Intelligent document processing decides how every AI agent in your account reads the documents you give it. When the setting is on, Celigo extracts the text and tables from each document — using OCR for scanned pages — before the AI model reads it, so PDFs, scans, and Word, Excel, and PowerPoint files work the same way on every model. When the setting is off, each file goes to the AI model as it is, and the model's own file support decides what it can read.
The setting is off by default. Account owners and administrators can turn it on for all agents in the account. Once turned on, it applies automatically—no updates to agents, flows, or mappings needed.
Prerequisites:
- You must be the account owner or an administrator.
- Sign in to your production environment. You can't change the setting from a sandbox or other non-production environment.
- Identify the agents that already receive PDFs. With the setting on, Celigo reads every PDF before the model does, including digital PDFs that the model could read by itself. Preview those agents with a sample file after you turn the setting on to confirm the output is what you expect.
Turn on or off intelligent document processing
- Sign in to integrator.io, go to Account, and then select Company.
- In the AI settings section, select or clear the Intelligent document processing option.
- Select Save.
What changes when the setting is on
| Setting off (default) | Setting on | |
|---|---|---|
| Who reads the document | The AI model's provider receives the file as it is. | Celigo extracts the content, and the model's provider receives the extracted content instead of the file. Images go to the provider in both states. |
| File types | PDF, images, and plain text, depending on the model | PDF, Word, Excel, PowerPoint, RTF, TIFF, BMP, images, and plain text, on every model |
| Scanned documents | Depends on the model's own vision | OCR reads scanned PDFs, TIFF, and BMP files |
| AI credits | The model's usage; some providers add an image of every PDF page, which uses more input | The model's usage; reading the document doesn't consume AI credits, and each file counts as one AI agent action |
Some model providers read other formats, such as Word or Excel files, when the setting is off. That's the provider's own behavior, which Celigo doesn't support and which can change without notice. To read Microsoft Office documents on any model, turn the setting on.
For the complete list of file types, how each is read, every limit, and the errors you see for files that can't be read, see Supported file types and limits for intelligent document processing.
FAQ
Can I turn on document processing for one agent only?
No. The setting applies to every AI agent in the account. To keep an agent from processing a file type, change what the flow sends to that agent.
Do I need to change my agents after I turn it on?
No. Instructions, output format, mappings, and error handling work as before. Preview agents that receive PDFs, because Celigo now extracts the content of those PDFs before the model reads them.
Does it work with bring-your-own-key (BYOK) models?
Yes. Document processing works the same way for Celigo AI models and BYOK connections.