Every software vendor now talks about AI. For finance and operations leaders running Dynamics 365 Finance & Operations, the real question is simpler: where will AI actually save time or reduce risk in our business, and what do we need in place first?
Here is the approach we use with clients.
1. Start with work, not technology
List the repetitive, judgment-light tasks that consume your team’s week. In ERP-heavy organizations, the same candidates come up again and again:
- Following up on overdue customer balances
- Answering vendor questions about payment and PO status
- Reviewing purchase order and delivery-date changes
- Preparing variance commentary for month-end
- Searching for policies, procedures and product information
Score each one on time spent, value of doing it better, and risk if AI gets it wrong. Start where value is high and risk is low.
2. Check your data and permissions
Copilot and AI agents work with the data and permissions you already have. Before any rollout, review:
- Master data quality in D365: customers, vendors, items and financial dimensions
- Oversharing in SharePoint and Teams, since Copilot can surface anything a user can open
- Security roles in Dynamics 365, so agents only see what the user should see
In our experience, permission clean-up is the most commonly underestimated step of an AI rollout.
3. Use what Microsoft already provides
Microsoft is steadily adding Copilot capabilities to Dynamics 365 Finance and Supply Chain Management, including in-app chat and record summaries. Turning on and configuring these features is usually the fastest, lowest-cost first step. Custom agents built with Copilot Studio come next, for processes unique to your business.
4. Pilot small, measure honestly
Pick one or two use cases and a group of 10 to 20 users. Define success before you start, for example “reduce time spent on collections follow-up by 30%”, and measure it. Kill pilots that don’t deliver, and scale the ones that do.
5. Put governance in place early
Decide who can build and publish agents, which data sources they may use, and how actions on ERP data are approved and logged. Copilot Studio environments, data loss prevention policies and Microsoft Purview give you the controls; you need to decide the rules.
Where to start
If you are unsure where AI fits, an AI readiness assessment gives you a prioritized list of use cases and a clear view of what needs fixing first. Or explore our full Copilot & AI services.