Golem.ai, the Paris-based startup that just raised €5 million according to Les Echos, offers companies artificial intelligence software capable of analyzing human language.

The technology Golem.ai technology converts human language into data that can be interpreted by the machine. The aim is to enable companies to automate certain repetitive tasks. For example, software with Golem.ai will be able to understand requests made by humans, for example through a chatbot or voice assistant. If the requests have low added value, such as train schedules, the software will automatically generate an appropriate response. Conversely, the program will contact a human operator if the problem is more complex.


In this way, Golem.ai offloads customer services and allows operators to focus on value-added tasks. Golem.ai's AI is also very good at document analysis and can speed up document searches or certification procedures. "Golem.ai can query a document of several thousand pages"Thomas Solignac, co-founder and director of the startup, summarizes in the columns of the Tribune. The startup sells the technology itself, but can also create custom software in exchange for a subscription.

Golem.ai on the international scene

Available in 8 languages, including French, Golem.ai distinguishes itself from its competitors with a linguistic approach that borders on the social sciences. The method, developed over six years, allows them to use little customer data, unlike commonly used machine learning tools. For example, when deploying a chatbot for a company's internal IT department, the startup's teams will communicate with the employees in charge of maintenance. They will use them to identify repetitions in queries, and then enter that specific vocabulary into Golem.ai. This could be the name of the workroom, proper nouns or the name of a specific material. After these few hours of vocabulary recording, the tool can already be put into production.

Machine learning, which is very popular in AI creation, suffers from the "black box" effect: we know the input and the output, but not the process between them. This is where Golem.ai stands out again:"Our algorithm's decisions can be evaluated by an auditor who can explain its reasoning", explains the co-founder, who regularly deconstructs the fantasy around AI. This traceability of decisions is very useful in banking, insurance or customer service ... The startup, which has 18 employees, claims to have a dozen customers in these sectors.

Two big fundraisings...

Founded in 2016 by young Epitech graduates, the Paris-based startup raised €1.3 million in two rounds of funding last year and again €5 million last March. The company's research and development department is working on creating industry-specific standards to offer products that are already tailored to customers' needs.