Every unanswered request chips away at the trust a customer has placed in a brand. That much has not changed. What has changed is that the delay no longer has to happen.
Dynamics 365 Customer Service, Microsoft’s enterprise customer service software and contact centre solution, now ships with an infrastructure that reads an incoming request before it reaches a service rep, pulls the relevant history, prepares a resolution suggestion, and evaluates its own quality. The four AI agents that make this possible reached general availability in October 2025. This is not a product announcement. It is a capability already running in production environments and changing how service teams operate.
From assistant to participant
AI entering customer service is nothing new. What is different this time is that it has moved from assisting to actually owning specific responsibilities.
Dynamics 365 Customer Service has more than twenty years of history behind it. Early versions gave you a form on screen: customer name, issue, priority. The rest was left to the human. Business rules, automatic assignments, and SLA tracking were layered on over the years, but the underlying logic did not shift: a person entered the data, a person found the resolution, a person closed the case.
The 2024–2026 period is where that logic breaks. Four agents now step into the process. Each one operates within defined boundaries and can run fully autonomously or require a rep’s confirmation before acting. The organisation decides which steps to hand over and which to keep under human control.
What each agent does
Cases open on their own: Case Management Agent
In the traditional flow, a case began with whatever the customer typed into a portal, or whatever the rep pieced together from an email or a call. More often than not, the record arrived incomplete.
The Case Management Agent intercepts this. When an email, a chat exchange, or a portal message triggers an interaction, the agent analyses the conversation and creates the case automatically. Subject, description, priority, and linked product are filled in by AI. The rep sees an enriched record from the start, with no need to look up the customer number, scroll through previous threads, or reconstruct the context themselves.
The agent also updates open cases based on rule sets the organisation configures: close automatically after a set period, re-prioritise when a score threshold is crossed, initiate a follow-up with the customer. Three activities and the case can cap itself, or it can wait for the rep’s sign-off. All of it is configurable from the admin panel in a handful of steps.
We use this module in Pargesoft’s own operations. Every request, complaint, or question that comes in from a customer is turned into a case the moment it enters the system and tracked from that point forward.
Reading what the customer actually wants: Customer Intent Agent
A case opens, but the real question is still unanswered: what is this customer actually looking for?
The same words can carry very different intentions. “My device isn’t working” might mean a request for technical support, a return, or simply a link to the manual. Getting that distinction right from the start has a direct effect on resolution time and on how productively the rep’s time is spent.
The Customer Intent Agent scans the full interaction history going back to before the sale, including past cases, correspondence, and purchase data, and classifies the customer’s intent. A CRM support query and an installation request land in different intent groups; each group triggers different content and a different workflow.
The agent learns rather than matching against rules. Every new interaction expands its intent library. Managers do not need to define a rule for every new scenario. The system updates itself from customer behaviour. Through the “Ask a question” tab in the Copilot help pane, the agent presents the rep with a summary of what the customer is likely after. Whether to use it is entirely up to the rep.
A knowledge base that never goes stale: Customer Knowledge Management Agent
This is one of the oldest problems in any contact centre: a case gets resolved, something genuinely useful is learned, and then it is never written down. The rep closes the ticket, the knowledge stays in their head, and the knowledge base continues aging.
The Customer Knowledge Management Agent breaks that cycle. As soon as a case closes, the agent analyses the case details, conversations, emails, and notes; identifies the gap in the knowledge base; and either drafts a new article or flags an existing one for an update. The rep reviews and approves. Before making any suggestion, the agent also checks whether a similar article already exists, preventing duplicate content from accumulating.
The agent supports over fifty languages, including Turkish, which removes a meaningful constraint for deployments in Turkey. There is also a preview feature worth noting: automatic updates to existing knowledge base articles based on the resolution patterns of subsequent cases. Full availability is expected before the end of 2026.
Operations that measure themselves: Quality Evaluation Agent
While the first three agents are handling the work, the fourth steps back and evaluates it.
The Quality Evaluation Agent assesses how both human reps and AI agents have handled customer interactions. Supervisors define the evaluation criteria and weightings; the agent scores every interaction against that framework. How many cases closed automatically, how many required a human hand, which channels are producing cases that stay open longest, whether things are improving or sliding. All of it is visible from a single interface without building a separate report.
In practice, this means the operation can measure itself. You do not need to wait for a monthly review to understand whether the system is working.
How Gartner sees this
Microsoft was named a Leader in the 2025 Gartner Magic Quadrant for the CRM Customer Engagement Center, published in October 2025. The report highlighted AI-powered automation, omnichannel integration, and the extensible architecture built on Power Platform as standout strengths.
This is not just a badge. For organisations comparing customer service technology options, independent analyst placements carry real weight in the decision process. Microsoft also held its Leader position in the Forrester Wave Customer Service Solutions Q1 2026 report published the same quarter.
Why having everything in one place matters
How well an AI agent performs is largely determined by how much data it can draw on.
If a customer’s purchase history, the products they own, previous service interactions, and open sales conversations all live on the same platform, the agent already knows the context when a new case arrives. It does not have to reconstruct any of it.
This is where Dynamics 365 Customer Service has a structural advantage. When Dynamics 365 Sales, Customer Insights, and service data sit in separate systems, even the best agent is working with an incomplete picture. When they share the same platform, the agent’s suggestions get sharper and its errors get fewer.
A concrete example: if a customer bought a washing machine and raises a service request, the agent should be looking at cases related to that product, not the customer’s refrigerator history. Knowing which product was sold, and when, is what allows the agent to make that distinction by itself.
Beyond the four agents: Copilot Studio
The four out-of-the-box agents cover a wide range of scenarios without any additional build work. But some industries and some organisations have processes that fall outside the standard.
Microsoft Copilot Studio is the development environment these agents are built on, and it gives organisations the tools to design their own agents from scratch. A custom classification logic, a connection to an industry-specific knowledge source, or an integration with an existing internal system can all be built in Copilot Studio and brought into Dynamics 365 Customer Service.
One technical point worth noting: all of these capabilities run in the cloud only. Deploying these agents in an on-premise environment is not supported.
Deployment is a strategic decision, not just a technical one
The setup itself is relatively accessible. The relevant tab in Admin Center, a few permission and licence assignments, the rule sets, and then publish via Copilot Studio. It can be done without heavy technical overhead.
What takes more thought is preparation. AI agents perform best in environments where the data is clean, the case categories are consistent, and the knowledge base is current. Badly structured queues, outdated articles, and inconsistent customer segmentation all limit what the agents can do. At Pargesoft, we treat this readiness work as part of the implementation itself, not a separate step. If you want to understand where your organisation stands before going further, our AI Readiness Score tool is a good starting point.
Frequently Asked Questions
Do the Dynamics 365 Customer Service AI agents work in an on-premise environment?
No. The Case Management Agent, Customer Intent Agent, Customer Knowledge Management Agent, and Quality Evaluation Agent are only available in the cloud version of Dynamics 365. Deployment in on-premise or hybrid environments is not supported.
Which channels does the Case Management Agent work with?
It works across all interaction channels supported by Dynamics 365 Customer Service, including email, live chat, portal messages, and social channel activities. You can also define custom activity types and bring them into the same flow.
Which Dynamics 365 licence do you need to use these agents?
Customer Service Professional covers core case management but does not include the AI agents. To access all four agents, you need Customer Service Enterprise or Dynamics 365 Contact Center. Some advanced Copilot features are billed on a consumption basis through Copilot Credits rather than a fixed seat fee.
Do the agents run fully automatically, or does a human still need to approve things?
Each agent can be configured to run fully autonomously, in a semi-autonomous mode where the rep confirms before the agent acts, or purely as a suggestion engine. The organisation decides which steps to hand over and which to keep under human control. For example, case creation can be fully automated while resolution suggestions are always presented to the rep for approval.
Can the Customer Knowledge Management Agent produce content in languages other than English?
Yes. The agent supports over fifty languages. For deployments in Turkey, it can generate and update knowledge base articles in Turkish without any additional configuration.
Licensing
A question that comes up often: which licence do you actually need?
Customer Service Professional covers the core case management capabilities but does not include the AI agents. To access all four agents, you need Customer Service Enterprise or Dynamics 365 Contact Center. Some advanced Copilot features are priced on a consumption basis via Copilot Credits rather than a fixed per-seat fee. We can help you map the right structure against your actual use case. Get in touch if you would like to work through it.
Full webinar recording
We covered all of this live, with a working demo, in our webinar presented by İlker Karaoglu, our CRM and Power Platform Solutions Director. If you missed it or want to watch it again, the full recording is below. For other upcoming sessions, visit our Events & Webinars page.
To discuss how AI agents could work in your customer service operation, see a live demo, or plan an implementation, get in touch with us.