From experimental bots to enterprise AI agents
Enterprise AI agents are software agents powered by artificial intelligence that can act on behalf of employees or systems, operating across business data and applications to automate tasks, make suggestions and trigger actions with built‑in governance, security and monitoring at organization-wide scale. After years of narrow pilots, these agents are now moving into the core of enterprise automation. The shift is marked by large, workforce-wide deployments where AI assistants are embedded in everyday tools and connected to back‑office systems. Instead of isolated proof‑of‑concepts, companies are building operating models that cover ownership, lifecycle management and risk. This change is driven by joint efforts between hyperscale platforms like Microsoft and service providers that understand industry processes, giving enterprises both the technology and the guidance needed to move from experimentation to production at agentic AI scale.
KPMG: Governing AI agents with Agent 365 and Copilot
KPMG is using Microsoft Agent 365 as a control plane to manage how enterprise AI agents are deployed, monitored and updated across its global organization. In parallel, KPMG member firms are expanding Microsoft 365 Copilot access across a workforce of more than 276,000 professionals, embedding agentic AI into daily work in audit, tax and advisory. According to KPMG’s Global Chief Digital Officer Lisa Heneghan, this expansion “requires strong foundations in governance, visibility and accountability.” Agent 365 extends KPMG’s Workbench ecosystem with centralized oversight of agents working across systems, data and business processes, while the KPMG Trusted AI framework adds structure for governance, risk and compliance. For clients, KPMG is turning these same tools into an enterprise AI operating model, helping them move beyond pilots by integrating agents into existing workflows, defining clear ownership, and securing sensitive data and intellectual property in Copilot deployment projects.
Atos: Workforce-wide agentic AI at production scale
Atos is demonstrating what full-scale deployment of enterprise AI agents looks like by rolling out Microsoft 365 Copilot to all 56,000 employees in 54 countries. The organization is standardizing on Microsoft 365 E7, which unifies Microsoft 365 E5, Copilot and Agent 365, and adds Entra, Defender, Intune and Purview for integrated security and compliance. Atos already manages a population of 19,000 AI agents through Agent 365, using the same admin and security workflows its IT and security teams run for other enterprise services. This is paired with Microsoft Copilot Studio and Microsoft Foundry to design, build and operate agents for internal IT, business functions and client solutions under a single operating model. Group Chief Digital & Information Officer Frédéric Aubrière calls this “the most significant technology investment in our people that Atos has made in a generation,” signaling a shift from pilots to AI‑first ways of working.
Why security, trust and governance now lead AI conversations
As organizations push agentic AI scale, security and trust frameworks have become the primary gatekeepers for production deployment. Both KPMG and Atos place Microsoft Agent 365 at the center of their approach, because it provides end‑to‑end governance over agents that act on behalf of users, operate with their own credentials, or come from partner ecosystems. This means AI agents are discovered, approved, monitored and retired through established workflows rather than ad‑hoc processes. KPMG’s Trusted AI framework adds risk, compliance and lifecycle controls on top, emphasizing clear ownership, accountability and auditability for every agent. Atos reinforces the model with security and compliance services from Microsoft 365 E5, Entra, Defender, Intune and Purview. Together, these stacks answer leadership questions about data protection, regulatory compliance and operational safety, turning AI agents from experimental tools into managed, enterprise-grade automation assets.
Integrating agents into systems, work and client delivery
The move from pilot projects to production hinges on how well enterprise AI agents integrate with existing systems and ways of working. KPMG is embedding Copilot into everyday Microsoft 365 applications while using Agent 365 to oversee agents that sit in its Workbench ecosystem and connect to client-facing platforms. Client stories such as Integra LifeSciences show phased roadmaps, where Copilot is integrated into core functions like global supply chain and regulatory affairs, with adoption and ROI tracked against an enterprise AI operating model. Atos is building Sovereign Agentic AI studios on top of Copilot Studio and Microsoft Foundry, giving customers a path to design and run agents within regulated environments. In both cases, strategic collaboration between platform vendor and service provider is key, blending cloud tools, security controls and process expertise so enterprise automation aligns with real work rather than staying in the lab.
