Killian Vermersch: Golem.ai simplifies work. Its Artificial Intelligence reads, understands and processes any text in any language. Unlike traditional solutions, Golem.ai is known for its symbolic AI approach, capable of modeling human reasoning without the need to collect data. It claims total transparency and explainability of the decision-making process of its AI and thus fights against the biases of machine learning.

Above all, this technology offers a sovereign alternative, capable of speaking all the languages and explaining each of its decisions, altogether with a frugal consumption of energy and resources.

Our main solution, InboxCare, allows us to analyze e-mails or any incoming message and categorize them. It is even possible thanks to InboxCare to automatically answer messages via an acknowledgment of receipt or a customizable automatic response. Our AI can also process contents from attachments, which is the reason why we need to work with a technology partner like PLANET AI, an expert in OCR processing.

PLANET AI: Why use an OCR?

Killian Vermersch: We have developed a natural language understanding Artificial Intelligence. It’s our specialty. In addition, our technology is based on symbolic AI, a method that requires no learning.

Since our technology processes raw text, it requires other solutions to get that text from other formats. Most of the content we process comes from documents, like PDFs, handwritten scans, images… There’s where an OCR comes in handy. After months of benchmarks, Golem.ai chose PLANET AI’s IDA as its main OCR partner to process all incoming content.

PLANET AI: What uses do you make of the extracted information?

Killian Vermersch: We work in different sectors of activity like banking, insurance, retail, defense, which need to analyze some specific intentions in all their messages. Specifically in the analysis of emails (content and attachments). All of our clients need to simplify message processing and extract information to make faster and better decisions.

For example in the field of retail, many companies are looking for the identification of contents related to the handling of customer complaints to respond quickly. Based on the data extracted by Golem.ai, companies can classify and prioritize requests or even respond to them directly.

For other companies, particularly in insurance, it is a way of quickly recognizing customer needs, such as following up on a claim, in order to be able to send the message to the right department.

PLANET AI: What are the benefits?

Killian Vermersch: The real value is saving time while increasing quality and customer satisfaction. It allows employees to spend more time on high-value work and less time doing tedious tasks. It is also a way, in the customer relations sector, to process customer requests more quickly and therefore the quality of request processing.

That is why it depends a lot on the choice of technology and its performance. This is why we work with PLANET AI. We need to have a high-performance rate (>90%) for the return on investment to be significant for the company.

PLANET AI: Is there any noticeable improvement in inbox automation with the implementation of the IDA Recognition feature?

Killian Vermersch: The Recognition feature both increased the quality of our results and made us able to process new types of documents, especially handwritten content.

When coupled with GPUs, PLANET AI’s IDA can also process a huge amount of documents with a small amount of hardware, allowing Golem.ai to process millions of documents while respecting its commitment to frugality.

Golem.ai also interviewed our COO, Jesper Kleinjohann, about technological challenges.