AI models are changing faster than most enterprise service roadmaps can keep up.
New models arrive with stronger reasoning, lower latency, broader language support, or more capable agentic behavior. Yet contact center leaders cannot rebuild their customer service architecture every time the AI market moves.
That leaves a more durable question for enterprises: what sits around the model?
For Microsoft, the answer is a connected environment where customer service, CRM, collaboration, and business operations work from a shared understanding of the customer and the work still required. The company’s argument is not that every organization needs to replace its contact center. It is that AI will struggle to deliver reliable outcomes if customer context, business tools, and governance remain separated. In a recent CX Today interview, Darya Mazandarany, Corporate Vice President of Dynamics 365 Contact Center at Microsoft, argued that the durable part of the architecture is not the model itself.
“I think what’s durable is the harness that the models are running in and the ability to give them the right tools and the right intelligence context. Models are going to come and go, but what’s fueling them is the key.”
A capable model can summarize a conversation or generate a response. It cannot reliably resolve a customer’s problem unless it can identify the customer, understand the current case, access the relevant policy, recognize previous decisions, and act through the systems where work gets done.
The “better together” case for contact center, CRM, and unified communications is therefore not a product-bundling argument. It is an operating model argument.
A service representative may be the person speaking to the customer, but they are rarely the person with all the information needed to resolve the issue.
A retention offer might need approval from a sales leader. A delayed delivery may require input from a supply chain team. A disputed charge may depend on a finance policy or a previous agreement reached in an account meeting.
In many businesses, the decision sits in an email thread, a Teams conversation, a shared document, or a meeting that the contact center cannot see. Representatives then have to search for information, ask colleagues for help, or tell the customer they will call back.
That process creates effort for everyone involved. It also limits what AI can do. Mazandarany used a commercial decision to illustrate the point:
“Should we give the customer the extra 20% discount on this deal? That doesn’t happen in the CRM or in the contact center. That’s happening in meetings, emails, documents and Teams chats.”
Microsoft’s proposition is that its contact center, Dynamics 365 applications, Microsoft 365 collaboration environment, and Power Platform can bring relevant business context closer to the point of service.
For an enterprise already using Teams, Outlook, SharePoint, and Dynamics 365, that may reduce the number of disconnected places representatives need to search. It could also help an AI agent ground an answer in more than a narrow contact center knowledge base.
The opportunity is real, but the architecture needs boundaries. An organization must decide which collaboration data is relevant to a customer interaction, who can access it, how long it is retained, and what an AI agent is allowed to do with it.
Service Intelligence Has Value Only When It Changes Something
Contact centers have collected interaction data for decades. Calls are recorded, chats are logged, cases are categorized, and customer feedback is measured.
The problem is not a shortage of insight. It is the gap between insight and action.
A transcript might reveal that customers are confused by a policy. A dashboard might identify a rise in repeat contacts. A quality review may show that representatives are struggling to find a particular answer. None of those findings improves the service experience until someone changes the policy, updates the knowledge, adjusts a workflow, coaches a team, or fixes the process creating the demand.
Microsoft is positioning its wider platform around that conversion from signal to action.
Mazandarany described the aim as improving customer outcomes across the whole journey, rather than simply optimizing a single interaction. That could mean using customer intent from a website session to give a representative better context when the customer subsequently calls. It could mean identifying a recurring operational issue and routing the evidence to the team responsible for fixing it.
The potential advantage comes from linking the contact center to systems where those actions can happen. A case can be updated. A workflow can be triggered. A field service request can be created. A customer record can be enriched. A knowledge article can be flagged for review. A supervisor can coach a representative using evidence from real interactions.
That is a broader ambition than conversational AI. It positions service as a source of operational intelligence for sales, marketing, commerce, and operations.
The contact center has often been treated as a destination for customer problems. The better-together model recasts it as a sensing layer for the wider business.
A customer interaction can reveal that a product description is unclear, a delivery process is failing, a billing policy is generating avoidable effort, or an account is at risk. Sales and marketing teams may benefit from that signal. Operations teams may need to act on it. Service leaders may need to adapt the knowledge and guidance available to representatives.
Each function has its own incentives, systems, and definitions of success. Service teams may focus on resolution and effort. Sales leaders may prioritize account growth. Operations may measure cost, throughput, and compliance. Creating shared context does not automatically create shared accountability.
CX leaders should therefore define the actions they want service intelligence to trigger before they invest in more data or AI. Which signals should create a sales task? Which should launch an operations investigation? Which require a knowledge change? Which can an AI agent act on without human approval?
Those decisions determine whether convergence produces a more responsive organization or merely a larger pool of data.
Governance Is What Makes Shared Context Usable
The case for connecting CC, CRM, and UC can sound straightforward until it reaches governance.
AI agents need clear controls over what data grounds a response, which tools they can access, when a human must approve an action, and how the organization can investigate an error. The more business context an agent can use, the more important those controls become.
Microsoft points to Dataverse, Copilot Studio, Power Platform, and its wider governance capabilities as the foundation for this controlled model. The company’s value proposition rests on keeping customer context and operational action inside a governed enterprise environment, rather than moving information between a growing collection of isolated point tools.
Buyers should test that proposition against their own operating reality. They should ask what context can be accessed in a live interaction, how permissions apply across CC, CRM, and UC, what evidence is captured when an agent acts, and who owns the remediation process when it gets something wrong.