your organization that perform document data extraction will automatically use OCR + AI technology;
No configuration is required in the agent — everything happens in the background;
Visually, nothing changes in your current interface.
⚠️ Important: In the future, you’ll be able to enable or disable this feature per agent. For now, it is a globally enabled resource for the entire organization.
💳 How Credit Consumption Works with Document Understanding
Whenever an AI Agent performs document data extraction with the Document Understanding feature enabled, the consumption will be:
2 standard AI Agent credits
+1 additional credit for OCR (Document Understanding)
Total: 3 credits per run involving document reading.
For more details on standard AI Agent credit usage, see "How AI Agents pricing works".
Applied limits: follow the same general AI Agent limits:
Maximum of 30 documents or images per run (e.g., 30 pages in a single file, 2 files with 15 pages each, 3 files with 10 pages each, etc.).
More details on the "AI Agent: Create an Agent" article.
🧠 Why Use It?
Significantly higher accuracy in extracted data, even from documents with varied formatting;
Ideal for critical use cases such as reimbursements, accounts payable, contract validations, and more;
Reduces rework and prevents errors in automated processes.
❓ How to Enable It
Contact your account manager or reach out to our support team via chat.
Once enabled, you can immediately test and monitor results directly in your existing processes.
📌 See also: