Lee Foster - 30.07.202620260730

Join our community of 1,000+ IT professionals, and receive tech tips and updates once a week.

You Already Own the Platform. Now It Is Time to Activate It

When IT leaders hear “agentic AI”, the instinct is often to ask what they need to buy.

A better question is what they already have.

For organisations invested in Microsoft 365, Azure and ServiceNow, much of the required foundation may already be in place. Identity services, business data, connectors, workflow engines, security controls and employee-facing applications can all become part of an agentic AI operating model.

Agentic AI does not necessarily require an organisation to replace its technology estate. In many cases, the opportunity is to activate and connect the platforms it has been building, governing and integrating for years.

From assistants to agents

Traditional AI assistants generally respond to a prompt and wait for the next instruction.

Agentic AI systems can go further. They can plan multistep work, use approved tools, interact with systems and continue working towards a defined outcome within established permissions and controls.¹

That does not mean agents should operate without oversight. The level of autonomy should reflect the sensitivity, risk and reversibility of each task.

Interest is already moving beyond experimentation. McKinsey’s 2025 global survey found that 23 per cent of respondents said their organisations were scaling an agentic AI system in at least one business function. A further 39 per cent said their organisations had begun experimenting with AI agents.²

However, adoption is still emerging. Most organisations that are scaling agents are doing so in only one or two business functions.

The practical advantage may therefore belong to organisations that can identify governed, high-value workflows and deploy agents through platforms employees already use.

Microsoft 365 provides the distribution layer

Microsoft 365 gives organisations an established environment through which agents can reach employees.

Microsoft reported more than 20 million paid Microsoft 365 Copilot seats in its FY26 Q3 earnings call, up from 15 million in the previous quarter. Microsoft also said Copilot seat additions increased by more than 250 per cent year on year and represented its fastest growth since launch.

That reach matters because agents do not create value in isolation. They need to appear where people communicate, find information and complete work.

Microsoft’s 2026 Release Wave 1 plans include AI-powered and agentic capabilities across Dynamics 365, Power Platform, Copilot Studio and role-based Microsoft 365 Copilot experiences. The planned releases span April to September 2026.

Depending on the use case and licensing model, organisations can publish or integrate agents through familiar Microsoft environments such as Teams and Microsoft 365 Copilot. This can reduce the need to introduce an entirely separate employee interface.

It does not eliminate change management. Employees still need to understand what an agent can do, what data it can access, when its work requires review and how to report unexpected behaviour.

Copilot Studio provides the business agent layer

Microsoft Copilot Studio is a platform for building, connecting, publishing and managing agents.

For organisations already using Power Platform and Microsoft 365, it offers a practical path for creating agents that work with existing data, connectors, workflows and channels.

One important development arrived in May 2026, when Microsoft announced the general availability of computer-using agents in Copilot Studio.

These agents can interact with websites and desktop applications through their user interfaces. They use vision and reasoning to identify elements and complete tasks, which can help organisations automate work in systems that do not provide suitable APIs.

This capability may be useful for legacy applications, but it requires careful testing. User interfaces can change, actions can have unintended consequences and sensitive workflows may require human approval before completion.

Governance capabilities are also becoming more visible within the authoring experience. Copilot Studio’s agent status summary helps makers identify issues affecting readiness, governance and security, while agent-level protection status provides information about an agent’s security posture.

These controls are important because the risk profile of an agent is determined by more than its language model. Identity, permissions, connectors, actions, accessible data and publishing settings all affect what the agent can see and do.

Microsoft Foundry and Agent Framework provide the engineering layer

Some business scenarios can be addressed through low-code tools. Others require deeper software engineering, custom orchestration or integration with specialised models and services.

Microsoft Foundry provides services for building, evaluating, deploying and governing AI applications and agents within the Azure ecosystem.

Microsoft Agent Framework 1.0 reached general availability on 2 April 2026. It brings concepts and capabilities from AutoGen and Semantic Kernel into a supported, open-source framework for building agent applications and multistep workflows across .NET and Python.

The framework supports components such as model clients, agent sessions, context providers, middleware, tool integration and multistep workflows.

This creates a useful division of responsibility.

Copilot Studio can support business-led and low-code agent development. Microsoft Foundry and Agent Framework can support engineering-led applications that need more control over architecture, orchestration, testing and deployment. Microsoft 365 can provide a delivery surface through which employees interact with selected agents.

These layers should not be treated as interchangeable. The appropriate layer depends on the use case, technical complexity, operational risk and governance requirements.

ServiceNow brings agents into enterprise workflows

For many IT organisations, ServiceNow already coordinates service delivery across IT, employee, customer and operational workflows.

ServiceNow AI Agents can use agentic workflows, platform records, approved tools and existing automation to pursue defined objectives. AI agent actions can also be incorporated into Flow Designer workflows, allowing outputs to be passed into subsequent workflow steps.

This means organisations can introduce agents into processes that are already configured within ServiceNow rather than rebuilding every workflow on a separate AI platform.

The governance layer is equally important.

ServiceNow launched AI Control Tower in 2025 as a central environment for managing AI agents, models and workflows across ServiceNow and third-party systems.¹¹

In November 2025, ServiceNow and Microsoft announced integrations intended to give AI Control Tower visibility over agents deployed through Microsoft Foundry and Copilot Studio.

At Knowledge 2026, ServiceNow expanded that relationship to Microsoft Agent 365. The integration is designed to extend AI Control Tower governance across Microsoft’s broader agent ecosystem and make ServiceNow AI specialists available through the Microsoft Agent 365 Marketplace.

For organisations operating agents across Microsoft and ServiceNow environments, this supports a more consistent approach to agent discovery, approval, monitoring and governance.

It does not automatically create one perfect control plane. Organisations still need to define ownership, access policies, approval requirements, logging standards, incident processes and accountability across both platforms.

The activation mindset

The opportunity for CIOs and CTOs is not simply to add agents to every workflow.

It is to identify where existing platforms can support useful, governed and measurable agentic work.

A Microsoft 365 environment can provide identity, collaboration and employee-facing delivery surfaces.

Azure and Microsoft Foundry can provide engineering, deployment and AI infrastructure.

Copilot Studio can provide a low-code environment for building and managing business agents.

ServiceNow can provide workflow execution, service management context and cross-platform AI governance.

The organisations likely to move most effectively will not be those that deploy the highest number of agents. They will be those that select the right processes, establish clear controls and demonstrate measurable improvements.

A practical activation program should begin with:

A review of the platforms, licences, connectors and automation already in use
A shortlist of repetitive or fragmented workflows with measurable business value
An assessment of data quality, identity controls and permissions
Defined boundaries for agent autonomy and human approval
Evaluation criteria covering accuracy, safety, cost and business outcomes
A governance model for monitoring, change management and accountability

Agentic AI should not be treated as an isolated technology experiment. It should be considered an evolution of the organisation’s existing digital workplace, cloud and workflow platforms.