Maximo APM 9.2: Condition Insight, Predict, and the Future of Reliability Engineering

The APM Stack in 2026: Four Applications, One Program

Maximo Application Suite has brought enterprise asset management and asset performance management into a single platform, and as of MAS 9.2, the APM stack is the most credible unified offering in the industry. The four applications that make up the APM story are Maximo Health for asset scoring, Maximo Predict for AI-driven failure forecasting, Maximo Monitor for sensor ingestion, and Maximo Reliability Strategies for the RCM methodology. With the 9.2 release, a fifth capability enters the picture: Maximo Condition Insight, an AI-powered diagnostic tool that aggregates data from across the APM stack and produces explainable recommendations in plain language.

For reliability engineers, this is the moment where the platform moves from a collection of related applications to an integrated reliability program. The challenge with APM has never been the individual capabilities. Health scoring, failure prediction, and sensor monitoring are all well-established practices. The challenge has been integrating them into a coherent workflow that a maintenance team can actually execute. Condition Insight addresses that challenge by sitting on top of the other applications, reading their outputs, and producing a unified assessment that a planner can act on without needing to be a data scientist.

The data flows are what matter. Maximo Manage holds the asset master data, work order history, failure codes, and maintenance costs. Maximo Monitor ingests sensor data from IoT devices and SCADA systems. Maximo Health reads data from Manage and produces health, criticality, and risk scores. Maximo Predict reads failure history from Manage and time-series data from Monitor, builds AI models, and forecasts failure probability and days to failure. Maximo Reliability Strategies ties it all together with RCM methodology, FMEA documentation, and maintenance strategy alignment. Condition Insight reads from all of them and tells you what to do next.

This article is a practitioner's guide to the APM stack in MAS 9.2, with particular focus on the new Condition Insight capability and how it changes the daily workflow of a reliability engineer.

Maximo Health: The Foundation of Asset Scoring

Maximo Health is the entry point for most APM programs. It provides asset health scoring, asset criticality scoring, and risk scoring, all driven by the data that already lives in Manage. The scoring is configurable, but the default scoring models that ship with the product are a reasonable starting point for most environments.

The default health score is a weighted combination of several factors: the number of open corrective work orders, the number of open service requests, the remaining useful life (if a Predict model exists), the chronological age as a proportion of expected life, and the asset's criticality. The weights are tunable, and the formula is visible in the Health application, which means you can defend the score in a meeting with the operations team. That matters more than it sounds. A black-box score gets challenged and ignored. A transparent score gets used and trusted.

The default criticality score is a function of the asset hierarchy and the consequence of failure. An asset at the top of a critical process has a higher criticality than an asset at the edge. The score is used as a multiplier in the health formula, so a high-criticality asset with a moderate health score is a higher risk than a low-criticality asset with a similar health score. The risk score, calculated as ((100 - health) / 100) * criticality, is the most actionable output. It is what you should be sorting your work queues by.

# Example: Health score configuration for rotating equipment
apiVersion: mas.ibm.com/v1
kind: HealthScoreProfile
metadata:
  name: rotating-equipment-v1
spec:
  appliesTo:
    assetClass: ROTATING
  factors:
    - name: openCorrectiveWO
      weight: 0.25
      maxValue: 10
    - name: openServiceRequests
      weight: 0.15
      maxValue: 5
    - name: remainingUsefulLife
      weight: 0.30
      maxValue: 1.0
    - name: chronologicalAge
      weight: 0.10
      maxValue: 1.0
    - name: criticality
      weight: 0.20
      maxValue: 100
  riskFormula: "((100 - health) / 100) * criticality"
  alertThresholds:
    - level: urgent
      riskAbove: 75
    - level: high
      riskAbove: 50
    - level: medium
      riskAbove: 25
    - level: low
      riskAbove: 0

A practical pattern is to start with the default scoring models and tune them based on what you learn from the first 90 days of use. The temptation is to spend weeks building a custom scoring model before you turn the feature on. That is a mistake. The default models are good enough to start, and you will learn more from real usage than you will from a whiteboard session. Create a Health Work Queue in the Work Queue Manager, surface the assets that exceed the risk threshold, and let the planner act on them. The queue is a living artifact: as work orders are created and closed, the assets move in and out of the queue automatically.

The 9.2 release adds new Health work queues that surface assets based on combined conditions from Health, Predict, and Monitor. This means you can create a queue that shows assets with a health score below 60, a Predict failure probability above 70 percent, and an active Monitor alert. This multi-signal queue is the foundation of condition-based maintenance, and it is now available out of the box.

Maximo Predict: AI Failure Forecasting That Works

Maximo Predict is the AI-driven failure forecasting capability in the suite. It takes historical failure data from Manage and time-series data from Monitor, builds a model for each asset or asset class, and forecasts the probability of failure, the days to failure, and the most likely failure mode. The output is a numeric score that can be consumed by Health, by a work queue, or by a custom dashboard.

The most important thing to understand about Predict is that it is a per-asset or per-asset-class model. There is no single Predict model that works for everything. The model that predicts failure on a centrifugal pump is different from the model that predicts failure on a switchgear cabinet. The training data comes from your environment, which means the model is only as good as the historical data you have.

The minimum data requirement is roughly 18 months of failure history with consistent failure codes. If your failure codes are inconsistent, the same defect logged under three different codes by three different crews, the model will treat three failures as one. The fix is the same as for any data quality initiative in Maximo: clean up the failure code hierarchy before you turn the model on. This is a data governance project that requires coordination between maintenance planners, reliability engineers, and the Maximo administrator. It is not a quick fix, but it is a prerequisite for credible AI predictions.

A practical deployment pattern is to pick a single asset class with a clean failure history, build a model, and validate it against the most recent 12 months. The validation step is where you learn whether the model is useful. A model that predicts failure 30 days in advance with 80 percent accuracy is a useful model. A model that predicts failure 30 days in advance with 50 percent accuracy is a model you should not put into production. The 9.x releases have made the validation step easier through a holdout test feature that lets you train on a subset of the history and validate against the rest, without exposing the model to production. Use it. A model that has not been validated is a model you do not understand yet.

The predictive maintenance outputs available in the Reliability Engineering Work Center include failure probability, failure contribution breakdown, predicted failure dates, anomaly detection, and the failure probability curve. These outputs give the reliability engineer multiple views into the same data, which supports different decision contexts. The failure probability curve, for example, is more useful for long-term planning, while the anomaly detection is more useful for immediate action.

Maximo Monitor: Sensor Ingestion and Anomaly Detection

Maximo Monitor is the sensor ingestion and processing application in the APM stack. It connects to IoT devices, SCADA systems, and other sensor sources through 1,600-plus pre-configured device partners. The application ingests time-series data, processes it through analytics pipelines, and generates alerts based on threshold conditions and anomaly detection models.

The architecture of Monitor is designed for scale. It can handle high-frequency sensor data from thousands of assets simultaneously, which is a significant capability for organizations with large rotating equipment fleets or distributed infrastructure. The data is stored in a time-series database that is optimized for the query patterns that APM requires, and the alerting engine can route alerts to Maximo work queues, email, or external systems through integration frameworks.

The 9.2 release strengthens the connection between Monitor and the rest of the APM stack. The out-of-the-box integration between Manage, Monitor, and Health enables condition-based maintenance workflows that were previously only possible with custom integration. Time-series data from Monitor can now flow seamlessly through MAS applications via Meters, enabling intelligent and automated CBM. This means a vibration sensor alert in Monitor can automatically update a meter reading in Manage, which triggers a health score recalculation in Health, which surfaces the asset in a work queue for the planner.

For organizations that have been building custom integrations between their sensor platforms and their EAM system, the native integration in MAS 9.2 eliminates a significant amount of integration overhead. The pre-configured device partnerships mean that many common sensor types can be connected without custom code, and the data flows through the standard MAS integration framework, which means it is monitored, logged, and supported by IBM.

Condition Insight: The AI Diagnostic That Changes the Workflow

Maximo Condition Insight is the headline APM feature of MAS 9.2. It is an AI-powered capability, built on IBM watsonx, that evaluates work orders, metrics, time-series data, meter readings, FMEA content, and alerts to assess asset condition, uncover performance patterns, and provide clear, actionable recommendations in plain language.

The problem that Condition Insight solves is the data overload problem. A reliability engineer managing 500 assets has access to work order history, inspection results, meter readings, sensor data, FMEA documentation, and alert histories for each asset. No human can review all of that data for 500 assets on a regular basis. The result is that important signals are missed, and failures occur that could have been predicted if someone had the time to look at all the data.

Condition Insight solves this by doing the analysis in seconds and returning an explainable summary of condition, trends, and recommended actions. The key word is explainable. The capability does not just produce a score. It produces a narrative that a reliability engineer can read, understand, and act on. It tells you what changed, why it matters, and what you should do about it. This explainability is what makes the tool practical for production use, because a recommendation without an explanation is a guess, and reliability engineers do not act on guesses.

The agentic AI aspect of Condition Insight means it can work in concert with other applications and AI capabilities in the MAS ecosystem. It does not operate in isolation. It reads from Health, Predict, Monitor, and Reliability Strategies, synthesizes the data, and produces a unified assessment. This is the integration layer that the APM stack has been missing.

For organizations moving from reactive to condition-based maintenance, Condition Insight is the bridge. CBM has traditionally required significant data analysis by experienced specialists, which limited its adoption. Condition Insight simplifies and scales that analysis, reducing the complexity and the effort required to interpret large volumes of asset data. Instead of spending time piecing together disconnected information, reliability teams can focus more quickly on the actions most likely to protect uptime and asset performance.

Maximo Reliability Strategies: The RCM Methodology Layer

Maximo Reliability Strategies is the application that ties Health, Predict, and Monitor together into a coherent reliability program. It provides pre-built RCM strategies, FMEA documentation templates, and the methodology framework that guides maintenance strategy decisions.

In MAS 9.2, Reliability Strategies includes settings to add RPN (Risk Priority Number) categories and action types. The FMEA content builder, introduced in 9.1 with watsonx support, allows reliability engineers to build FMEA documentation with AI assistance, which accelerates what has traditionally been a time-consuming manual process. The expanded integration with other APM applications means that the strategies defined in Reliability Strategies can be directly connected to the work queues, health scores, and prediction models that drive daily maintenance execution.

The RCM methodology layer is what distinguishes Maximo APM from a collection of monitoring tools. Without a methodology, sensor data and health scores are interesting but not actionable. With a methodology, they become the basis for maintenance strategy decisions that are defensible, repeatable, and aligned with business objectives. The pre-built strategies in Reliability Strategies give organizations a starting point that they can adapt to their specific environment, rather than building an RCM program from scratch.

Practical Implications

For reliability engineers, the 9.2 APM stack is the first version where the integration between the applications is tight enough to support a real condition-based maintenance program without significant custom integration work. The data flows from Manage to Monitor to Health to Predict to Condition Insight to work queues, and the methodology layer in Reliability Strategies provides the framework for turning insights into actions.

The implementation path is not simultaneous. Start with Health, because it depends only on data that already exists in Manage. Turn on the default scoring models, create work queues, and let the planners start using them. Once the Health scores are trusted, add Predict for a single asset class with clean failure data. Validate the model, compare its predictions to actual failures, and iterate. Add Monitor for assets where sensor data is available, starting with the pre-configured device integrations. Add Condition Insight last, because it depends on data from the other applications. If the underlying data is poor, Condition Insight will produce poor recommendations.

The investment is in data quality, not in software licenses. All the APM applications are part of the MAS suite, and the AppPoints model means you can activate them without a separate procurement cycle. The real cost is the time it takes to clean up failure codes, validate sensor data, build and validate Predict models, and train the reliability team on the new workflows. This is months of work, not weeks, but the payoff is a maintenance program that is proactive rather than reactive, and that can be defended with data rather than instinct.

Bottom Line

Maximo APM 9.2 is a credible platform for asset performance management. Health provides the scoring. Predict provides the forecasting. Monitor provides the data. Reliability Strategies provides the methodology. Condition Insight provides the synthesis. For the first time, the integration between these applications is tight enough that a reliability engineer can run a real condition-based maintenance program without building custom integrations or hiring a data science team. The AI capabilities are explainable, the data flows are native, and the methodology framework is built in. For organizations that have been waiting for APM to mature before investing, the wait is over. The platform is ready. The question is whether your data is.