Self-Training, Mappings, and Workflows: How They Work Together

Self-Training and Its Relationship with Mappings and Workflows

Self-training, mappings, and workflows complement each other:

• Self-training improves the quality of data extraction (what the AI reads from the document).
• Mappings apply business rules to that data (for example, cost centers or accounting accounts).
• Workflows manage approvals based on that data (for example, amount, supplier, etc.).

A good practice is:

• Use self-training to ensure Dost correctly reads key data.
• Use mappings to transform or complete that data according to your business rules.
• Base workflows on reliable data (amount, supplier, etc.).

Best Practices

Prioritize high-volume suppliers

Start by training the AI on suppliers that generate the most invoices, as this will have the greatest impact.

Use unique and consistent tags

Recommendation: use the supplier’s VAT ID as the training tag to maintain a clear and consistent identifier.

Avoid training on exceptional documents

Do not use unusual invoices, poorly formatted documents, or one-off supplier errors for training purposes.

Review results periodically

From time to time, check whether the AI is still correctly interpreting trained documents.

If you detect new recurring errors, add more training examples or verify whether the supplier format has changed.