Microsoft continues to push Copilot from a conversational assistant toward an enterprise agent platform. One of the most interesting developments is Custom Skills for Declarative Agents, giving organizations another way to extend agents with capabilities designed around their own business processes.
What changes with Custom Skills?
Declarative agents can be configured for specific roles, knowledge and scenarios. Custom Skills take that idea further by allowing an agent to connect its reasoning and conversation experience with purpose-built capabilities.
Instead of an agent only answering a question, the enterprise vision becomes much more powerful: understand the user's intent, apply organizational context, invoke an appropriate skill, work with business systems and return an actionable result.
Why this matters for Dynamics 365 & Power Platform
For organizations already invested in Microsoft Dynamics 365, Power Platform and Dataverse, agent skills create interesting opportunities to place AI closer to operational processes.
Consider scenarios such as project delivery, sales, customer service, finance, supply chain or internal operations. An agent could potentially help interpret a request, retrieve the right business context, trigger a governed capability and guide the user through the next action.
Imagine these enterprise scenarios
- Project Management: Ask an agent to summarize project risks, identify overdue actions and prepare a management update.
- Dynamics 365: Let an agent retrieve contextual business information and guide users through complex operational processes.
- Delivery Governance: Combine project data, actions, RAID information and status signals to improve leadership visibility.
- Customer Operations: Give service-oriented agents specialized skills for resolving requests while maintaining enterprise controls.
- Knowledge + Action: Move beyond searching organizational information by connecting knowledge to an executable business capability.
The architecture still matters
Custom Skills do not remove the need for good enterprise architecture. In fact, agents that can take meaningful actions make governance even more important.
Identity, authorization, data boundaries, environment strategy, Dataverse design, API management, monitoring, auditability and ALM all need to be considered. AI may make the user experience simpler, but the enterprise controls behind that experience must remain intentional.
What should delivery leaders be thinking about?
The important question is no longer only, “Where can we deploy Copilot?” A stronger question is:
That changes the conversation from experimentation to operating model. Organizations will need to identify high-value processes, define agent boundaries, establish governance, measure adoption and continuously evaluate business value.
My perspective
The future of enterprise applications will increasingly combine systems of record, systems of automation and systems of intelligence. Dynamics 365 and Power Platform already provide much of the business-data and workflow foundation. Agent capabilities add a new interaction layer on top.
The winners will not necessarily be the organizations that create the largest number of agents. They will be the organizations that design the right agents, connect them to the right capabilities and govern them with the same discipline expected from every other enterprise system.