Enterprise AI agents move from pilots to production
Enterprise AI agents are software-based assistants that act on behalf of employees or systems, using organizational data and tools to perform tasks, automate workflows, and support decision-making within secure, governed technology environments. KPMG and Atos are now turning this concept into day-to-day reality by deploying Microsoft’s agentic AI stack across their organizations and client ecosystems. Both firms are standardizing on Microsoft 365 Copilot deployment, combined with Agent 365 as the control plane for managing AI agents across complex environments. Rather than isolated experiments, these projects treat AI agents as core infrastructure, embedded into office productivity tools, service delivery platforms, and industry solutions. The aim is to improve speed and consistency of work while maintaining strong enterprise AI security, so AI agents can operate at scale without exposing data or breaking compliance requirements.
KPMG’s Agent 365 strategy: trusted AI at network scale
KPMG is adopting Microsoft Agent 365 to govern how AI agents are deployed, monitored, and updated across its global organization, while expanding Microsoft 365 Copilot access to more than 276,000 professionals. This combination strengthens the KPMG Trusted AI framework, giving clients a repeatable way to move from trials to enterprise AI agents in production. According to KPMG Global Chief Digital Officer Lisa Heneghan, scaling AI “requires strong foundations in governance, visibility and accountability” so responsible AI sits at the center of culture and client work. Agent 365 extends KPMG’s Workbench ecosystem, providing centralized control of AI agents that operate across systems, data, and business processes. That control layer is key to enterprise AI security: it supports policy enforcement, lifecycle management and clear ownership, while KPMG teams use Copilot in everyday tasks across audit, tax and advisory to raise quality and consistency of services.
Atos: workforce-wide secure agentic AI on Microsoft 365 E7
Atos is taking a workforce-wide approach, rolling out Microsoft 365 Copilot to all 56,000 employees across 54 countries on the new Microsoft 365 E7 platform. E7 combines Microsoft 365 E5, Copilot and Agent 365, reinforced with Microsoft Entra Suite plus advanced Defender, Intune and Purview capabilities, to create a tightly governed environment for secure agentic AI. This gives Atos a single stack where productivity tools and AI agents share the same security and compliance baseline. Agent 365 provides end-to-end agent governance over a fast-growing population of 19,000 AI agents, including those acting on behalf of users and those operating with their own credentials. By using the same admin and security workflows its IT and security teams already run, Atos reduces operational friction while keeping central oversight of agent behavior, access rights and risk exposure across its ecosystem.
Embedding AI agents into tools, workflows and client services
A shared pattern in both initiatives is tight integration of enterprise AI agents into existing tools and workflows rather than building separate AI silos. KPMG is weaving Copilot into Microsoft 365 so professionals can use AI directly in Outlook, Teams, Word, Excel, PowerPoint and SharePoint, while Agent 365 links those experiences to back-end systems through the Workbench ecosystem. Atos is doing the same with Microsoft 365 Copilot, plus Copilot Studio and Microsoft Foundry, creating an operating model where agents are designed, built and operated as standard components of IT and business processes. These capabilities also underpin Atos’ Sovereign Agentic AI studios and KPMG’s client-facing engagements, where lessons from internal use feed into client implementations. The result is a more consistent approach to enterprise AI security, governance and adoption that clients in regulated sectors can reuse with clearer guardrails.
From security-first foundations to new operating models
Security and trust are central to both rollouts, and they are reshaping how work is organized. By treating Agent 365 as an enterprise control plane, KPMG and Atos are standing up new AI operating models that define ownership, risk and compliance processes for agents from day one. Integra LifeSciences, a KPMG client, shows how this plays out: the company has set up an enterprise AI operating model and a dedicated team to ensure “every deployment is responsible, secure, and compliant,” while embedding Copilot into functions such as Global Supply Chain, Regulatory Affairs and Medical Affairs. Atos, in turn, is using its own transformation as a template for client programs, especially in regulated industries. In both cases, secure agentic AI is not a separate project: it is becoming part of how organizations plan operating models, measure ROI, and design work around AI-powered assistance.
