Power Automate and Copilot Studio: Build Smarter AI Agents
From starting out answering predefined questions to intelligent assistants that understand natural language, access business data, and complete meaningful work, AI-powered agents have come a long way.
As far as they’ve come, there is a way to take them even higher. Combining Power Automate and Copilot Studio lets you create agents that take actions on your behalf, going above and beyond simply answering questions.
Integrating smart automation into AI-powered agents allows you to transform your conversational experience into practical business tools that retrieve information, create records, initiate workflows, and even make policy-based recommendations.
In this blog and video, we’ll show you what this partnership between Copilot Studio and Power Automate looks like and explore why this combination creates a new generation of business "helpers."
Evolution from Bots to AI Agents to "Helpers"
To understand why Power Automate is such a beneficial addition to Copilot Studio, it helps to look at how Microsoft's AI capabilities have evolved over the years:
Stage One: Traditional Bots
The first generation of business bots was designed around predefined conversations. They answered specific questions, guided you through structured options, and completed simple, repetitive tasks.
A common example is an online government or customer service chatbot that helps users:
- Make a payment
- Find a service
- Answer common questions
- Navigate a website
These bots work well within the scenarios they've been programmed for, but they typically can't move beyond those predefined conversations.
Stage Two: AI Agents
The introduction of generative AI fundamentally changed what agents can do. Rather than responding only to predefined prompts, AI agents can interpret natural language and carry on much more dynamic conversations. You can ask questions conversationally instead of navigating rigid decision trees.
However, even with that flexibility, there was a key limitation. These agents could tell you what to do but couldn’t do it for you.
Stage Three: The "Helper"
Thinking of AI agents as “helpers” is where we are today. “Helper” agents combine the conversational intelligence of an AI agent with the execution capabilities of a tool like Power Automate.
Instead of simply providing information, the helper performs actions in the background.
For example, rather than recommending a recipe, a helper could:
- Email the recipe
- Send it to a family member
- Complete follow-up tasks automatically
When these two tools work together, Power Automate transforms an AI conversation into an automated business process that drives value and saves time for your organization.
Agents in Action: What You Can Do with Power Automate and Copilot Studio
Now that you have the background, let’s dive a little bit deeper into what exactly you can do when you combine Copilot Studio and Power Automate.
Demo: Inventory Agent
Let’s introduce you to Order Smith, an AI agent designed for warehouse operations. Although the scenario focuses on inventory management, the concepts can easily be applied to virtually any department.
Asking Questions Instead of Clicking Through Applications
Let’s begin inside a Power Apps model-driven application. Rather than navigating through multiple tables and records, you can interact directly with the Copilot agent using natural language.
The first question is simple: "How many Windex do we have on hand?"
Instead of manually searching inventory records, the agent accesses inventory information in the background and immediately reports that three units are currently in stock. The value goes beyond simply getting a faster answer; it’s about avoiding unnecessary navigation through the application.
Looking Beyond Internal Data
Next, let’s shift from internal business data to external information. After viewing the current purchase price for Windex, the user asks the agent to search the web for pricing on the original product. Because web access has been enabled, the agent retrieves current pricing information and presents it back to the user.
This allows you to quickly compare existing purchase costs with external pricing before placing an order. Rather than opening a browser, searching multiple websites, and returning to the ERP system, the agent performs the research within the conversation.
Creating an Order Through Conversation
Once the pricing has been confirmed, the final step is placing the order. Instead of navigating to another screen and entering information manually, you can simply ask Order Smith to order additional Windex.
The helper processes the request and confirms that the order has been created.
In just a few conversational prompts, you have:
- Checked inventory
- Reviewed current pricing
- Compared external prices
- Created a new order
The kicker is you can do all of that without leaving the chat and switching between windows and applications.
Multi-Item Order Creation Using AI and Automation
Ordering one product is useful, but ordering multiple products from multiple vendors through a single request is where the automation becomes especially impressive.
Accepting Natural Language
Instead of asking for products individually, the user provides a list of products in one prompt. The helper accepts the entire request and begins processing it automatically.
From your perspective, the process is incredibly simple.
Behind the scenes, however, Power Automate is doing the heavy lifting. Let’s look at how we built the solution inside of Copilot Studio with Power Automate:
Step 1: Capture the User's Request
The Copilot agent receives your natural language request inside the model-driven application. A custom topic created in Copilot Studio sends that text directly into a Power Automate flow.
Step 2: Transform the Text
Inside Power Automate, the raw text is passed into an AI Hub. Rather than treating the request as plain text, the AI transforms it into structured purchase order information.
The result becomes individual purchase order product lines containing the necessary values extracted from the user's request. In our example, the user wanted to transform the purchase order into a JSON file.
Step 3: Process Each Product
Once the product lines have been generated, Power Automate loops through each one individually. For every product identified, the flow automatically creates the appropriate purchase order line.
Instead of building separate automation for each possible product, the AI interprets the request while Power Automate handles the repetitive processing.
This combination allows a single conversational request to generate multiple purchase order lines automatically.
Purchase Order Approval Agent
The second AI helper, Gina, focuses on purchase order approvals. Rather than helping create orders, Gina evaluates whether purchase orders comply with company policy.
Reviewing Purchase Orders
Once purchase orders have been created, Gina reviews each one. The agent can be started manually or triggered automatically whenever a purchase order status changes.
This makes the review process flexible enough to support different business workflows.
Using Company Policies as Knowledge
One of Gina's most important capabilities is having access to your organizational policies. Those policies define exactly what should be approved automatically and what should be escalated for manager review. In our example, we will use the following approval rules:
- Quantities greater than 10
- Purchase orders exceeding $1,000
If either condition is met, Gina recommends rejection so the order can be reviewed by a manager before going out.
Recording the Decision
Once Gina evaluates the purchase order, Power Automate takes over once again:
- Calls Gina for an evaluation.
- Receives her response.
- Passes that response through another AI Hub.
- Determines whether the purchase order was accepted or rejected.
- Creates a note explaining the decision.
- Updates the purchase order status.
When we refresh the model-driven application, purchase orders display their updated status. Additionally, opening a rejected purchase order reveals the note explaining exactly why it failed approval. This allows you to go in and resolve issues quickly.
You can also customize explanations to meet your specific and unique organizational requirements.
Key Benefits of Combining AI Agents and Power Automate
Throughout all the key enhancements made to Copilot agents, Power Automate enables them to move beyond conversation and into execution. Agents will now perform meaningful work to reduce manual effort.
This combination will deliver several benefits:
- Save Your Employees Time: Routine activities no longer require your employees to navigate multiple tables or complete repetitive administrative work.
- Reduce Manual Navigation: Instead of clicking through applications to locate information, your team can retrieve what they need through natural conversation.
- Improve Decision-Making: Employees receive the information they need more quickly, allowing them to make informed decisions faster.
- Increase Your Employee Satisfaction: Removing repetitive administrative work allows your employees to spend more time on higher-value, strategic activities rather than routine tasks.
- Deliver Better Customer Service: When employees have faster access to information and automated support behind the scenes, they're better equipped to serve customers efficiently.
Ultimately, the goal isn't to replace employees; it's to give them intelligent helpers that handle repetitive work while allowing people to focus on the tasks that require expertise, judgment, and strategic thinking.
Ready to Turn Your AI Agents into Intelligent Helpers?
Building an AI agent is only the beginning. By integrating Power Automate with Copilot Studio, you can create intelligent helpers that streamline processes, automate repetitive tasks, and help your teams work more efficiently.
At Stoneridge Software, we've helped organizations get more value from Microsoft's business applications by designing practical, scalable solutions that align with real business processes. Get in touch with our team today if you are ready to take your Copilot Studio agents to the next level with Power Automate!
Under the terms of this license, you are authorized to share and redistribute the content across various mediums, subject to adherence to the specified conditions: you must provide proper attribution to Stoneridge as the original creator in a manner that does not imply their endorsement of your use, the material is to be utilized solely for non-commercial purposes, and alterations, modifications, or derivative works based on the original material are strictly prohibited.
Responsibility rests with the licensee to ensure that their use of the material does not violate any other rights.