AI enters the customer process

From 1 September, Microsoft combined its customer data platform and real-time marketing in Dynamics 365 Customer Insights. Later that month, it described Copilot for campaign content and segments, sales preparation, and service case summaries. Some capabilities were generally available; others remained in preview or were scheduled for September and October. That distinction matters when an organisation plans adoption.

The news therefore concerned more than a writing assistant. It placed generative AI in the transition between customer data and the next action. A segment can become the starting point for a campaign, an opportunity can be followed up from Outlook or Teams, and a completed conversation can give a service agent a shorter case summary. When several stages use the same information, errors in consent, customer status, or ownership also spread faster. Marketing and sales leaders consequently need shared rules for permitted data, approval of customer-facing copy, and review of machine-generated recommendations before delivery.

Early service figures need cautious interpretation

Microsoft reported an internal evaluation involving 11,500 support agents. In one limited area handling lower-severity chats, active handling time fell by 12 percent. In one support unit, 10 percent of cases that normally required peer assistance were resolved without it. These figures concern specific workflows at the vendor itself and are early signals, not a general forecast of impact.

That same month, Forrester identified content creation for customer support channels as the most selected generative AI use case in a survey of AI decision-makers. The analyst also stressed the need to synthesise large volumes of customer information before the tool can provide useful support. A service manager should therefore set pilot measures close to the task: time needed to understand a case, proportion of accurate summaries, need for peer assistance, and quality of the answer the customer receives. Customer satisfaction and reopened cases should be monitored alongside time saved.

Field service connects with office work

Later in the month, Microsoft presented closer connections between Dynamics 365 Field Service and Outlook, Teams, and Viva Connections. Work orders could be created and managed in Outlook, while Teams provided access within the collaboration surface. This can reduce manual handovers between the contact centre, dispatcher, and technician. It also makes the work order the central information carrier: a wrong address, a missing parts requirement, or outdated asset history follows the assignment into the field.

Microsoft highlighted a UK energy supplier where more than 20 legacy applications moved to a solution for over 1,500 technicians and 600 other users. Its expected cost reduction was a vendor- and partner-presented forecast, not evidence of what others should expect. The story still illustrates what management needs to observe: fewer duplicate entries, better preparation for the first visit, complete work-order data, and actual adoption among technicians.

Management sits between data and action

Automation should be bounded according to decision risk. A summary reviewed by an employee before use can tolerate a different error rate from customer communication sent automatically. A recommended next activity needs to be traceable to current customer status, while routing and resource suggestions must be checked against capacity and skills. Analytics then becomes more than a reporting layer: it should reveal how often suggestions were accepted, corrected, or rejected, and whether the customer outcome improved.

The September announcements gave organisations several concrete places to test generative AI and automation. A sound starting point is one connected customer journey with a named process owner, a limited data set, and measures that can be compared before and after adoption. Marketing, sales, customer service, and field service can then assess the same outcome without mistaking vendor examples for their own results.

Sources

Microsoft Dynamics 365: The new Customer Insights became generally available, 1 September 2023

Microsoft Dynamics 365: Copilot across service, sales and marketing, 7 September 2023

Microsoft Dynamics 365: Connected services in Field Service, 28 September 2023

Forrester: Product managers and generative AI, 15 September 2023

Join the discussion

How do you check that a customer journey gives recipients relevant follow-up? Share a practical example in the discussion.