Affinda's solutions have been designed as a 'human-in-the-loop' solution to ensure that data extracted from our platform has 100% accuracy. We recommend that users validate the information extracted by our AI models via our Validation Interface, which creates a feedback loop that allows the models to learn and improve over time (where a tailored model is set up).
Using the validation tool
The standard workflow we recommend customers implement is a simple four-stage process that ensures accurate and continually improving data.
- Upload - invoice is uploaded via API or our web app for parsing
- Extract - Affinda's AI models will predict the various data fields within the invoice
- Validate - a 'human in the loop' validates the data extracted by the model and makes any corrections necessary. Rules are created to reduce the amount of human intervention required, by auto-validating fields if they are above a defined confidence threshold or other rules. The human-validated data is fed back into the model for re-training to improve accuracy over time
- Export - once a human has validated all of the required fields, data is exported typically via JSON file into the customer platform for further processing
Updated 14 days ago