Predictive Coding is an AI feature in Routable that automatically predicts invoice coding classifications, such as account, class, location, department, and other custom fields. These predictions help reduce manual data entry, speed up invoice processing time, and improve coding accuracy. AI takes care of bill coding. You just review and approve.
How It Works
AI Predictive Coding works alongside Routable’s Invoice Capture OCR to leverage extracted data from incoming bills, then predicting the right values for your accounting and ERP fields. By learning from your historical invoices and coding patterns, AI Predictive Coding helps your Accounts Payable team process bills in a fraction of the time.
-
Historical Data Analysis
The Predictive Coding AI uses your past invoices and posting history, ideally one year, but effective with as little as two weeks, to learn your unique chart of accounts and coding rules.
-
OCR Data Extraction
Routable’s Invoice Capture OCR reads the bill, extracting header details, line items, and other visible information and sends this information to various AI models which predict header and line item classifications.
-
AI Engine Collaboration
Multiple specialized AI models work together:
- Some excel at standard GL code prediction.
- Others specialize in project or dimension coding.
- The system ranks predictions by confidence
-
Field-Level Predictions
- Header-Level: Account, class, location, customer, and other top-level fields.
- Line-Level: Categorization of items as inventory, expenses, or other custom ledger fields.
- Vendor-Specific: Models are fine-tuned for each vendor’s coding patterns.
-
Continuous Learning
As you approve or correct coding, Routable updates the AI to reflect your latest accounting practices.
Features
- Coding predictions for all your vendor bills
- Auto-categorization of line items as inventory or expenses
- Header and line level coding predictions
- Predictions for accounting software fields like account, class, location, and customer
- Predictions for custom ledger fields and dimensions
- Models trained on your data and rigorously tested to deliver incredible accuracy
- Support for QBO, Netsuite, Sage Intacct, Xero
Key Benefits
- Less Manual Data Entry – Routable automatically predicts and fills coding fields like account, class, location, and custom dimensions.
- Higher Accuracy – Predictions are based on your historical data, vendor-specific trends, and continuously improving AI models.
- Faster Processing – Shorten your invoice lifecycle by eliminating repetitive, manual coding steps.
- Smarter Over Time – Every confirmation, correction, or change feeds back into the AI, improving future predictions.
The AI
All customers are different, so our models are trained to target the unique chart-of-accounts and accounting practices for different customers and their vendors.
The foundation of the coding prediction AI is customer historical invoices and postings. The ideal dataset contains one year of data, but Routable reaches a good performance with just 2 weeks of invoice and posting history assuming 100+ invoices per week processed.
The AI Engine consists of multiple AI models working together to reach the highest possible accounting automation performance.
- Some models work well with traditional GL codes, whereas others are great at project coding.
- Models analyze field data individually, and entire line items (groups of fields)
- Models sort data into high, medium, and low confidence categories
Routable automatically combines these models into one combined model targeted to each of your vendors. As you process bills, the coding prediction AI engine keeps improving. This feedback loop continues forever, making sure the AI is up-to-date with the latest accounting practices for your company.
Use Cases
Predictive Coding is your AP automation accelerator, cutting down manual coding work, boosting accuracy, and continuously improving with every invoice you process. Paired with Routable’s OCR, automated approvals, other automation tools, it enables an end-to-end streamlined payables workflow.
- Standard Accounting Dimensions: Auto-fill standard fields used in your accounting software.
- Custom Accounting Dimensions: Auto-fill non-standard fields unique to your business.
- Multi-Line Invoices: Automatically categorize and classify each line item accurately.
- Recurring Vendor Bills: Predict the same coding without repetitive manual entry.
FAQs
Does predictive coding require any manual training? No! If predictions are wrong or out of date, the updates you make when creating a payable feeds directly back into the training
How long does it take for the model to start predicting fields? AI Predictive Coding starts to make predictions with as little as 250 invoices with historical coding data. While the model is training, you'll see the message "We scanned your attachment, but this vendor's predictive coding model is still learning. We'll learn from each bill you code."