Document processing infrastructure isn’t the sexiest catch-phrase, but real productivity gains are pretty cool. Send AI allows companies to fine-tune AI models for their specific data-extraction use case. An example use case are insurance companies who receive claim in the most creative formats possible (mobile phone pictures, multiple docs in one frame, noisy backgrounds) and need to extract data & input these in their systems to action them.
Although the model can work independently, humans are very much at the steering wheel. A user can quickly open a document, and the model will classify it and label pieces of information, which the user can then verify through an intuitive visual scaler. “Yes, this is indeed a GP prescription, and that is a patient ID number on the form,” check!
The user’s growing trust in the model’s ability to classify and label data, coupled with their validation of the outcomes, co-trained the model. This human-in-the-loop philosophy gradually makes the user’s job easier while the model undergoes constant reinforcement learning, becoming increasingly precise. It’s a powerful use-case specific feedback loop that generic AI models simply cannot match.
But it wasn’t just the product and market opportunity that captured our attention; it was their entrepreneurial grit. In our view, here were two humble Dutch techies, creating a state-of-the-art enterprise solution from their bedroom office, capable of rivaling or even outperforming the solutions of UIpath and others. We believe Thom and Philip have a real chance to build a giant company.
