Deprecated: announced 14 September 2026. Support ends 30 October 2026, after which the feature is removed. Move to forecast scenarios in workforce engagement management. See the deprecation notice.

What it does

AI-backed report that predicts conversation volumes and staffing demand from historical data. Produces daily forecasts up to 6 months ahead and 15-minute intraday forecasts up to 6 weeks ahead. Configurable for SLA targets, shrinkage, and concurrency per channel.

Key facts

• Standalone only: not available in embedded deployments
• Minimum 2 weeks of historical data required before forecasts generate
• Intraday (15-min) forecast uses most recent 6 weeks only: older history is not used for the intraday model
• Daily forecast range: up to 6 months if ≥12 months of history available; otherwise half the available history
• AI bot conversations (unescalated) are excluded from staffing demand calculation
• First enable: data can take up to 24 hours to appear
• Global configuration: Required Service Level (%), Target Answer Time, Shrinkage (%), Concurrency
• Channel overrides allow different concurrency/SLA settings per channel (e.g. voice concurrency = 1, chat concurrency = 3)
• Holiday calendar integration available for seasonality modelling
• Read privilege on msdyn_dataanalyticsreport_forecast table required for non-admin users to view the report

When to use / skip

Turn this on for any standalone deployment that needs staffing forecasts. Enable early, more history = better model, so turn it on during UAT to have data ready for go-live.

Configuration decisions

• Concurrency per channel: most critical accuracy lever; voice = 1, chat/messaging = 2–5 depending on client practice; validate with operations team before configuring
• Shrinkage %: most clients underestimate; typical range is 20–35%; underconfiguring understates staffing need
• Channel overrides: required for any mixed-channel deployment; a single global concurrency value cannot serve both voice and digital channels accurately
• Historical data start date: defaults to earliest records up to 2 years back; adjust if there is a known data quality issue in older records

Gotchas

• Deprecated 14 September 2026: forecasting for case and conversation volumes, and for representatives on conversations, is deprecated in Contact Center and Customer Service. Support ends 30 October 2026 and it's removed after that. The replacement is forecast scenarios in workforce engagement management, which also bring capacity planning, scheduling and intraday. See the deprecation page.
• Concurrency misconfiguration kills accuracy. Chat reps handling 3 conversations at once but concurrency set to 1 = 3x overstated staffing demand.
• 24-hour data lag catches everyone. Tell the client up front: enable this, then wait a day for data to appear.
• This is not the same as WFM forecasting (feature-wfm-forecasting-capacity.md). Both exist. This AI report does SLA/concurrency modelling. WFM forecasting plugs into capacity plans and scenario modelling.

Consultant notes

• Enable this early in the project, not at go-live. More historical data equals a better model. Turning it on during UAT rather than the day of go-live gives the model a few weeks of training data and avoids the "the forecast looks wrong" conversation in the first post-go-live week.
• Concurrency misconfiguration is the accuracy killer to verify upfront. Ask the client's ops team how many conversations a chat rep actually handles simultaneously, not how many the system allows. If reps regularly take 3 simultaneous chats and concurrency is set to 1, the staffing demand output is meaningless.
• This is not the same as WFM forecasting: they do different things. If the client asks "which one should we use?" the answer is usually both, but for different purposes: this for SLA/demand modelling, WFM forecasting for capacity planning and schedule building.

---

Deprecated — support ends 30 October 2026. Revisit only to confirm the removal date and that migrations to forecast scenarios are done.