Hi All,
I’m looking for guidance or best practices for implementing a document verification mechanism in OutSystems 11, preferably using an AI-based approach. Requirement: I need to validate customer-submitted form data against the information available in an uploaded document (e.g., ID proof, invoice, or other supporting documents). The system should: Extract data from the uploaded document Compare it with the data entered by the customer in the form Allow form submission only if the data matches successfully.The primary goal is to reduce misuse and detect potential fraud during form submission. My questions: 1)Are there any AI tools or services that can be integrated with OutSystems 11 to achieve this (e.g., OCR + data validation using AI)? 2)Has anyone implemented a similar use case using OutSystems integrations (such as Azure AI, Google Vision, or other OCR/ML services)? 3)Is it feasible to implement this functionality without AI (rule-based or standard validation), and what would be the limitations compared to an AI-based approach? 4)Any recommendations on architecture, components, or Forge assets that can help with this scenario?
I would really appreciate any suggestions, references, or examples from your experience. Thanks in advance!
Based on my personal experience, I implemented the solution as follows:
First, data is extracted from the uploaded document using AI. In cases where the AI extraction fails or produces low-quality results, a C# fallback mechanism is used to manually extract the required information from the document.
After that, the extracted data is compared internally and also validated against the employee’s stored data in the database, using AI to improve the accuracy of the verification process.
Then, a comprehensive evaluation system is applied to the output based on several criteria, including:
Finally, a complete output report is generated for each document, which includes:
Thank you for your response.
May I know which AI model have you integrated in Outsystems 11 and which technical implementation steps have you followed to achieve the same ?
Regards,
Ajit Kurane
I tried several models, and Azure delivered the best results while following the exact same steps and sequence across all tests. Of course, performance can vary from one business to another depending on the specific use case.
Could you please provide guidance on which technical implementation steps you followed and best practices for secure implementation ....?