Analysts say enterprises face a trade-off: reduced operational toil versus increased administrative overhead and governance requirements.
Microsoft is betting big on agentic AI to simplify and automate cloud operations by introducing an agentic mode to Azure Copilot that could surface insights and provide recommendations, but not take any actions.
While the hyperscaler is pitching the upgraded Copilot as a tool that simplifies cloud operations at scale by taking actions on its own, as opposed to traditional management tools that would require manual intervention and specialized skills, analysts say that instead of reducing complexities, the tool could introduce a few of its own.
How is the new agentic Azure Copilot different?
The agentic Azure Copilot, showcased at the company’s annual Ignite conference, utilizes a reasoning model to orchestrate six agents, easing the complexity of tasks, such as modernization and migration of legacy applications, deployment of cloud infrastructure, optimization, observability, resiliency, and troubleshooting.
This ability to orchestrate between agents on its own, based on user intent, is what differentiates the agentic Copilot from its previous version that depended on a more general-purpose large language model (LLM) to recommend tools and actions.
The agentic Azure Copilot, which can be enabled by toggling a setting in the Azure portal, becomes accessible once activated across all workflows in Azure, including the command line interface (CLI).
Enterprises also have the option of saving chats on the upgraded Copilot to a location of their choice for governance and audit purposes.
Is the juice worth the squeeze?
Analysts don’t seem to agree with Microsoft’s assertions about traditional tools and their impact on complicating management of cloud operations, and thus the need for an agentic Copilot.
“While the upgraded Copilot can orchestrate and execute workflows on behalf of users in cloud operations, I’m not sure that this was a common problem that needed a solution, and traditional tools worked just fine. Enterprises should look at what really changes while considering adoption. Is it for the better, or are we just ‘agentic washing’ existing tech that works just fine,” said David Linthicum, independent consultant and retired chief cloud strategy officer at Deloitte Consulting.
The cloud strategy consultant was referring to the conundrum that he says most enterprises would have to face: the promised upside of simplified cloud operations versus the rising complexity of an administrator’s role in deploying the upgraded Copilot.
“There’s a fundamental tension here. On one hand, Copilot’s agent mode simplifies many day-to-day operational tasks for end users and operators. On the other hand, it introduces new administrative considerations — particularly around policy-setting, access control, and compliance,” Linthicum said.
“Setting up granular spend permissions, managing access to sensitive datasets, and enforcing retention or storage controls — all require a more thoughtful approach from administrators. I would advise my clients who are considering this to consider the additional work it would require, including increased complexity, and whether there would be value to be gained,” Linthicum added.
Derek Ashmore, agentic AI enablement principal at Asperitas, pointed out that agentic cloud operations don’t eliminate governance but instead amplify the need for it, complicating the role of administrators.
“…most enterprises won’t turn this on in a week. A realistic adoption cycle is 3 to 9 months, depending on cloud maturity. Early pilots will happen in non-production environments with narrow use cases before expanding to anything that can be remediated automatically,” Ashmore said.
Real benefit or just some tactical improvements?
Despite his concerns about the complexity for administrators, Ashmore said that enterprises could see some tangible benefit post-adoption and deployment of the upgraded Copilot.
“Once the policy layer and data-access patterns are well-defined, agents can dramatically reduce the day-to-day toil; drift management, cost optimization, incident triage, and cross-service troubleshooting become much lighter,” Ashmore said, adding that Azure Copilot could mirror the trajectory that enterprises saw with Infrastructure-as-Code: the setup phase required discipline, but the long-term payoff was consistency and velocity.
In contrast to Ashmore, Linthicum sees nothing substantial for enterprises beyond “just a few tactical improvements,” as he says that enterprises wouldn’t be able to point to any core improvements.
“In short, Microsoft is inventing a problem that they are solving with their new magic software,” the cloud strategy consultant said.
The analysts also pointed out that Microsoft isn’t alone or early to frame cloud operations explicitly in “agent” language inside the core cloud console, and AWS and Google are clearly on the same trajectory.
While AWS has Amazon Q for AWS, which already helps ops teams design, troubleshoot, and even provision resources from chat and CLI, Google Cloud has Gemini Cloud Assist for lifecycle management, Ashmore pointed out.
“Both the hyperscalers are looking to advance their capabilities with more agents,” Ashmore added. Microsoft’s upgraded Azure Copilot is currently in preview.
More Microsoft Ignite 2025 news:
- Microsoft Fabric IQ adds ‘semantic intelligence’ layer to Fabric
- Microsoft unveils Agent 365 to help IT manage AI ‘agent sprawl’
- Microsoft touts scalability of its new PostgreSQL-compatible managed database
- Microsoft unveils Agent 365 to help IT manage AI ‘agent sprawl’
