Conversations with more EU IDP companies

last updated:

Conversations with more EU IDP companies

by:
last updated:

With the loudest noises coming from the Silicon Valley, New York City, London, and India, it’s easy to overlook the impressive AI companies growing inside the EU.

The European Union continues to produce technically-impressive and innovative IDP startups. We were briefed this week by two EU-based companies with similar origin stories: Alphamoon and Parashift. Both companies began as AI consultancies and custom software development specialists during the 3rd Wave of IDP, before they decided to build their own IDP products. 

Alphamoon (Poland)

Alphamoon has the honor of being the first Polish IDP company we’ve spoken to. The founders are AI experts who opened an AI consultancy around 2016. A lot of work involved document processing and data extraction, and in 2022 the decision was made to turn their document AI creations into an IDP platform. The founders brought in Norbert Raus as CEO to lead the efforts. The company has 30 employees and has raised $2.5 million so far. 

As to be expected from AI experts, the product line is powered by advanced 3rd and 4th Wave document AI methods such as visual document understanding (VDU), multi-modal AI, and large language models (LLMs).

Alphamoon has developed an innovative continuous learning method that it claims can improve the models incrementally (such as 1 document at a time) and can update the models in near real-time. This of course would be pretty cool, because document AI model training typically takes a day to a week and consumes a ton of GPU cycles. Not something you normally want to do incrementally. When I challenged Norbert on that claim, his response was, OK that’s our secret sauce and I cannot let you print the recipe. Fair enough. We’ll keep a close eye on this development. 

On the generative AI front, Alphamoon has a head start on most of the wolfpack. While most IDP vendors are still in alpha or beta, Alphamoon has just released version 2.0 of its zero shot Custom Model builder. Including a more powerful module to extract tabular data from any document and increased data extraction accuracy. The custom model also has the ability to edit masks so that you extract only the data you need. I saw it in action on my favorite unstructured test document – the stock purchase agreement – and it worked well. And what would an EU story be without a GDPR mention? In the latest release, the models are now hosted at Microsoft Azure UK, so EU customers can meet GDPR compliance. 

Parashift (Switzerland)

I always enjoy a good origin story from founders, and Alain Veuve, CEO and co-founder, did not disappoint. Around 2016, Alain was working on a general ledger automation project and wanted to add invoice data extraction. He searched for a cloud-based single API with very little training required. He look at Kofax and ABBYY but neither was suitable and both required too much training and finetuning for each new document and customer.

Amazed that an API solution able to handle different document types out of the box did not yet exist, Alain and his co-founders thought, we can do this. So they started Parashift in 2018 and have since built an impressive IDP API cloud service sold mainly through their partner network. Their mission is to become the central point in the enterprise for document separation, classification and extraction. 

The company has grown to over 40 employees and 150 customers. How do they compete and win with so many other options on the market? 

Firstly, Parashift is laser-focused on offering an easy to use single API call that can provide document separation, classification, and data extraction across hundreds of document classes.

Secondly, Parashift is not limited to any one document use case or industry. The team has built a general purpose data network with 400+ ML models, all based on their proprietary Document Swarm Learning® technology that maximizes continuous AI-based learning across all use cases and customers in a fully GDPR-compliant way. It was further described as an extremely well-annotated data set trained on millions of documents; each model benefits by tapping into the Swarm’s training on common fields.   

This is not currently meant as an end to end IDP platform. Parashift designed the APIs to be integrated easily with other software that will provide the additional process functionality such as HITL interfaces for error correction, workflows, external database lookup and validation, and connectors to enterprise software. 

As for generative AI, Parashift is still experimenting with LLMs and formulating their product roadmap. For now, their document-specific ML models seem to work fine out of the box with no hallucinations to worry about.

Don’t under-estimate the EU

With the loudest noises coming from the Silicon Valley, New York City, London, and India, it’s easy to overlook the impressive AI companies growing inside the EU. Everyone seems to know about the Rossum.ai story (and rightfully so, they have executed well). Our EU coverage also includes Acodis, AIDA, Altilia, Anyline, Cortical.io, EXB Labs, Hypatos, Insiders, Klippa, Konfuzio, Levity, Metamaze, Planet.ai, ReciTAL, Retarus, Scalehub, Scandit, Skilja, Skwiz, and TCG Process. 

 

Leave a Comment

We use cookies to ensure that we give you the best experience on our website. If you continue to use this site we will assume that you are happy with it.

Work Intelligence Market Analysis 2024-2029