Building AI agents with Microsoft Copilot Studio

Microsoft Copilot Studio is part of Microsoft's comprehensive AI offering, which supports companies in developing and implementing customised AI solutions. The offering is primarily aimed at companies that want to use AI at various levels to increase their efficiency and optimise business processes. The following is an overview of Microsoft's entire Copilot portfolio:

‍

1. Copilot (with commercial data protection)

free of charge

This version is freely accessible and offers AI-based support for general search queries via the browser. However, this version is not included in Microsoft's Data Processing Agreement (DPA). Therefore, it cannot be assumed that it is GDPR-compliant when connecting internal data, especially not for sensitive data.

• Area of application: General information requests, suitable for all users.
• Availability: Via the Edge browser or at https://copilot.microsoft.com/.
  ‍

2. Copilot for MS365

€28.10 per user/month (as of Oct. 2024)

This version goes one step further and integrates the AI functions directly into the Microsoft 365 applications. The AI can access company or user data to provide advanced search options and personalised recommendations. This solution is particularly interesting for employees who regularly work with the Office applications.

• Area of application: Suitable for selected employees who want to complete specific tasks more efficiently.
• Availability: Integration with numerous MS365 applications such as Word, Excel, PowerPoint, Teams, OneNote, Outlook and more.

‍

3. Copilots via Copilot Studio

€187.20 per month for 25,000 messages (as of Oct. 2024)

Copilot Studio is a specialised platform for creating individual copilots, which can be enriched with company-specific data, connected to internal systems and can support the automation of processes through actions. Once created, (custom) copilots can be made available throughout the company via various channels, such as Microsoft Teams, or made accessible to external persons via a website.

• Area of application: Only for administrators and developers who want to create customised AI solutions for the company.‍
• Availability: Creation exclusively in Copilot Studio, but the created Copilots can be rolled out company-wide.

‍

With these different solutions, Microsoft covers the entire spectrum of companies' needs - from general AI applications to specific, customised solutions for complex requirements. For German SMEs in particular, Copilot Studio offers the potential to introduce AI-supported processes in a targeted manner and remain competitive in the long term.

AI agents are specialised software programs that use artificial intelligence to perform tasks independently, make decisions and interact with their environment They process data, make rational decisions and autonomously pursue specific goals that are set for them by humans They use sensor data or other inputs and can independently search for and analyse information and act on the basis of this data Examples range from simple chatbots that process customer enquiries to workflow automation and complex autonomous systems such as LLM Agent Operating Systems (AIOS).

The key features of AI agents include:

1. Autonomy: They act independently and perform tasks without the need for constant human intervention
2. Learning: By using machine learning, AI agents continuously improve their performance based on new data and experience
3. Goal-orientation: AI agents pursue clearly defined goals and can dynamically adapt their actions to achieve them as efficiently as possible

‍

Competitive advantages through AI agents

‍

1. Automation and efficiency

AI agents take over repetitive tasks and relieve employees, which increases productivity and reduces costs.

‍

2. Error minimisation‍‍

Automation reduces human error, which is particularly important in critical areas such as accounting or customer service.

‍

3. Faster decisions

Real-time analyses enable well-founded, quick decisions and proactive action in areas such as marketing or supply chain management.

‍

4. Scalability

AI agents adapt to growing business requirements and enable efficient processes even with increasing volumes.

‍

5. Personalisation‍

‍Customer data is analysed to provide tailored recommendations, increasing customer satisfaction and loyalty.

‍

6. Process optimisation‍‍

AI agents identify inefficient processes and offer solutions for optimisation, e.g. through predictive maintenance in production.

‍

AI agents offer SMEs the opportunity to organise their business processes more efficiently, focus on value-adding activities and remain competitive through the use of modern technologies. With platforms such as Microsoft Copilot Studio, companies can develop and implement these intelligent solutions in a customised manner in order to achieve specific business goals and secure long-term competitive advantages.

The possible applications of AI agents created with Microsoft Copilot Studio are diverse and can optimise different business areas. Here are three specific use cases that show how AI agents can help solve specific challenges:

‍

1. Qualification of sales leads

Microsoft Copilot Studio enables the creation of agents that support sales teams in prioritising leads and automating the contact process. The Sales Qualification Agent, for example, independently researches potential customers and helps the sales team to focus on high-priority leads and personalise outreach, increasing efficiency and improving conversion rates.

‍

2. Optimisation of customer service

Companies facing high call volumes and staff shortages benefit from Copilot Customer Intent and Knowledge Management Agents. These agents independently handle common customer queries and update knowledge bases, allowing customer service teams to provide consistent, high-quality service and quickly onboard new agents by extending best practices to the entire team.

‍

3. Supply chain management‍‍

For operational teams, the Supplier Communications Agent optimises the supply chain by independently monitoring supplier performance and resolving delays. This proactive approach minimises costly disruptions and allows procurement teams to focus on strategic tasks rather than manual tracking.

‍

Microsoft plans to offer some of these agents as turnkey solutions in the future to enable organisations to get started quickly and easily with the automation of key business processes. For precisely customised agents for individual processes, it is nevertheless worth experimenting with Copilot Studio and exploring your own solutions.

Before you develop your copilot with Microsoft Copilot Studio, clear planning is essential. Define the goals and tasks of the copilot as well as the benefits for your company and your target group. Gather relevant data sources, both internal and external, and ensure that the use of this data complies with data protection guidelines and the GDPR. To do this, clarify the permitted use of data with your IT department and data protection officer at an early stage.

To build the AI agents with Microsoft Copilot Studio, follow these steps based on the information from the quick start guide:

‍

1. Registration and getting started

• Log in to Microsoft Copilot Studio or create an account. After logging in, you will be taken to the Copilot Studio start page.

2. Create a new Copilot

• Select the ‘Create “ option in the left-hand navigation area and then ”New Copilot’.
• Describe in the chat what you want the copilot to do. Specify which tasks it will perform and the style (e.g. formal, colloquial) in which it should communicate with users.

3. Add name, description and icon

• Give your co-pilot a name and add an icon (in PNG format).
• Write a short description and specific instructions for the co-pilot. For example: ‘Friendly Tutor Copilot’ could act as a friendly tutor who explains how Microsoft Copilot Studio works.

4. Add knowledge

• You can add knowledge sources to Copilot. These can be public websites or company-specific data sources. To do this, click on "Add knowledge ’ and enter the relevant URL or other data that you want Copilot to use.

5. Add actions

• Open the ‘Actions’ area and click on ‘+ Add action’. Select an action, e.g. a Power Automate flow or a Bot Framework skill. Follow the instructions in the wizard to configure the action and establish connections if necessary. Save the action to add it to the copilot and test its function to ensure that the responses are generated as desired.

‍

‍

6. Test and customise

• Test your co-pilot by asking questions and checking how they respond. Experiment with changes to the instructions or tone of voice to see how the interaction changes.
• You can always add new instructions or knowledge sources and test the reactions in the chat to further optimise the co-pilot.

7. Publish

• When the copilot is ready, you can publish it on a demo website. Click on ‘Publish’ to make it available. You will then receive a URL that you can share with others to test the Copilot.

Once published, integrate the Copilot into your business processes and tools. Train your employees to use it and provide supporting materials. Monitor performance regularly and continuously optimise the Copilot based on user feedback.

‍