Almost every employee who sits in an office knows the problem. There are always multiple windows, tabs, and apps open, and for every little thing, there’s a separate service that is specialized and perfected for its task. Communication between services? Often poor. “We humans are the glue between the apps,” says Stefan Mesken, Chief Scientist at DeepL. And he immediately presents the solution.
DeepL Dialogues is the in-house event for introducing new products. There, Mesken presents the DeepL Agent, a classic AI agent, which is on par with those from other providers. It offers the same capabilities, is based on a reasoning model, can be adapted to one’s own requirements, and al…
Almost every employee who sits in an office knows the problem. There are always multiple windows, tabs, and apps open, and for every little thing, there’s a separate service that is specialized and perfected for its task. Communication between services? Often poor. “We humans are the glue between the apps,” says Stefan Mesken, Chief Scientist at DeepL. And he immediately presents the solution.
DeepL Dialogues is the in-house event for introducing new products. There, Mesken presents the DeepL Agent, a classic AI agent, which is on par with those from other providers. It offers the same capabilities, is based on a reasoning model, can be adapted to one’s own requirements, and already complies with ISO certification (27001). And above all, this AI agent is intended to be the new adhesive between services. Since agents can do everything that humans do on a computer, it can handle the integration and merging of different tasks, as DeepL envisions it. This, in turn, sounds quite different from many other AI agent providers. While they often say that agents make people more efficient by directly taking over tasks—whatever and however they may be—DeepL addresses a more overarching problem. Although the DeepL Agent can also create analyses, reports, and more, even across different languages.
Stefan Mesken at DeepL Dialogues.
Of course, you also have to tell the DeepL Agent what to do. “It’s no surprise that it’s a chat interface,” says Mesken. But it is just an interface through which a lot can be achieved—in natural language, without major hurdles or tricks, or tabs, buttons, and hidden checkboxes that need to be set. According to DeepL, so-called knowledge workers switch between different applications 1200 times a day, and eleven hours a week are spent searching for data in different systems.
The agent has been available to around 1000 testers for two months now. Now it’s here—for everyone. “We want to unlock human potential by transforming work related to language and beyond. Translation remains our core, and with today’s announcements, we are expanding this foundation while setting new standards for agentic AI,” says DeepL CEO and founder Jarek Kutylowski.
Customization Hub for Your Needs
In addition to the DeepL Agent, the company now offers a Customization Hub. Through this platform, individual needs can be defined—for example, glossaries, style guides, and Translation Memories, which are recurring translation guidelines.
DeepL’s customers are companies that operate internationally and accordingly have employees in different locations and languages. Translation remains an important part of DeepL’s products. The language offering for corporate customers is significantly expanded again—by 70 languages. In the EMEA region, all 24 official EU languages will be available in the future, as well as Croatian, Bosnian, Serbian, Swahili, Afrikaans, and Malagasy.
Kutylowski believes that so far we are only scratching the surface of what AI can bring to the corporate world. Innovation happens, but often in the shadows. Employees experiment find ways AI can help them, but there is often a lack of thinking bigger about these ideas and solutions. “The culture in companies must also allow for failures give employees the opportunity to experiment,” says the CEO.
(emw)
Don’t miss any news – follow us on Facebook, LinkedIn or Mastodon.
This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.