Oracle Cloud EPM IPM Best Practices: A Finance-Led AI Adoption Playbook

AI in finance is no longer a future concept. It is becoming part of how organizations forecast, analyze, explain and improve performance. But successful AI adoption in finance does not happen by simply enabling a feature. It requires the right use case, reliable data, business ownership, testing discipline, measurable outcomes and a clear plan for ongoing model care. Oracle Cloud EPM Intelligent Performance Management, or IPM, gives finance teams a practical path to bring AI into planning, forecasting, insights and narrative analysis. Oracle’s July 2025 Cloud Customer Connect session on IPM – AI in EPM Best Practices, Use Cases and Implementation Considerations focuses on how organizations can select the right AI use cases, prepare data, structure pilots, tune models and move toward steady-state adoption. For CFOs, FP&A leaders and EPM administrators, the most important takeaway is simple: IPM works best when AI is owned by finance, governed by clear process and measured against real business outcomes.

Why Oracle Cloud EPM IPM Matters Now

Finance teams face constant pressure to improve forecast accuracy, reduce manual analysis, explain performance faster and support better business decisions. Traditional planning and reporting processes often depend on historical trends, spreadsheet reviews, manual variance commentary and repeated forecast cycles. IPM changes this by embedding AI into the finance environment itself. Oracle Cloud EPM has continued to expand AI capabilities across Predictive Planning, Auto Predict, IPM Insights, Bring Your Own ML, Predictive Cash Forecasting, Generative AI for management reporting narratives, Gen AI in Insights, Insights in Smart View, Dynamic Parent Insights and Predictions, PCM Agent and Advanced Predictions. This creates a broader AI foundation for finance teams across three areas:

Analysis Automation

Insights and Gen AI can help surface anomalies, explain variances and reduce manual review effort.

Reliable Forecasting

Predictive Planning, Auto Predict and Advanced Predictions help finance teams compare traditional forecasts with AI-generated projections.

Task Automation

PCM Agent and future transaction-matching-related capabilities indicate a growing direction toward AI-assisted task execution in EPM. The opportunity is to forecast faster. The opportunity is to build a more intelligent finance operating model.

Start with the Right Use Case

A common mistake in AI adoption is starting with the most complex or high-profile use case first. Oracle recommends a phased approach where organizations begin by selecting the right use case, understanding which EPM tools support it and evaluating viability based on AI maturity and data availability. The best first use case should be easy enough to configure but meaningful enough to create business confidence.

Good starting points may include:

• Operating expense predictions
• Salary and utility cost forecasting
• Basic anomaly insights
• Forecast variance monitoring
• Revenue or COGS insights where historical data is strong
• Headcount trend analysis
• Cash flow forecasting
• Inventory or sales volume predictions

Oracle’s use case framework compares ease and impact. Simple use cases help teams learn quickly, while high-impact use cases create stronger FP&A buy-in. A practical adoption strategy is to start with one focused use case, prove value, then expand.

Where IPM Can Create Business Value

Oracle identifies several business areas where AI in EPM can support financial and operational improvement. These include demand risk and rising costs, sustainability planning, service-based business models, workforce retention and recruitment, cash optimization, project financial management and sales performance management.

Examples include:

Demand Risk and Rising Costs

Organizations can analyze shipped units, margin, backlog, revenue, expense, energy costs, cancellations, spoilage and rebates. Drivers may include price changes, promotions, GDP, inflation, interest rates, freight indexes and supply chain pressure.

Sustainability Planning

Teams can plan and report on energy consumption, shipping distance, product volume, order volume, customer mix, supplier mix, commodity mix and packaging.

Workforce Planning

HR and finance teams can analyze headcount, attrition, leaves, employee tenure, job classifications and employee satisfaction.

Working Capital

Organizations can review cash from accounts receivable and payable, customer revenue, credit score, supplier data, purchase volumes and average order value.

Project Financial Management

Project-oriented industries can forecast revenue, costs, margins, billable hours, staffing mix, chargeability rates and project run rates.

Sales Performance Management

Sales teams can plan sales targets, commissions, bookings, sales rep profiles, account attributes and territory attributes. This range of use cases shows that IPM can support more than finance-only forecasting. It can help connect operational drivers with financial outcomes.

Data Readiness Is the First Real Test

AI adoption fails when teams underestimate the importance of data readiness. Oracle recommends confirming that the required data exists in the EPM application for the specific data slices being used. The suggested validation method is to create a Smart View report using those intersections. Oracle also recommends that data should be pre-processed, accurate and consistent. For meaningful insights or predictions, organizations should have at least 2x historical data. For example, predicting one year requires two years of historical data. For stronger back testing, three years of historical data is preferred.

Other important data-readiness checks include:

• Ensure at least 50% or more data availability across historical periods.
• Ensure data exists in the last period before the prediction period.
• Maintain a strong forecast archival process for variance insights.
• For new products, stores or territories, consider using similar historical data where direct history does not exist.

The message is clear: there is no AI without usable data. But Oracle also makes an important practical point — organizations do not need massive “big data” volumes, and the data does not need to be perfect because EPM includes data-cleansing techniques.

Build the Pilot Around Business Ownership

A strong IPM pilot requires cross-functional ownership.

Oracle identifies four participant groups:

• Project Sponsor
• Finance User
• EPM Admin
• Data Analytics, optional

Each role has a different responsibility. The project sponsor provides leadership, strategic alignment, stakeholder management and change management. The finance user identifies business use cases, input drivers, KPIs, success criteria and user feedback. The EPM admin configures jobs, validates data availability, prepares forms and dashboards, manages EPM changes and supports data integration. A data analytics resource can help measure accuracy, compare models and advise on algorithms, but Oracle notes this role is optional because EPM handles much of the AI work behind the scenes. This structure is important. AI adoption should not be left only to IT or data science. Finance users must own the business problem, success criteria and interpretation of results.

Plan Realistic Time Commitment

Oracle provides practical pilot effort guidance. During the pilot, EPM Admin involvement may be around 5–15 hours per week, while finance users may need 10–15 hours per week. After go-live, the EPM Admin commitment may reduce to 2–4 hours per month, while finance users may need 1–2 hours per month. If a data analytics resource is included, pilot involvement may be 5–10 hours per week. This helps organizations set the right expectations. IPM is not a heavy data science implementation, but it still requires business focus. The pilot needs enough time for configuration, review, testing, tuning and user feedback.

Configure IPM with the Right EPM Architecture

Oracle recommends considering both BSO and ASO design for IPM. IPM works on both BSO and ASO, and Hybrid should be enabled to use IPM features.

Oracle’s guidance is practical:

• Predictions are best done in BSO because they can be baselined against forecasts and compared at the same level as traditional forecasts.
• Insights are best done in ASO because ASO typically has more actuals and reporting data.
• IPM works across monthly, weekly, daily and custom period granularities.
• Application design should be future-proofed for increased data volume.

This means IPM should be considered during EPM design and optimization. The application must be structured to support the AI use cases the business wants to run.

Use Multiple IPM Jobs for Better Control

Oracle recommends breaking IPM jobs into multiple jobs because measures, thresholds and model settings can vary.

Jobs can be grouped by:

• Entity, account or product group
• Insight jobs versus prediction jobs
• Variations with events and without events

This is a good governance practice. A sales forecast job may need different drivers, settings and thresholds from an OpEx forecast. A revenue insight job may need a different anomaly method from a workforce trend job. Separating jobs makes monitoring, tuning and ownership easier.

Preserve Prediction Traceability

AI output should never become a black box inside the forecast. Oracle recommends creating a new Scenario, Version or Plan Element member to store predictions. This allows predictions to be compared against the existing forecast. Oracle also recommends adding an adjustment member so finance users can adjust the forecast on top of the prediction while maintaining traceability.

A practical structure may include:

• Forecast
• Prediction
• Adjustment
• Prediction plus Adjustment

This separates the AI-generated prediction from human judgment. It also helps finance leaders understand where the final forecast changed and why.

Measure AI Success with Forecast Value Add

AI should be measured against business value, not novelty. Oracle recommends measuring comparative accuracy against actuals and different forecast methods. Oracle also highlights Forecast Value Add, or FVA, which measures improvement over the existing forecast.

A strong IPM success framework should include:

• Accuracy against actuals
• Forecast Value Add
• Improvement over current forecast methods
• Speed of analysis
• Reduction in manual review effort
• User satisfaction
• Adoption by finance users
• Ability to support decisions earlier

This is where AI becomes credible. Finance teams need evidence that predictions are better, faster or more useful than the current process.

Use Back Testing to Build Trust

Back testing is essential for AI adoption. Oracle recommends at least 24 months of historical actuals to ensure adequate data for algorithms, especially seasonal algorithms. Ideally, organizations should have 36 months of history to back test on one year of data. Back testing allows teams to generate predictions for past periods and compare them with actuals. Historical forecasts are optional, but Oracle recommends optimizing the forecast archival process so historical forecasts are available for comparison. This helps answer a critical business question: are AI predictions better than the existing forecast?

Dynamic Parent Predictions and Insights

Not every organization has clean historical data at the lowest level of detail. In many cases, leaf-level data is incomplete, spotty, periodic or too granular for meaningful prediction. Oracle highlights Dynamic Parent Predictions and Insights as a way to configure AI at parent-level members. This provides greater flexibility, improves accuracy and reduces workarounds.

Example use cases include:

• Total OpEx forecasting when sub-expense history is incomplete
• Department-level headcount planning when team-level history is unavailable
• Program-level project forecasting where project-level detail is missing
• Revenue planning by brand, product group or territory
• Marketing spend forecasting by category rather than campaign-level detail

This is highly relevant for FP&A teams because real-world planning data is rarely perfect at every intersection.

Tune the Model Based on Business Context

The refinement phase should focus on accuracy, performance, advanced configuration and how AI changes the end-to-end business process.

Oracle identifies several tuning options for Auto Predict:

• Events
• Data screening
• Data attributes
• Seasonal and non-seasonal methods
• Missing value treatment

Events can represent recurring, one-off or skipped events such as holidays, supply disruptions or pandemics. Data screening helps with missing values and outliers. Method selection helps align prediction logic with seasonal or non-seasonal data behavior. For Insights, events can be tagged for anomalies, and different metric methods can be selected depending on the data and business question.

Choose the Right Metric for the Right Question

Oracle identifies multiple anomaly, variance and bias methods.

For anomaly detection:

• Z-score is best for clean, symmetric data that looks roughly normal.
• Modified Z-score works better with outliers or smaller sample sizes.
• Interquartile Range is useful when data is not normally distributed.

For variance measurement:

• MAPE shows average forecast error as a percentage.
• MAD shows average error in the same unit as the data.
• RMSE penalizes larger errors more heavily.
• Total Deviation shows the total size of all errors.
• Total Deviation Percentage expresses total error as a percentage of actuals.

For bias analysis, Oracle highlights Mean Percentage Error and Relative Percentage Difference. Bias analysis helps determine whether the forecast consistently overestimates or underestimates actual results. This matters because the “best” metric depends on the business question. A forecast used for revenue planning may need a different accuracy lens from a forecast used for cost control or capacity planning.

Advanced Predictions: Moving Beyond Historical Trend Forecasting

Advanced Predictions, also described as multivariate predictions, allow finance teams to use multiple data points and more sophisticated algorithms. Oracle positions Advanced Predictions as powerful, accuracy-focused, easy to configure through a step-by-step wizard and designed for finance users inside EPM without requiring data science skills. The difference between univariate and multivariate prediction is important. A univariate prediction mainly uses historical target data. A multivariate prediction can include operational drivers and contextual factors such as marketing promotions, average sales price, industry volumes and GDP rate. This helps finance teams build predictions that reflect business reality, not just historical patterns.

Steady-State Care: Keep the Model Alive

The final stage of IPM adoption is end-user rollout and steady-state care. Oracle identifies this phase as including change management, end-user education, user experience, success evaluation, model monitoring and continued evolution of business processes. This is where IPM becomes a real operating capability.

After go-live, organizations should monitor:

• Forecast accuracy
• Model drift
• User adoption
• Manual overrides
• Business feedback
• New data availability
• New use case opportunities
• Changes in business drivers
• Performance of insights and predictions

AI adoption is not complete at launch. It must be monitored, tuned and evolved as the business changes.

Addressing Common AI Adoption Concerns

Oracle addresses several common objections to AI adoption in EPM.

“Our data is not clean enough.”

Oracle notes that organizations need two to three years of historical data, but they do not need large “big data” volumes. Data also does not need to be perfectly clean because EPM Cloud includes data-cleansing techniques.

“Our team lacks AI knowledge.”

Oracle states that AI knowledge is helpful but not mandatory. The product is designed for EPM users with typical administrative skills such as loading data, running rules, building forms and creating dashboards.

“Finance users will resist AI.”

IPM does not necessarily replace the existing EPM process. It can help finance teams consolidate, forecast and report faster and more accurately.

“We do not have time.”

Oracle states that IPM is not a significant implementation effort and that many customers can be up and running in production with a pilot in a matter of weeks. These points are useful for finance leaders trying to move AI from discussion to action.

How NexInfo Can Help

NexInfo helps organizations adopt Oracle Cloud EPM IPM with a practical, finance-first approach across readiness, data validation, pilot planning, configuration, model tuning and managed support. NexInfo’s delivery model is supported by ISO 9001 for Quality Management and ISO 27001 for Information Security, helping organizations strengthen process quality, delivery discipline and secure enterprise transformation practices. NexInfo has also received the AI-Enabled Workforce Excellence Award at the 1st Annual Long Beach Business AI Summit, reinforcing its focus on practical AI adoption across workforce enablement, enterprise systems and operational transformation.

NexInfo can support organizations with:

• Oracle EPM IPM readiness assessment
• AI use case selection and prioritization
• Data readiness review
• Smart View validation of data slices
• IPM pilot planning
• BSO, ASO and Hybrid readiness
• Insights configuration
• Auto Predict setup
• Advanced Predictions implementation
• Driver selection and event configuration
• Model tuning and metric selection
• Prediction storage and adjustment design
• Back testing and Forecast Value Add measurement
• Dynamic Parent Predictions and Insights planning
• User training and change management
• Steady-state monitoring and managed support

NexInfo helps finance teams move AI in EPM from experimentation to measurable operating value.

Conclusion

Oracle Cloud EPM IPM gives finance teams a practical way to embed AI into forecasting, planning, insights and reporting. But the strongest results come from disciplined adoption. Organizations should begin with a use case that balances ease and impact, validate data readiness, run a focused pilot, store predictions separately for traceability, measure Forecast Value Add, use back testing to build confidence and monitor the model after go-live. The real value of IPM is not only better predictions. It is better finance decision-making supported by governed AI, stronger data discipline and measurable business outcomes.

NexInfo helps organizations adopt Oracle Cloud EPM IPM with Oracle EPM expertise, ISO-certified delivery governance, AI-enabled transformation experience and a finance-led implementation approach.

FAQ

What is Oracle Cloud EPM IPM?

Oracle Cloud EPM IPM, or Intelligent Performance Management, is Oracle’s AI and machine learning capability for EPM. It supports capabilities such as Insights, Predictive Planning, Auto Predict, Advanced Predictions and Gen AI-enabled narrative and insight use cases.

What is the best way to start with IPM?

The best way to start is to select a focused use case, validate data availability, run a pilot, define success metrics, tune the model and then roll out to end users with steady-state monitoring. Oracle recommends a four-phase approach covering use case selection, pilot, refinement and rollout.

What IPM use cases should finance teams consider first?

Finance teams can start with use cases such as OpEx prediction, basic insights, revenue or COGS analysis, sales volume forecasting, cash flow forecasting, headcount planning, inventory planning or top expense trend analysis.

How much historical data is needed for IPM?

Oracle recommends at least 24 months of history for algorithms and ideally 36 months for back testing on one year of data.

Does IPM require perfect data?

No. Oracle notes that data does not need to be perfectly clean and that EPM Cloud includes data-cleansing techniques. However, data should be accurate, consistent and available for the slices used in the AI job.

Who should participate in an IPM pilot?

An IPM pilot should include a project sponsor, finance user, EPM admin and optional data analytics support. Finance users provide domain knowledge, while EPM admins handle configuration and data readiness.

Should predictions be stored separately?

Yes. Oracle recommends creating a new Scenario, Version or Plan Element member to store predictions and adding an adjustment member to maintain traceability between AI prediction and human adjustment.

What does Forecast Value Add?

Forecast Value Add, or FVA, measures improvement over the existing forecast. It helps determine whether AI predictions create measurable forecasting improvement.

What is back testing in Oracle EPM IPM?

Back testing uses historical actuals to generate predictions for past periods and then compares those predictions with actuals or historical forecasts. It helps finance teams evaluate prediction quality before broader adoption.

What are Dynamic Parent Predictions and Insights?

Dynamic Parent Predictions and Insights allow AI to work at parent-level members when leaf-level data is incomplete, spotty or too detailed. Use cases include OpEx, headcount, program budgets, revenue by category and marketing spend planning.

What are Advanced Predictions?

Advanced Predictions are multivariate predictions that use multiple data points, drivers and algorithms. They are embedded in EPM and designed for finance users through a step-by-step configuration wizard.

What is the difference between univariate and multivariate prediction?

Univariate prediction mainly uses historical target data. Multivariate prediction can include operational drivers and contextual factors such as promotions, average sales price, industry volumes and GDP rate.

What metrics can be used to evaluate IPM accuracy?

Oracle identifies methods such as Z-score, Modified Z-score, Interquartile Range, MAPE, MAD, RMSE, Total Deviation, Total Deviation Percentage, Mean Percentage Error and Relative Percentage Difference.

Do finance teams need data scientists to use IPM?

Oracle states that AI knowledge is helpful but not mandatory. IPM is designed for EPM users with typical administrative skills such as loading data, running rules, building forms and creating dashboards.

How can NexInfo help with Oracle Cloud EPM IPM?

NexInfo can help with IPM readiness assessment, use case selection, data validation, pilot planning, Insights setup, Auto Predict, Advanced Predictions, model tuning, back testing, Forecast Value Add measurement, user adoption and managed support. Received the updated file. Use this citation for this uploaded source going forward