You have large number of documents, user will upload all at once you have to differentiate them into different categories. Which Appian object will you recommend for this problem?
You have large number of documents, user will upload all at once you have to differentiate them into different categories. Which Appian object will you recommend for this problem?
Answer
I recommend the Document Classification AI Skill for this. It uses machine learning to identify and categorize uploaded documents such as Invoices, Purchase Orders, and Contracts.
That cuts down the manual sorting, speeds up document processing, and lets you route each document automatically based on its type.
Implementation steps
1. Create an AI skill
- Navigate to New → AI Skill → Document Classification.
- Define the document categories (Invoice, Purchase Order, Contract).
- Upload at least 10 unique PDF documents per category for training. Appian recommends starting with at least 50 documents per category for better results.
2. Train and publish the model
- Train the AI model using the uploaded sample documents.
- Review its performance and test it with additional documents.
- Publish the model once the classification results meet your requirements.
3. Implement bulk document classification
- Create an Appian interface that allows users to upload multiple PDF documents.
- Pass the uploaded documents as a list to the Classify Documents smart service in a process model.
- Configure a confidence threshold (for example, 80%) to determine which predictions require manual verification.
4. Handle classification results
The smart service returns three outputs:
Output | Description | Action |
|---|---|---|
Above Threshold | Documents classified with confidence above the configured threshold | Process automatically |
Below Threshold | Documents classified with lower confidence | Route for manual review |
Failed | Documents that could not be classified | Log errors and investigate |
Each successful classification result contains the document ID, predicted classification, and confidence score.
5. Store and route documents
- Store the document ID, predicted category, confidence score, and processing status in an Appian record type backed by a suitable data source.
- Automatically route successfully classified documents to the appropriate workflow.
- Assign low-confidence documents to users for manual verification.
- Track failed documents for troubleshooting or reprocessing.
Example: Vendor document processing
Scenario: A company receives hundreds of documents from vendors every day, including invoices, purchase orders, and contracts. Right now, employees open each document by hand, work out what type it is, and forward it to the right department.
We can automate this with the Document Classification AI Skill.
Example: A vendor uploads five PDF documents through an Appian portal, and we set the confidence threshold to 80%.
Uploaded document | AI prediction | Confidence | Action |
|---|---|---|---|
invoice_101.pdf | Invoice | 97% | Send to Finance |
po_202.pdf | Purchase Order | 94% | Send to Procurement |
contract_303.pdf | Contract | 91% | Send to Legal |
document_404.pdf | Invoice | 62% | Manual verification |
damaged_505.pdf | Failed | n/a | Error handling |
The predictions and confidence scores above are illustrative.
How I would implement this in Appian
- Create a Document Classification AI Skill with three categories: Invoice, Purchase Order, and Contract.
- Train the model on sample PDFs and publish it.
- Build an interface where vendors can upload multiple documents.
- Start a process model and pass the uploaded document IDs to the Classify Documents smart service, with a confidence threshold of 80.
- Process the results:
- Above Threshold: Route each document to its department.
- Below Threshold: Create a user task to confirm or correct the classification.
- Failed: Record the error and request a valid document if necessary.
- Store the document ID, classification, confidence, review status, and assigned department in a record type for tracking.
Important considerations
- Execution limits: A single smart service execution handles a maximum of 100 PDF documents or 100 MB combined, whichever limit is reached first.
- Batch processing: Split larger uploads into smaller batches and process them separately.
- Training limits: A skill supports a maximum of 20 document categories and 100 training PDFs per category.
- Confidence threshold: If no threshold is configured, all successful predictions are returned in Above Threshold, regardless of confidence.
- Asynchronous processing: The smart service executes asynchronously and does not support activity chaining.
Official Appian Documentation:
Check yourself
Question 1 of 3, on what the answer above covers.
- 1.Your Appian application uses a Document Classification AI Skill to classify vendor documents. The confidence threshold is set to 80%, but an uploaded invoice is classified with 65% confidence.What should your process model do?
- 2.A user uploads 150 PDF documents totaling 90 MB. You need to classify all of them using the Classify Documents smart service.What is the best approach?
- 3.Your Appian AI Skill correctly classifies an uploaded document as an Invoice with 96% confidence.The business now wants to automatically retrieve the invoice number, vendor name, and total amount.What should you implement next?