Core distinction: Auto Predict and Advanced Predictions are part of IPM Insights. Predictive Planning is different: it is launched directly from Planning forms and ad hoc grids.
1. Which capability fits which use case?
Predictive Planning
Planner-driven prediction launched from a data form or ad hoc grid. Best for giving users a quick forecast starting point based on history.
Auto Predict
Scheduled, centralized time-series prediction for a defined slice of data. Useful for recurring forecast review and comparison.
Advanced Predictions
Driver-based machine-learning forecasting that can use multiple variables when a simple time-series pattern is not enough.
Generative AI / Agents
Generative AI explains and summarizes. Agents can go further and orchestrate controlled actions when the required Fusion environment and integrations are available.
2. IPM Insights: the central intelligent analysis layer
IPM Insights is the broader framework for intelligent signals, prediction, anomaly detection, and forecast review. It does not make decisions automatically; it helps users focus on areas that deserve attention.
Forecast Bias
Highlights repeated over- or under-forecasting across entities, regions, or business owners.
Anomaly
Flags unusual values or patterns that may indicate a posting issue, data-quality problem, one-off event, or genuine operational change.
Prediction
Compares submitted forecasts with system-generated predicted values and identifies areas for review.
Period Movement
Surfaces unusually large period-to-period changes and potential breaks in an established trend.
3. Auto Predict vs. Advanced Predictions
Auto Predict
Best for: recurring, scheduled prediction based primarily on historical time-series patterns.
Typical example: generate a monthly revenue or expense prediction and compare it with the submitted forecast.
Setup complexity: medium.
Advanced Predictions
Best for: forecasts where business drivers materially influence the outcome.
Typical example: forecast sales using historical volume plus price, promotion, channel mix, FX, or macroeconomic drivers.
Setup complexity: medium to high, mainly because of driver-data quality and model validation.
4. Predictive Planning: prediction inside the planning workflow
Predictive Planning is tied directly to Planning forms and ad hoc grids. The user requests a prediction while already working in the planning interface, making it a practical productivity feature for bottom-up planning.
5. Quick comparison
| Capability | Where it belongs | Best use | Typical effort |
|---|---|---|---|
| Predictive Planning | Planning forms / ad hoc grids | User-level forecast starting point | Low–Medium |
| Auto Predict | IPM Insights | Scheduled time-series prediction | Medium |
| Advanced Predictions | IPM Insights | Driver-based ML forecasting | Medium–High |
| Generative AI | Reporting / narrative experiences | Explain, summarize, answer questions | Low–Medium |
| AI Agents | Fusion AI Agent Studio ecosystem | Controlled multi-step task execution | High |
6. Generative AI, Reporting Agent, and application-specific intelligence
Generative AI
Turns analysis into business language: summaries, explanations, key movements, and narrative commentary. It complements forecasting rather than replacing it.
Reporting Agent / Ask Oracle
Provides a conversational way to interact with report content and ask questions about financial or planning information where the capability is available.
FCCS / ARCS / EDM intelligence
AI-related features can also appear inside individual EPM business processes, for example around reconciliation, data quality, matching, or metadata governance.
Cash forecasting & simulation
Predictive Cash Forecasting and Monte Carlo-style simulation extend intelligent planning into liquidity, uncertainty, and scenario analysis.
7. AI Agents: powerful, but with the strongest prerequisites
AI Agents are different from ordinary EPM prediction features. They are designed to coordinate a sequence of tasks and actions, potentially combining LLM interaction, business objects, APIs, and enterprise workflows.
Possible EPM-oriented use cases
Launch an approved integration, retrieve status, summarize an exception, guide a user to a report, or coordinate an API-driven workflow.
Why setup is harder
Security, API permissions, Fusion connectivity, action design, testing, and governance all need to be addressed before an agent is suitable for production use.
8. Practical adoption path
- Start with clear business problems. Do not enable AI simply because the feature exists.
- Validate data quality first. Prediction quality depends heavily on reliable history and drivers.
- Separate prediction from explanation. Forecasting and Generative AI solve different problems.
- Use IPM for centralized review. Auto Predict and Advanced Predictions belong in that operating model.
- Use Predictive Planning for planner productivity. It fits directly into form-based planning work.
- Treat Agents as an advanced phase. Platform prerequisites, security, and governance matter more than the demo experience.