From Translation Excellence to Agentic Intelligence

When software starts to think, plan, and act across your business systems on its own, it is more than automation – it is collaboration.

At this year’s DeepL Dialogues user conference, our portfolio company DeepL introduced DeepL Agent, the next chapter in a journey the company has been shaping for years and one eagerly awaited by its global user base.

Following a successful beta program with over 1,000 early adopters, DeepL Agent is now generally available and ready to transform how organizations collaborate, automate, and get work done.

We were fortunate to be among the early testers of DeepL Agent and experience its capabilities firsthand. What impressed us most was its ability to plan and execute complex, multi-step business tasks with real autonomy, showing a level of robustness and persistence we haven’t seen in other AI tools. Rather than another incremental productivity feature, DeepL Agent represents a genuine shift in how we think about intelligent systems inside the enterprise. (See my initial review here.)

Replacing Macros with Autonomy: How DeepL Agent Changes the Game

Most AI agents today resemble glorified macros: pre-scripted workflows occasionally enhanced with an LLM-generated sentence. They can assist, but not truly think.

DeepL Agent is built differently. It autonomously plans and executes complex, multi-step business tasks based purely on natural language prompts. It understands intent, determines the best way to achieve a goal, and then carries out the necessary actions through standard browser interfaces – just like a human would.

This browser-based interaction model is a critical design choice. Instead of requiring new APIs, connectors, or integrations, DeepL Agent can access and control existing systems directly, using the same interfaces employees already rely on. This allows it to work across a company’s full technology stack with no additional infrastructure or rewiring required.

In our early tests, DeepL Agent executed a complex analytical workflow that involved OCR, accessing dozens of websites, and updating our Salesforce CRM, running flawlessly for nine hours straight. No other agentic system we have tried even came close in stability or persistence.

This robustness and autonomy make DeepL Agent a new benchmark for real-world business use. It is designed to handle unpredictable digital environments, recover from transient errors, and maintain full task continuity over extended periods.

Governance as a Feature, Not an Afterthought

DeepL Agent is built from the ground up for enterprise deployment. It includes human-in-the-loop controls, ensuring users can review, guide, or intervene in any workflow when necessary. Every action is logged and auditable, giving organizations the transparency they need to meet internal compliance and governance requirements.

Data security is also central to the system’s design. DeepL operates under ISO 27001 certification and follows EU AI Act principles, ensuring that enterprise data remains private, protected, and never used for model training.

The result is a powerful, secure foundation for the next generation of workplace automation.

Not Theory, Real Stories: Early Adopters Already Reshaping Work

During the DeepL Dialogues launch, Chief Scientist Stefan Mesken shared stories of organizations like Perk, Caritas Schweiz, and the University of Kassel that are already leveraging DeepL Agent to transform their operations.

Perk uses DeepL Agent to accelerate complex multilingual workflows, reducing manual work and increasing quality. Caritas Schweiz applies it to streamline administrative processes, freeing up time for mission-critical tasks. At the University of Kassel, the agent supports data management and analysis across research workflows.

These examples show that DeepL Agent is not a prototype or experiment but is already delivering tangible business outcomes across industries.

A Human–Machine Partnership, Not Replacement

DeepL describes the product best: “DeepL Agent isn’t just another tool; it’s the AI collaborator that gets every tool and every system working for you.”

This captures what makes the product unique. DeepL Agent does not replace existing tools; it orchestrates them. It works across applications, departments, and workflows, acting as a cohesive layer of intelligence that unites the organization’s digital systems.

It also reflects DeepL’s broader philosophy: AI should augment human capability rather than replace it. By handling repetitive or complex procedural work, DeepL Agent allows people to focus on creativity, problem-solving, and strategy.

The Dawn of Autonomous Business Collaboration and the Broader Implication for AI in Europe

For Europe’s AI ecosystem, DeepL Agent’s launch is a significant milestone. It highlights how European innovators are moving beyond foundational models to create applied AI systems that combine intelligence, reliability, and regulatory compliance.

In a landscape often defined by regulation and caution, DeepL Agent is a reminder that European AI can lead not by scale, but by trust, precision, and depth.

With DeepL Agent, the company that revolutionized translation is now pioneering a new category: autonomous business collaboration.

We are proud to have supported Jarek Kutylowski and the DeepL team from their early days and to follow their journey closely as they continue to build technology that is both deeply intelligent and profoundly useful.