If you are familiar with Microsoft Copilot, you’ll know its ability to draft emails and search for information. Microsoft has now introduced AI agents, which add extra automation to complete tasks and connect data. This article looks at how Copilot and agents work together and how you can modernise processes to scale AI impact by starting with a description.
Comparing Copilot and Agents
Introduced in 2023, Copilot is a personal digital assistant that drafts content, asks questions, and summarises information. The more you use it, the more personalised it becomes, boosting productivity.
In contrast, agents work on behalf of teams alongside people to manage business processes with additional automation.
Using your organisation’s data, files, and other internal knowledge resources, a simple agent can replace repetitive tasks by retrieving information and acting when prompted. However, autonomous agents offer the most significant potential to optimise processes.
What are Autonomous Agents?
Autonomous agents can adapt to changing conditions and work independently without requiring constant human intervention. For instance, a lead generation agent can follow instructions to research leads and identify the best ones for review. An approvals agent could automate review procedures by validating uploaded documents against your checklist criteria.
Autonomous agents can execute an entire end-to-end process, but we anticipate agents will be more commonly deployed to re-engineer specific process steps that involve manual administration.
Agents for Multiple Business Processes
Each agent can apply reason, remember, and seek clarification if needed. You’ll be able to train each agent with specific skills, such as sending emails, updating CRM records, posting Teams messages, or retrieving information. Crucially, agents are empowered to act using your defined knowledge.
Microsoft likens agents to unique apps that can run different processes. Because teams will have various use cases, agent libraries will serve these needs within organisations.
While agents are the new apps for AI-powered processes, Copilot is the user interface where you’ll primarily interact with agents across any workflow.
Using Copilot Studio to Build Agents
Copilot Studio is the unified platform where authorised users can build, connect and manage agents.
Because each agent brings together your organisation’s data and knowledge, their actions and decisions will fit your process and business context.
To understand opportunities for Copilot agent automation, identify triggers that signal when an agent should act. For example, a trigger could be a record update in Dynamics 365, receiving a specific email, or uploading a document to SharePoint.
To instruct agents, you can design their behaviour using natural language prompts, templates, or code. For instance, you could use conversational text to state the process goal and desired outcome. Each agent can be instructed with example scenarios and details of what should happen and which internal data sources and documented rules should be used.
Once instructed, agents can be enabled to take action by connecting them to your applications and knowledge resources. Depending on the type of agent, this can include Dataverse tables, Teams channels, SharePoint libraries, Power BI reports, and external systems.
A range of agent actions can be enabled, including updating a Dynamics 365 record, sending an email or retrieving an order history.
After configuring these steps, an agent generates responses in Copilot or extends Power Platform apps, portals, and other channels.
Building Agents in Power Apps
Another approach to creating agents will be to use the knowledge, logic and actions already built into your apps. When released, this will bring Copilot Studio directly into Power Apps to accelerate the agent creation process.
How Do Agents Differ from Power Automate?
You may wonder how agent automation capabilities differ from what’s already available in Power Automate?
Each is a complementary automation tool. Agents work alongside Power Automate, triggering flows or using flow outputs to enhance their capabilities. However, agents differ by using reasoning and using your organisation’s knowledge to provide responses and perform actions. Agents excel at understanding context and logic to complete complex tasks that cannot be managed by using Power Automate alone.
Agent and Copilot Security
Microsoft announced the Copilot Control System at Microsoft Ignite in November 2024. This provides additional tools to securely manage Copilot and agents within your organisation.
The Copilot Control Studio will roll out capabilities in early 2025 to help IT teams enforce permissions and governance policies, ensuring security and compliance, including:
- Data Protection with intelligent grounding on enterprise data, including SharePoint Advanced Management, at no additional charge for Copilot customers.
- Management Controls in the Microsoft 365 admin centre to govern Copilot and agent access and usage.
- Measurement and Reporting to track adoption and ROI, including Copilot Analytics and a Copilot Business Impact Report.
Microsoft also confirmed that Power Apps, Power Automate, Microsoft Copilot Studio, and Dynamics 365 are now delivered on a fully managed platform.
Prebuilt Dynamics 365 Agents
Alongside the ability to create your own customisable agents, Microsoft has also announced preconfigured Dynamics 365 autonomous agents. These prebuilt Dynamics 365 Copilot agents can be tuned with your organisational knowledge using Copilot Studio in Power Apps.
At the time of writing, the Dynamics 365 Copilot agents designed for sales, service, finance, and operational teams include:
Sales Qualification Agent to research leads, help prioritise opportunities and guide customer outreach activities.
Supplier Communications Agent to optimise your supply chain and avoid disruptions by autonomously tracking supplier performance, detecting delays and responding accordingly.
Customer Knowledge Management Agent that can autonomously add knowledge articles to scale your best practices.
Case Management Agent for D365 Customer Service to automate tasks throughout the case lifecycle. It can be triggered in several ways, including inbound emails and live chat sessions, to reduce admin time by automating case details and follow-ups. See a demonstration below.
The first Dynamics 365 Copilot agents will be available in public preview in early 2025.
Copilot Studio Licensing Changes
As Copilot Studio usage increases, a new licensing option enables the billing of Copilot Studio messages on a pay-as-you-go basis. This consumptive billing through Azure subscriptions will be available as an alternative to capacity-based billing.
AI Agents: Customer Examples
Recently success stories from early AI agent deployments include:
A retailer that uses an agent within its profit protection team to identify transactions for human review. This combats fraud by analysing transactions at scale to find patterns, such as repeated use of the same photo in refund claims. The agent also identifies and highlights recurring product problems.
To shortcut the validation of legal documentation, a financial services company uses an agent to review each document. The agent automatically completes a series of checks to authenticate or reject documents in line with the defined validation policies in SharePoint.
In streamlining its legal due diligence processes, an information provider has found that an agent can complete specific tasks 50% quicker.
A travel arrangement provider uses an agent to provide advisors and customers with personalised support. The agent connects with various back-end systems to provide informed, accurate responses that resolve chat queries.
A manufacturer using an agent to increase billing accuracy and save shipping costs by assessing and tracking invoicing errors that incur substantial overpayments.
A global management consulting firm uses an agent to streamline its onboarding process. This includes processing emails, checking existing engagements, identifying the right expert resources, and notifying the assigned teams.
Implementing Autonomous Agents
Autonomous agents work alongside teams to sharpen the efficiency of repeatable processes, as the above examples demonstrate.
We are incredibly excited about AI agents’ vast potential to help organisations of any size modernise their business processes. By going beyond Power Automate capabilities, AI agents can automate additional manual admin processes by leveraging your organisation’s collective knowledge.
At this early stage, we recommend considering which processes within your business could be modernised before agents are generally available in 2025. If you’d like to explore potential use cases and understand how ServerSys can help you, please get in touch.