Microsoft AI Agents: How Your Organization Can Prepare

Microsoft AI agents are advanced tools that leverage generative AI to assist with a wide range of tasks. Unlike traditional artificial intelligence (AI) assistants, these agents can work alongside you or even autonomously on your behalf. They can handle almost everything from simple prompt-and-response interactions to complex, multistep workflows.

AI agents can be customized to have specific knowledge, making them capable of managing tasks like reconciling financial statements, processing customer returns, or providing detailed product information. Furthermore, AI agents can personalize interactions and provide faster, more efficient support, helping improve the customer and employee experience.

Preparing for AI Agents

As more organizations prepare to implement and use AI agents to help automate routine tasks and improve efficiency, we’ll highlight key considerations to help guide AI deployment at your organization.

According to a Forrester report, “The State of AI Agents,” AI agents are emerging as a top technology for 2025 and beyond, with diverse use cases ranging from consumer-facing applications to business operations. However, their development and maturity vary widely depending on the specific use case and industry trends.1

Approach AI agents as you would onboarding a new employee. Begin by defining the role of the new team member, then refine their skills and knowledge to fit their business role. As they grasp their duties, their skills can help improve internal and client-facing efficiencies. Similarly, AI agents build from knowledge, and decisions are needed to prepare for the new technology.

The preparation for AI agents begins with defining your agent’s role and clearly outlining what you want to achieve. This could include productivity, customer support, or assisting with select internal tasks to drive efficiency. As you define the role, you will want to consider three important resources for your agent:

User experience journey
Data knowledge sources
Employee preparation

User Experience Journey

To effectively adopt Microsoft AI agents, start by outlining a road map of your business goals and use cases. This journey can create opportunities to share knowledge and enhance services across your organization. An organization’s strategic goals for agents can be unique, depending on business needs, clients, processes, technology, and cybersecurity in the user experience.

Focus areas can include internal streamlining of redundant tasks, external knowledge sharing, and improvements in customer service, sales, and marketing. The possibilities for AI agent automation are vast and require careful planning on how to guide the conversation.

When developing use cases, consider how the AI agent can help enhance workforce efficiency, drive productivity, and streamline workflows within your organization. After defining the functionality goals, plan how the agent will be tested. Next steps involve setting timeline expectations. When planning your implementation timeline, a best practice is to prioritize AI agent adoption and refine development over time, considering data source availability and necessary transitions for your staff.

Data Knowledge Sources

Knowledge sources can enhance AI agents by connecting data from various platforms like Microsoft Power Platform, Dynamics 365, websites, and others. These sources can enable agents to provide relevant information and insights. Knowledge can be added during agent creation or later and is used in generative answers to respond to user queries effectively. This is further explained on the Microsoft Copilot Studio knowledge sources overview page.

Confirming Copilot users have access to controlled data can significantly help enhance the results’ quality. By sourcing data from reliable and focused channels, content delivery and governance can be easier to manage. Data mapping can be derived from various internal and external sources, providing flexibility and control. For example:

Source	Name	Description	Authentication
External	Public Website	Searches the query input on Bing, only returns results from provided websites	None
Internal	Documents	Searches documents uploaded to Dataverse; returns results from the document contents	None
Internal	SharePoint	Connects to a SharePoint URL, uses GraphSearch to return results	Agent user’s Microsoft Entra ID
Internal	Dataverse	Connects to the configured Dataverse environment and uses a retrieval-augmented generative technique in Dataverse to return results	Agent user’s Microsoft Entra ID
Internal	Enterprise Data Using Copilot Connectors	Connects to Copilot connectors where your organization data is indexed by Microsoft Search	Agent user’s Microsoft Entra ID

These integrations can power your Microsoft AI agent decision-making results. It’s important to note that if an agent has permissions to a website or document data, all users of the agent also have accessibility permission to the data. However, if the data requires Microsoft Entra ID authentication, only authorized users can access and interact with the data within their permissions. This focused data approach can help mitigate the risk of oversharing concerns and preclude Copilot from processing sensitive information.

Authoring and collecting knowledge materials may initially seem daunting; however, there are many resources you can use to get started. The table above illustrates that agent knowledge sources can be drawn from various locations. While defining AI agent knowledge, consider having experienced employees draft how-to articles to provide best practices, tips, and tricks. Challenge employees to document frequently asked questions (FAQs), which can help feed an AI agent’s knowledge so it can respond appropriately. Take proactive steps to clean up your SharePoint site or website, generate PDFs and spreadsheets, and consider other data sources to help fill and update data knowledge sources.

Lastly, consider different anomalies while defining and testing an agent’s chat style. Some AI chat experiences extract knowledge from multiple data sources, which may present challenges for the AI agent to respond appropriately. An open-style question bot, where interactions include various slang, dialects, and languages, needs to be considered due to the limited visibility of predictable behaviors. This method of automation does not allow for a controlled list of question possibilities. Therefore, your data should be reliable and well vetted. The saying “garbage in, garbage out” applies when working with agents. Visibility of these communications can be crucial to success while supporting the knowledge sources with consistent updates. Conversely, other agent methods allow you to control the communication in a tree structure, guiding users through predefined contexts.

Employee Preparation

Employee engagement is important to help support the onboarding of AI agents. Their responsibilities, processes, and tasks may change when deploying AI technology. To support these initiatives, training and clear communication can help ease the transition and support desired outcomes. Employees will need training on bot-to-human interaction and prompt engineering as this new method of interaction may present novel experiences and findings. Throughout the communications to your team, presenting potential benefits and challenges upfront can provide clarity on expectations. Potential challenges for preparation and adoption include:

Change management: Employees may resist changes to their workflow; however, allowing them to participate in the onboarding process can help prepare them for these new tools.
Initial productivity dip: There might be a temporary decrease in productivity as employees adapt to the technology.
Data governance and security: Data privacy and security is paramount, and organizations need robust controls to help govern agent access to data.
Managing agent access and permissions: IT and system administrators require tools to manage agent access and permissions to help oversee data governance and security.
Blocking or unblocking agents: Administrators also need tools to block or unblock agents for specific users or groups to control their availability and functionality.
Cost and training: The cost of training employees on AI agents can vary significantly. Simpler rule-based systems generally require less training compared to more complex machine learning-based agents. Factors such as training materials, workshops, and live sessions can influence costs, too.

Benefits of Microsoft AI Agents

Microsoft AI agents are integrating with Dynamics 365 applications to help enhance various processes—for finance, sales, service, marketing, and more. The agents can automate tasks, provide insights, and assist in making informed decisions, helping to improve customer and employee engagement and business efficiency. For example, Microsoft’s AI Accelerator for Sales program aims to transform sales processes using Microsoft 365 Copilot and AI agents, including prebuilt agents like the Sales Research Agent and custom agents tailored to specific sales processes.2

Adoption Strategies

Adoption of AI agents can be facilitated by utilizing strategic approaches:

In-depth training programs: Offer ongoing training and support.
Clear communication: Keep employees informed about potential benefits and changes.
Involve employees early: Engage employees in the adoption process to get their buy-in.
Monitor and adjust: Consistently monitor the adoption process and make necessary adjustments.

Managing Copilot Agents in Integrated Apps

The Microsoft 365 guide on managing Copilot agents in integrated apps explains how administrators can oversee the deployment and management of Copilot agents. Agents can enhance Microsoft 365 Copilot by adding custom actions, search capabilities, and integrations with third-party apps like Jira and Dynamics 365. Administrators can enable, disable, assign, block, or remove agents to align with organizational needs and data privacy standards.3

Organizational guidelines can outline acceptable ethics in communications learned and shared within chat conversations. Training employees on managing privacy and defining responsibilities for the data collected from chat conversations can affect employee processes while adopting AI agents. Like the use-case road map, preparing your employees also involves making many decisions throughout the AI adoption process.

How Forvis Mazars Can Help

The results of adding AI agents to your workflows can be far-ranging and inspiring. The productivity gains powered by AI agents, through streamlined tasks and repeated small wins, can scale to significantly enhance efficiency and drive innovation within your organization.

In summary, Microsoft AI agents represent a significant evolution in workflow technology, offering capabilities from automating tasks to providing data-driven insights. Their integration with platforms like Dynamics 365 highlights the potential to enhance various business functions. Preparing for these agents involves defining roles, considering user experience, managing data knowledge sources, and employee preparation. By applying effective adoption strategies and considering management and ethical factors, organizations can effectively leverage AI agents.

As a top 1% Microsoft Business Applications Partner, the Microsoft team at Forvis Mazars can help you prepare for and implement AI agents. Connect with us today to start your journey with agents.

Related FORsights