AI-generated insights for assets

Starting in Maximo® Application Suite 9. 2, you can use AI-generated insights to get a data-driven snapshot of your asset's performance. The system analyzes historical data to identify trends, risks, and maintenance needs, which can help you make informed decisions about asset health and reliability.

Before you begin

Ensure that AI service enabled by your administrator. For more information, see Enabling AI insights for assets.

Video overview

Watch this video to see how AI-generated asset insights work in Maximo Application Suite 9.2:

Generate insights for an asset

Complete the following steps to generate AI-powered insights:

From the side navigation menu, select Assets and locations.
Select the asset that you want to analyze.
On the Insights card, click Generate insights.
Note: If you do not see the Insights card on the asset dashboard, the AI service might not be enabled yet. Ask your administrator to configure it. For more information, see Enabling AI insights for assets.
Wait for the analysis to complete. The analysis typically takes 2-5 minutes depending on the amount of data and system performance.

The AI engine analyzes available asset data and displays results on the card. You can regenerate insights when new data becomes available to keep the recommendations current.

Tip: Regenerate insights after maintenance work is completed or when significant changes occur.

Understanding asset status

The Insights card displays a status that summarizes your asset's current condition. Use this status to assess whether action is needed.

Table 1. Asset status indicators
Status	Meaning	Recommended action
Needs attention	The asset condition requires maintenance or intervention.	Review insights immediately and schedule maintenance.
Not enough data	Insufficient historical data for AI-based analysis.	Continue creating work orders and recording meter readings, then regenerate insights.
Normal	The asset is healthy and operating as expected.	Continue regular monitoring and preventive maintenance schedule.

Understanding confidence levels

Each AI-generated insight includes a confidence score that indicates how strongly the system supports its analysis. Use these levels to gauge the reliability of insights before you take action.

Table 2. Confidence levels
Confidence level	Meaning
High	The insight is strongly supported by data and can be acted on with confidence. Recommended for immediate action.
Medium	The insight is reasonably reliable but might require additional validation. Consider verifying with subject matter experts before you act.
Low	The insight is based on limited or uncertain data. Use caution and gather more information before you act.

Data sources used for analysis

The AI engine combines multiple data sources to generate accurate insights. Understanding which data is analyzed helps you ensure data quality and completeness.

Table 3. Data types analyzed
Data type	Description	Requirement
Work orders	All types of work orders, including preventive, corrective, and failure, are analyzed to identify maintenance patterns and issues.	Required
Asset details	Basic information such as specifications, age, expected service life, and attributes.	Required
Reliability strategy	Documented failure modes and mitigation actions that provide context for risk assessment.	Optional but recommended
Meter data	Usage and performance readings that help detect anomalies and trends.	Optional but improves accuracy
Health score	Overall health rating that provides a baseline for condition assessment.	Optional but improves accuracy

Important: Data quality directly impacts insight accuracy. Ensure that work orders include problem codes, accurate timestamps, and detailed descriptions.

Components of the Insights card

The Insights card presents information in three sections. Each section serves a specific purpose.

Table 4. Insights card components
Component	Description
Condition summary	A concise overview of the asset's health based on historical data, current status, and identified patterns. This section provides context for the detailed insights.
Insights	Specific observations about maintenance trends, overdue tasks, anomalies, or data quality issues. Each insight highlights a particular aspect of asset performance or maintenance history.
Recommendations	Prioritized, actionable steps to improve asset reliability and performance. Recommendations are ordered by urgency and impact, with the most critical actions listed first.

Example: Interpreting insights for an air handling unit

This example demonstrates how to interpret AI-generated insights for a 72-year-old air handling unit with overdue maintenance. Use this example as a guide for understanding insights on your own assets.

Scenario: Air Handling Unit #1 is operating but its health score is Fair at 56.44, three work orders are overdue, and it is 72 years old with zero remaining useful life.

Table 5. Condition summary interpretation
Summary	What this means
Air Handling Unit #1 is operating but its health score is Fair at 56.44, three work orders are overdue, and it is 72 years old with zero remaining useful life, indicating the unit needs attention.	The asset is aging and at increased risk due to missed maintenance. Current health is only fair. The unit has exceeded its expected service life and requires immediate attention to prevent unexpected failures and ensure continued operation.

Table 6. Insights interpretation
Insight	Interpretation
Three work orders are overdue for the air handling unit according to work order data.	Indicates missed scheduled maintenance, which increases failure risk and highlights gaps in planning or scheduling processes.
The health score of the air handling unit is 56.44, which falls in the Fair range.	Health is fair, which suggests potential degradation. Not critical now, but monitor the trend closely.
Filter replacement and intake filter inspections have recurred multiple times in corrective work orders.	Recurring filter issues indicate a need for proactive maintenance scheduling to prevent airflow obstruction and system inefficiency.
The asset is 72 years old and has zero remaining useful life according to the health score contributors.	End-of-life condition with higher probability of age-related failures. Begin replacement or major refurbishment planning.
No meter anomalies or changepoints have been recorded for the air handling unit.	No sensor-indicated abnormalities, but absence of anomalies doesn't negate risk from age and missed preventive maintenance. Consider installing performance meters for early anomaly detection.

Table 7. Recommendations interpretation
Recommendation	Expected outcome
Complete the overdue filter maintenance work orders to prevent airflow obstruction.	Rapid risk reduction and baseline restoration. Prevents avoidable failures and maintains system efficiency.
Establish a proactive filter replacement schedule based on usage to avoid recurring corrective actions.	Reduces recurring maintenance issues and improves system reliability through preventive scheduling.
Conduct a comprehensive condition assessment of the 72-year-old air handling unit to evaluate replacement or major refurbishment.	Plan capital and downtime proactively to avoid unplanned outages. Consider condition-based justification for replacement.
Review and prioritize open service requests to lower the health score contributors and improve the overall health rating.	Improves asset health score and reduces risk of unexpected failures through systematic issue resolution.
Install or activate performance meters to capture operational data for early anomaly detection.	Augments meter data, which enables early anomaly detection, predictive maintenance, and alerting.

Taking action on insights

After you review the insights, follow these best practices to maximize their value.

Prioritize high-risk assets: If insights indicate that an asset is flagged as high risk or shows recurring issues, schedule maintenance within 24-48 hours. Acting on these alerts helps prevent unexpected failures and reduces downtime.
Plan preventive actions: Look for patterns in the insights, such as repeated failures or anomalies in meter readings. Use these patterns to plan preventive maintenance tasks before issues escalate. This approach improves reliability and extends asset life.
Optimize resource allocation: Focus your resources on assets that show declining performance or fair health scores. By targeting these assets, you can avoid over-maintaining healthy equipment and ensure that critical assets receive timely attention.
Improve data quality: If insights highlight missing problem codes or data inconsistencies, work with your team to improve data entry practices. Better data quality leads to more accurate insights over time.
Track outcomes: After you implement the recommendations, regenerate insights to verify improvements. This regeneration helps you measure the effectiveness of your actions and refine your maintenance strategy.

FAQs for insights issues

Why does it take so long to generate insights?: Generating insights is a long-running process that depends on the speed of the Maximo Application Suite environment and the amount of available data. Wait for the process to complete. After generation, insights are stored in memory for faster retrieval.
What can I do if there is not enough data?: If there is not enough data, the feature might not generate accurate insights. Update the data range to include historical information.
How do I include or exclude data sources?: This option is not available.
Can I add or replace the data sources that are used for insights?: No, you cannot add or replace the data sources that are used to generate insights. The feature uses predefined data sources within Maximo Application Suite.
Why are my insights incorrect?: Incorrect insights may occur due to customizations in Maximo Application Suite that affect how AI interprets data. Review your Maximo Application Suite customizations and verify that the data is accurate.
How do I check for issues with the AI service?: Track the number of AI tokens that are used in the service and ensure that you do not exceed the defined usage limit.