Munich – November 03, 2026 -- Atos is pitching its Sovereign Agentic Studios as a route for German companies to move agentic AI out of pilot mode, as the share of firms using AI jumped from 17% to 41% in one year.

The figures come from a March 2026 representative survey by the digital association Bitkom, covering companies with at least 20 employees. A further 48% are planning AI deployments or discussing them. More than half of current users report a measurable contribution to business success.

Atos Sovereign Agentic Studios cover design, build and operation of AI agents

The approach spans the full value chain of agentic AI: conception, development and operation, plus the governance structures and underlying infrastructure. Each project starts from a prioritized business process with defined goals and measurable results, not from the technology.

Data sources are then consolidated, agentic functions developed and processes orchestrated, with security requirements, governance and sovereign provision of data and AI built in. Scaling into production should follow only once benefit, security and operability are proven, according to Thomas Götz, Head of Technology & Innovation at Atos Germany, Austria & Central Europe.

Operationalization, not model performance, is the main scaling hurdle

Götz identifies operationalization as the biggest challenge. A pilot can work in lab mode with a small team and few data sources, he said, but production adds legacy systems, governance and compliance. Data protection, regulatory requirements, integration into operational processes, cost-effectiveness and scalability must also be addressed.

Atos calls for human approvals and audit logs on autonomous agents

Because agentic systems do not always react predictably, Atos says companies must limit their room for action. Required controls include:

• Clearly defined identities and mandates for individual agents
• Permissions set by humans
• Approval processes for critical decisions
• Documented activity and continuous monitoring of behavior, quality, cost and results

Critical decisions should be secured through human-in-the-loop approvals, and all activity should be logged in an auditable way, Götz said. Companies must also be able to adjust permissions dynamically and block individual actions. The degree of autonomy should be set by risk, regulatory requirements and business benefit.

Atos ties digital sovereignty to open standards and hybrid cloud

Atos argues that companies must keep control over where and how their data, models and workloads run and who can access them. It names open standards, interoperable architectures and a flexibly orchestrated hybrid-cloud model as prerequisites. The aim, per Götz, is to create room for in-house innovation rather than purchasing it exclusively.