Customer processes gained more acting agents

Microsoft’s wave 1 launch on 30 April described Dynamics 365 Sales agents that could research and prioritise leads, initiate conversations and follow up sellers. Customer Service and Contact Center gained proactive conversational journeys and three agents intended to work together on contact-centre tasks. The release extended over six months, so availability needed checking for each capability and environment.

For marketing and sales, this enabled a faster transition from signal to relevant contact. In service, agents could use case and knowledge context before routing work. The Field Service plan also described Copilot-first experiences across web, mobile and Microsoft 365. An enquiry could move from digital dialogue to a work order only when identity, consent, product and agreement travelled with it.

Human work shifted towards choices and exceptions

At its sales conference on 20 May, Gartner argued that sellers should develop skills where people add value instead of competing with AI on tasks the technology handles well. This offered a useful boundary: an agent might collect information and propose an activity, while the seller assessed the customer situation, the promise being made and the commercial priority.

The next day, Gartner recommended small agent experiments where data was available, behaviour verifiable and the expected return high. Although its analysis concerned supply chains, the test principle applied to customer processes. An agent creating a case or work order could be checked for correct fields, duplicates, poor priority and manual correction.

The process owner also needed to verify licensing and availability before changing responsibilities. A demonstrated agent could be in preview, require separate capacity or lack support in the region. The test protocol should record version, environment and date with its results. The team could then repeat the test after an update without confusing a product change with a better process.

The result belonged at the handover

A suitable experiment could follow one lead or case type from the first signal to agreed follow-up. The team could measure time to a relevant response, complete handovers, reopened cases and whether customers repeated information. A manual sample would expose errors hidden by volume measures. Only when both flow and quality improved was there evidence for expanding the agent’s mandate.

Sources

Microsoft Dynamics 365 Blog: 2025 release wave 1 brings hundreds of updates, 30 April 2025.

Microsoft Learn: Dynamics 365 Field Service 2025 release wave 1, April 2025.

Gartner: CSO & Sales Leader Conference: Day 1 highlights, 20 May 2025.

Gartner: Agentic AI in supply chain management solutions, 21 May 2025.

For discussion

Which customer handover can be tested end to end with a verifiable action and measurable outcome?