Not a pilot, a company-wide transformation

While most large companies are still testing AI assistants in pilot programs for a few hundred employees, Atos took a step that has no parallel in the corporate world. Microsoft 365 Copilot is being deployed company-wide — to all 56,000 employees across 54 countries. The applications people use every day (Outlook, Teams, Word, Excel, PowerPoint, and SharePoint) are getting integrated artificial intelligence.

"This is the most significant technology investment in our people in a generation," said Frédéric Aubrière, Group Chief Digital & Information Officer at Atos Group. "We're not piloting. We're deploying at enterprise scale, because we believe that agentic AI will define the next decade of professional services."

Atos's strategy is built on the "Client Zero" approach — the company first transforms itself and its own processes, and only then offers the same model to clients. It already manages a population of 19,000 AI agents across its entire ecosystem.

What Microsoft 365 E7 brings and why it matters

The foundation of the deployment is the Microsoft 365 E7 package, which combines three key layers into a single solution:

Microsoft 365 Copilot — a personal AI assistant integrated directly into office applications. It can summarize email conversations, generate presentations from documents, analyze spreadsheets, or automatically draft meeting notes from Teams calls.

Microsoft 365 E5 security & compliance — a security layer including Entra Suite for identity management, Defender for endpoint protection, Intune for device management, and Purview for sensitive data governance and regulatory compliance.

Microsoft Agent 365 — a central "control tower" for observing, managing, and securing all AI agents in the enterprise. It enables unified management of agents acting on behalf of users, agents with their own credentials, and third-party agents — all through the same administrative workflows that IT teams already know.

This architecture is critical for Atos because the company operates in regulated sectors such as defense, finance, healthcare, and public administration. Without unified identity management, security, and agent governance, deploying AI at this scale would not be possible.

Atos and its Czech branches — what it means for the local market

Atos has a long-standing presence in the Czech Republic — it operates branches in Prague, Ostrava, and Brno and employs hundreds of specialists in IT services, cybersecurity, and cloud solutions. The Czech branch of Atos, part of one of Europe's largest technology players, is directly impacted by this transformation.

For Czech companies and public institutions that use Atos's services — whether for IT infrastructure management, cybersecurity, or cloud migrations — this means their supplier will work more efficiently. Atos expects a 40–60% reduction in first- and second-level service tickets and 55–75% of incidents resolved entirely without human intervention.

In a broader context, this is a signal for the entire Czech market: if a European giant with 56,000 employees can securely deploy agentic AI in regulated industries, it opens the door for banks, insurance companies, and government institutions in Czechia as well. A model that domestic organizations can follow is now on the table.

Security as the foundation, not an add-on

Atos is no newcomer to cybersecurity — it is Europe's number one in cybersecurity. That's why it's fitting that when deploying AI across the entire company, it made no compromises. The architecture built on Microsoft Entra, Defender, Intune, and Purview means that every AI agent, every data access, and every action is under supervision.

Jared Spataro, Chief Marketing Officer for AI at Work at Microsoft, said: "We have entered a new era of work — one where the most ambitious companies aren't just adopting AI, but are rebuilding work around it from the ground up. That's exactly what Atos is doing."

Atos's strategy rests on three pillars: agentic AI for mission-critical environments, digital sovereignty (the customer retains control over their data and models), and end-to-end cybersecurity. At a time when the EU AI Act sets strict requirements for deploying high-risk AI systems, this is a model approach.

How Atos develops its own agents

Beyond Copilot itself, Atos is deploying Microsoft Copilot Studio and Microsoft Foundry — tools for designing, developing, and operating custom AI agents. It uses these agents internally for IT support, business functions, and client projects. The entire effort is wrapped under a unified operating model.

A key part of Atos's offering is its Sovereign Agentic Studios — specialized workplaces that help organizations get AI safely into production. The studios operate from hubs in Birmingham (for regulated industries), Irving, Texas, France, and Germany. They offer a structured process: identifying use cases, validating agentic workflows in controlled environments, and industrializing proven solutions.

In its startup accelerator Scaler, Atos collaborates with six technology companies — including Poolside (sovereign AI models), Ema (platform for enterprise AI agents), Klarity (automated contract and document review), and Noma Security (AI lifecycle security).

Concrete results: fewer tickets, faster contracts, more accurate proposals

Atos is already reporting measurable results from its first agentic deployments:

• 40–60% fewer tickets at L1 and L2 levels in IT support
• 55–75% of incidents resolved fully autonomously, without human intervention
• 20–40% reduction in mean time to repair (MTTR)
• 60% faster creation of first-draft business proposals
• 20% labor savings in contract review and 25% savings in procurement

One of the most impressive examples comes from client Satair, where AI agent Lilly shortened price quote processing from 24 hours to 2 minutes with 99% accuracy.

What this means for other companies

Atos shows that enterprise-wide deployment of agentic AI at scale is a reality today — provided it's built on the right security and governance foundations. Key lessons for other companies:

• Don't run pilots endlessly — Atos went straight to full-scale deployment
• Security and governance must be part of the architecture from day one
• The "Client Zero" model (first transform yourself, then help clients) is more credible than theoretical consulting
• 19,000 agents cannot be managed manually — a central control plane (Agent 365) is a necessity

For companies considering a similar path, Atos offers its Sovereign Agentic Studios as a service. And for Czech companies, there is a direct route — through the local branch in Prague.