While our clients continue to bring incredible projects to the table requiring serious Power Platform expertise, the capabilities of Copilot Studio remain largely untapped. Throughout this year (2025) and last, Microsoft Copilot Studio has gone from “bot builder” to a serious agent platform with the ability to orchestrate work across the Power Platform applications and the wider Microsoft cloud products suite. 

For CIO’s, the question is no longer “Is this real?” but “Where does it fit in our tech stack, governance model and budget for the next 12 – 24 months?”

In this briefing, we explore:

• What Copilot Studio actually is in practical terms;
• How Copilot Studio differs from the standard Copilot product available across the M365 suite; and
• What’s changed in 2025 (autonomous agents, computer use, ROI analytics, governance controls).

What is Microsoft Copilot Studio – in practical terms?

At the time of writing this article, Microsoft describes Copilot Studio as a graphical, low-code tool for building AI agents and agent flows, with native ties into Microsoft 365 and the Power Platform… Which in itself should give you a sense of how fast Microsoft is moving with this product. Tomorrow its capability may be vastly expanded.

In practice, one can build and configure agents to achieve the following:

• Build agents that can answer questions, call actions, and now even act autonomously based on events.
  As an example, using the in-built agent prompting tool (which operates like ChatGPT for building a Copilot agent) you may configure a simple agent that provides insights on your website content as a knowledge source. This could be useful for websites with a high volume of content (such as a knowledge base). 
• Agents connect to your data using prebuilt or custom connectors and Dataverse, the same way Power Apps and Power Automate do.
  Imagine connecting your agent to your dataverse so you can ask questions of your Dynamics 365 or Power Apps data (ie Dataverse is offered as a standard connector). 
• Extend Microsoft 365 Copilot, or publish custom agents into websites, Power Pages, mobile apps and Teams.
  Consider an agent built into your Sharepoint site allowing users to ask questions about documents or find information faster. Or an agent in Teams that automatically alerts customer service staff to new tickets in Dynamics 365.
• Everything sits inside your existing Microsoft tenant with environments, security roles and DLP policies alongside the rest of the Power Platform.
  So, information shared by users with the agent will not be used to train the underlying Large Language Model (LLM). 

With child agents, workflow builders and an array of pre-built MCP tools, it is an incredibly deep and capable product that aims to provide significant productivity gains and/or boost customer experiences. 

For organisations already invested in Microsoft 365, Dynamics 365 or Power BI, Copilot Studio is effectively the front door for AI agents across the estate – not a separate AI island. And this is where the rubber hits the road for organisations striving to reduce data silos and extract the maximum value from their data sets. Notwithstanding GAI’s excellent ability to analyse non-structured data (i.e. documents), data governance practices must be considered in your equation. However, this is a story for another day. 

How does Copilot Studio differ from the standard Copilot product available across the M365 suite?

Copilot on its own is the AI assistant sitting inside most Microsoft products. Think of it as the off-the-shelf user-facing helper that you may see in any of the M365 (Office) products, Dynamics 365, Power BI and so forth. You don’t build Copilot – you simply use it as a feature. 

Whereas Copilot Studio is a builder platform enabling you to build new agents with specialist functions tied to your knowledge resources and databases. As a builder, one can control where the agents are published – (website, Power Pages, Teams, M365, etc). It is a no/low-code platform aimed at solution designers (and developers) rather than end users. 

Put together:

Copilot = the thing people see and talk to
Copilot Studio = the place you build and control those things

For instance, a solution builder could configure…

• departmental/organisation-specific copilots (e.g. create a custom “Case Triage Copilot” that understands your organisation’s Dataverse model, policies, and workflows)
• sector-specific agents (e.g. “NDIS intake assistant”, “Grant triage assistant”)
• back-office agents that run behind the scenes via flows and events (e.g. wire it up to Power Automate so it can actually create cases/actions, route work, send notifications, etc).

What’s new in Copilot Studio in 2025 – and why it matters

Beyond the standard functionality, here are some of the new features that have shipped by Microsoft this year. Incredibly, Microsoft has been shipping monthly updates, but four themes in 2025 are especially relevant for CIOs. 

1. Autonomous and “agentic” capabilities mature

Copilot Studio now supports autonomous agents that can be triggered by events (rather than waiting for a user to start a conversation) and then reason over data and tools to complete long-running tasks.

In short: agents can monitor, decide and act inside governed guardrails.

• Copilot Studio allows users to build agents that run sophisticated operations on behalf of users, not just answer FAQs.
• Autonomous agents as either GA “General Availability” or broad preview in many enterprise tenants, with event triggers modelled as Power Automate flows under the hood.

Why CIOs should care

• This is a step beyond “chatbot at the front door”; agents can now watch for business events (new case lodged, referral created, claim status changed) and act without a human nudge.
• It pushes Copilot Studio into the same conversation as other agentic platforms – but with Microsoft’s identity, security and audit story layered in.

2. “Computer use” opens up legacy web and desktop apps

One of the most significant late-2025 additions is computer use – a Copilot Studio tool that lets agents drive a Windows desktop or browser session using a virtual mouse and keyboard.

• You describe the task in natural language; the agent then clicks, types and navigates its way through websites or desktop apps as a human would.
• This is now in public preview and documented as a way to handle systems with no API or connector.

Why CIOs should care

For many Australian gov, health and industrial environments (to name a few), key processes still depend on legacy line-of-business systems that are hard to integrate:

• ageing line-of-business apps on-prem
• supplier or regulator portals with brittle web UIs
• older clinical or finance systems without modern APIs.

Computer use means agents can:

• extract data from on-screen dashboards
• fill and submit complex forms
• bridge processes that currently rely on manual swivel-chair work.

Combined with Power Automate (cloud and desktop), this extends automation and AI into the last 20–30 per cent of workflows that have resisted integration providing fascinating opportunities for innovation. 

3. Analytics and ROI measurement become first-class

CIOs have been asking the same question: “If we invest in AI agents, how do we prove value?” Microsoft’s answer is agent analytics and savings tracking:

• Copilot Studio now includes a Savings tile on each agent’s Analytics page, where you can define expected time and cost savings per run or per tool, and then track realised savings over time.
• Analytics cover sessions, satisfaction and topic performance, and can feed into Power BI reports for exec visibility.

Why CIOs should care

• You can finally talk about agents in hard numbers: minutes saved, cost avoided, calls deflected.
• This supports business cases for further rollout and gives boards a clearer line of sight over AI impact.

4. Governance, configuration and roadmap clarity

Microsoft has also been tightening the governance story:

• The “What’s new” and release-wave documentation now breaks out capabilities for builders, admins and security teams, including environment-level controls and Copilot configuration options.
• Admin-facing updates in the Power Platform admin centre give more control over who can create, publish and share agents.

On the commercial side, Microsoft has also started simplifying Copilot licensing by bundling more Copilot capabilities into a single enterprise price point, which will affect how organisations plan AI budgets across M365, Dynamics and Copilot Studio.

Why CIOs should care

• Copilot Studio now aligns more cleanly with existing tenant, environment, DLP and RBAC structures – crucial for health, justice, human services and child safety workloads.
• There’s more predictability around the roadmap (through the Wave 1 and Wave 2 release plans) and a clearer admin experience for central ICT teams.

Where Newpath fits in your Copilot and Power Platform journey

For those already operating or building solutions in the Microsoft stack, the next 12–18 months are about moving from pilots to platform:

• selecting one or two high-value agent scenarios (typically intake/triage, knowledge-assisted case work, or executive insight)
• designing the right mix of Copilot Studio, Power Apps, Pages, Automate and BI
• putting governance, security and ROI measurement in place from day one.

Newpath can help by:

• running focused AI on Power Platform discovery workshops with your architecture and business teams
• designing the solution and governance architecture around Copilot Studio and the Power Platform
• delivering end-to-end pilots – including legacy integration via computer use – and then scaling out to additional programs and services.

If you’d like a deeper dive into how these late-2025 Copilot Studio capabilities map to your roadmap, we can start with a short workshop focused on one concrete scenario in your portfolio and build from there.