Summary

Last week, Ohpen participated in AWS Summit Amsterdam 2026, where our own Wessel Stoop took the stage together with Anja Pettersson from Amazon Web Services (AWS). 

Their breakout session, “Agentic AI for Faster Mortgages: Ohpen and AWS on Shipping Agents to Production,” focused on a practical example of how Ohpen is applying AI agents within its mortgage technology platform. 

The session reflected a broader shift in artificial intelligence: the focus is moving beyond experimentation towards practical applications that can be integrated into everyday operations. Rather than exploring AI’s potential in theory, organisations are increasingly looking at how AI can be deployed securely, scaled effectively, and applied to solve real business challenges.

Applying agentic AI to financial document processing

Financial document processing plays an important role in mortgage journeys. Processing financial information efficiently and accurately is essential for creating smooth digital experiences and supporting faster mortgage processes. 

During the session, Wessel demonstrated an example of how Ohpen is using an AI agent to improve its financial document processing pipeline. 

The agent was built to autonomously: 

  • detect pain points within the document processing pipeline;  
  • propose improved extraction prompts; and  
  • verify whether those proposed improvements deliver better results.  

This creates a continuous feedback loop in which the agent identifies opportunities for improvement, suggests changes and validates whether those changes have a positive impact before recommendations are evaluated by human colleagues. 

Moving from AI experimentation to production

A key theme of the session was the practical application of Agentic AI. Rather than assisting with individual tasks, AI agents can work towards a defined objective by selecting appropriate tools, analysing system feedback and determining their next steps. 

The example presented at AWS Summit Amsterdam demonstrated how Agentic AI can be used as part of a continuous improvement approach, helping teams identify issues and optimise processes more efficiently. 

One of the key lessons shared during the session was the importance of context. Effective AI agents require more than instructions alone. Providing the right tools, meaningful context and informative error messages enables agents to make better decisions and contribute more effectively to existing technology workflows. 

When combined with appropriate human oversight, this enables AI agents to move beyond technology demonstrations and become part of practical production environments. 

AI solutions deployed today

The session also highlighted examples of AI capabilities that are already part of Ohpen’s platform:

iWoud – Agentic Data Extraction Configurator 

iWoud monitors and autonomously improves the performance of data extraction from financial documents. It inspects live production data, identifies fields that are frequently corrected, and proposes improved extraction configurations, reducing the need for manual tuning. 

DocStreet / Collect & Conclude – Data Extraction via LLM 

LLMs running on AWS Bedrock extract structured data from financial documents, including salary slips and employer statements, using zero-shot prompting without requiring training data. This replaces the legacy manual rule-based extraction process and improves data extraction performance on column level to above 99%.

Conclusion

AWS Summit Amsterdam brings together technology professionals and organisations to exchange ideas, share practical experiences and explore cloud innovation. 

For Ohpen, the event provided an opportunity to share how Agentic AI is being applied in mortgage technology and to contribute to the broader conversation about bringing AI into production. 

We would like to thank Amazon Web Services for the invitation to present and for the continued collaboration as we explore new opportunities to apply technology in meaningful ways. 

 

Wessel Stoop
Product Manager Innovation, Ohpen