AI in Maximo in 2026: Condition Insight, Predict, Visual Inspection, and the Agentic Shift

AI in Maximo in 2026: Condition Insight, Predict, Visual Inspection, and the Agentic ShiftAI inside Maximo is no longer a research project. As of the Maximo Application Suite 9.2 release in June 2026, every reliability engineer, planner, technician, and safety lead in a Maximo shop is working alongside AI capabilities that have shipped to production in hundreds of organizations. The shift over the last eighteen months has been less about new AI capabilities and more about where the AI lives. The dominant pattern in 2026 is not "AI as a separate layer that integrates with Maximo" but "AI inside the workflow, embedded in the screens that reliability and maintenance teams already use every day."

This article is a practitioner's guide to the AI capabilities inside MAS 9.2: what each one does, where it fits in the maintenance workflow, what the realistic value looks like in production, and how to sequence the rollout so that the data foundation is in place before the AI capabilities go live. We focus on the AI capabilities that ship with Maximo or with the watsonx platform that IBM bundles with Maximo deployments, not on third-party AI integrations. Those are real and increasingly important, but the first-party capabilities are where most organizations should start.

The Asset-First AI Shift in MAS 9.2

The defining characteristic of the AI capabilities inside MAS 9.2 is that they are asset-first, not data-first. Previous generations of AI in asset management required reliability engineers to leave Maximo, open a separate analytics environment, export asset data to a data lake, build or train a model, and then bring the model's predictions back into Maximo as a manual overlay. This worked for the data scientists who built the models. It did not work for the reliability engineers, planners, and technicians who needed the predictions to make daily decisions.

MAS 9.2 reverses this pattern. AI lives inside the Maximo screens that the maintenance and reliability team already uses. The predictions, condition assessments, recommended actions, and visual inspection results appear in context, alongside the asset record, the work order, the inspection form, or the safety checklist. The reliability engineer does not need to know which model produced the recommendation. They need to know that the recommendation is grounded in the right data, that it explains its reasoning, and that they can act on it without leaving the screen.

The architectural change that enables this is the deeper integration between Maximo and the IBM watsonx platform. watsonx provides the model serving, the data ingestion, and the agentic orchestration that the Maximo AI capabilities call into. From the Maximo user's perspective, the AI is just there. From the architect's perspective, the AI is a set of services on the watsonx platform that Maximo calls through well-defined APIs.

The five AI capabilities that most organizations will encounter first in MAS 9.2 are:

• Maximo Condition Insight, which provides AI-driven condition-based maintenance recommendations powered by watsonx.
• Maximo Predict, which builds and deploys machine-learning models that forecast days to failure, probability of failure, and remaining useful life.
• Maximo Monitor (formerly Maximo Health), which provides AI-enabled remote monitoring of asset condition at scale.
• Maximo Visual Inspection, which uses computer vision models to detect defects from images and video, with the option to run inference locally on mobile devices.
• Maximo Assistant on Mobile, which uses natural-language AI to help technicians find asset information, review history, and complete work efficiently in the field.

Each of these addresses a different maintenance problem. Together they cover the AI surface that most asset-intensive organizations need.

Maximo Condition Insight: Agentic CBM at Scale

Condition Insight is the flagship AI capability in MAS 9.2. It is an agentic AI capability within Maximo Asset Performance Management (APM) that evaluates work orders, metrics, time-series data, meter readings, FMEA (Failure Mode and Effects Analysis) records, and alerts to produce a clear, plain-language assessment of asset condition, emerging trends, and recommended corrective actions. The output is not a model score. It is a written explanation that a reliability engineer or maintenance planner can read, evaluate, and act on.

The traditional barrier to condition-based maintenance at scale has been the data analysis itself. CBM has always made sense conceptually, but in practice it requires a specialist to look at sensor data, correlate it with work order history, factor in the FMEA knowledge, and produce a recommendation. That specialist time does not scale. Most organizations have one or two reliability engineers who can do this work, and they can cover maybe 5% of the asset base.

Condition Insight removes that barrier by analyzing asset data in seconds and returning a clear, explainable summary of condition, trends and recommended actions. The agentic AI pattern means the model is not just running a single inference. It is breaking the question down into sub-tasks, pulling the relevant data from each Maximo module, reconciling the data, applying the FMEA logic, and producing a recommendation that explains its reasoning. The reliability engineer reviews the recommendation, accepts it, adjusts it, or rejects it, and the system learns from the feedback over time.

A practical example from a utility deployment in early 2026: a transmission transformer with multiple sensor inputs (top-oil temperature, load current, dissolved gas analysis readings) was producing intermittent Condition Insight alerts. The reliability team reviewed the first batch of recommendations, accepted about 70% as written, modified about 20%, and rejected about 10% (where the model's recommendation did not match the engineer's domain knowledge). After three months of feedback, the acceptance rate had climbed to about 85%, and the team reported that Condition Insight was surfacing at least two emerging issues per month that they would not have caught through their existing manual review process. The value was not in replacing the reliability engineer. It was in giving the reliability engineer leverage.

For organizations rolling out Condition Insight, the realistic timeline is:

• Month 1 to 2: Data assessment. Identify the assets where the data foundation is mature enough to support CBM (clean meter history, complete FMEA records, consistent work order history). Do not try to roll out Condition Insight on every asset on day one.
• Month 3 to 4: Pilot on a subset of 20 to 50 critical assets. The reliability team reviews every Condition Insight recommendation, builds confidence in the system, and establishes the feedback loop.
• Month 5 to 12: Phased rollout to additional asset classes as the data foundation matures and the team builds operational muscle.

Maximo Predict: ML Models for Failure Forecasting

Predict is the more traditional machine-learning capability inside MAS. It uses AI and machine learning to predict asset performance and maintenance needs by analyzing time-series data from Maximo Monitor and failure data from Maximo Manage. The output is a forecast: probability of failure, days to failure, remaining useful life, and other key indicators that planners and reliability engineers can act on.

The data scientist workflow inside Predict is structured but flexible. The reliability engineer or data scientist creates an asset group (for example, all 480V motor starters in the Plant 1 motor control center), connects the relevant data sources, and uses a default notebook to build and train the predictive model. The default notebooks ship with sensible defaults for common asset types (rotating equipment, electrical assets, vehicles). For specialized assets, the data scientist can configure custom notebooks that extend the defaults or are completely custom.

A critical operational detail: the trained models must be deployed in Watson Machine Learning, which is the model-serving environment inside the watsonx platform. The integration between Maximo and Watson Machine Learning is what makes the predictions appear in the Maximo UI without custom development. Once a model is deployed, the Predictions section on each asset record is populated automatically.

The use cases where Predict produces the most value are:

• Rotating equipment (pumps, motors, compressors) where vibration, temperature, and current data correlates strongly with failure modes.
• Electrical assets (transformers, switchgear) where dissolved gas analysis, partial discharge, and load data predict failures weeks or months in advance.
• Vehicle and mobile equipment where usage data, telematics, and maintenance history combine to predict component failures.
• Instrumentation (control valves, transmitters) where calibration drift and historical failure patterns are predictable.

The use cases where Predict produces less value are:

• Assets with sparse data history (less than two years of consistent records).
• Assets with no instrumentation (no sensor data, only manual inspections).
• Assets with one-off failure modes that do not repeat.

For organizations new to Predict, the recommendation is to start with a single asset class that has rich data history, train and deploy a model, validate the predictions against actual failures for three to six months, and then expand to additional asset classes once the validation is solid.

Maximo Visual Inspection: Computer Vision in the Field

Visual Inspection uses computer vision models to detect defects from images and video. The training process is well within the reach of a maintenance engineer with no formal machine-learning background. The workflow is:

• Capture a set of reference images showing the asset in good condition, with examples of the various defect types you want to detect.
• Train the model using those images, typically in less than an hour for a focused use case.
• Deploy the model to the mobile app (or to the inference server for centralized inspection).
• Technicians in the field capture new images, the model scores them, and the inspection result (pass, fail, defect type, defect severity) is recorded against the asset.

The MAS 9.2 release added a particularly important capability: local inference directly on the mobile device. This means that a technician with a phone or tablet in a facility with no network connectivity can still run visual inspection against a trained model. The model runs on the device, the inference happens locally, and the result syncs back to Maximo when connectivity is restored. For industrial environments where network coverage is unreliable (underground, remote, inside metal enclosures), this is the difference between a usable capability and an unusable one.

The use cases that work well with Visual Inspection are:

• Visual inspection of assets where the defects are visually distinguishable (rust, corrosion, cracks, leaks, wear patterns).
• Quality control on manufactured components where the pass/fail criteria can be expressed visually.
• Safety inspections where personal protective equipment compliance, signage, or hazard markers need to be checked at scale.

The use cases that work less well are:

• Defects that require domain expertise to interpret (subtle changes in equipment behavior, sounds, vibrations).
• Defects inside equipment enclosures that cannot be photographed without disassembly.

The most common production pattern in 2026 is to deploy Visual Inspection on a small number of high-value inspection types (corrosion on outdoor equipment, PPE compliance at construction sites, weld quality on manufactured components) and expand from there based on technician feedback.

Maximo Assistant on Mobile and Agentic Workflows

Maximo Assistant on Mobile is the natural-language interface that MAS 9.2 introduces for technicians in the field. The assistant uses natural-language AI to help technicians find asset information, review work order history, get step-by-step guidance on tasks, and complete work efficiently. The key design principle is that the assistant is grounded in the asset data inside Maximo. It does not hallucinate asset specifications or invent maintenance procedures. It retrieves the relevant records and explains them in plain language.

A typical interaction might be:

• Technician: "What is the lubrication interval for the gearbox on Pump 3?"
• Assistant: "The lubrication interval for GEARBOX-PUMP3 is every 6 months or 4,000 operating hours, whichever comes first. The last lubrication was on April 12, 2026, by Tech ID 4421. The next PM is scheduled for October 12, 2026. The current operating hours are 2,847."
• Technician: "Show me the work history for the past year."
• Assistant: "Pump 3 has had 4 work orders in the past 12 months: WO-44102 (vibration anomaly, closed in 3 days), WO-44218 (seal replacement, closed in 5 days), WO-44389 (lubrication PM, closed in 1 day), WO-44501 (coupling inspection, closed in 2 days). The mean time between failures is 73 days."

The value of this pattern is not that the AI replaces the technician's knowledge. It is that the technician can get to the answer in seconds rather than navigating through Maximo menus, running queries, and assembling the answer manually. For senior technicians, this is a productivity gain. For newer technicians, it is a force multiplier that helps them perform at a higher level sooner.

Beyond the mobile assistant, MAS 9.2 introduces agentic workflows that guide decisions and move work forward across reliability, maintenance, and safety processes. These are not chatbots. They are workflow automations that use AI to make decisions within guardrails. For example, an agentic workflow for safety incident triage might read the incident description, classify it against historical incident types, recommend the initial response team, and create the appropriate work orders and notifications, all within the limits defined by the safety team.

Sequencing an AI Rollout in Maximo

The right sequence for rolling out AI capabilities inside Maximo depends on the data foundation. The capabilities that ship inside MAS 9.2 are mature, but they all rely on data quality, and the data quality is something that has to be built over time.

The recommended sequence for organizations starting their Maximo AI journey in 2026 is:

• Audit the data foundation. Look at the asset hierarchy, the meter history, the work order history, and the failure records. Identify the asset classes where the data is complete and consistent enough to support AI.
• Start with Monitor and Predict on a single asset class. These capabilities have the most mature data requirements and produce measurable value within the first six months.
• Add Visual Inspection for high-value inspection types. Visual Inspection does not depend on historical data, so it can be deployed in parallel with Predict.
• Pilot Condition Insight on the same asset classes where Predict and Monitor are deployed. The agentic recommendations layer on top of the underlying predictions.
• Roll out Maximo Assistant on Mobile to the field technician population. The mobile assistant does not require a strong data foundation, but it does require API stability and good integration with the asset and work order data.
• Mature the agentic workflows over time. The agentic workflows are the most experimental of the AI capabilities. Start with the simplest workflows (safety incident triage, MOC routing) and build up to more complex orchestration.

Practical Implications

For reliability engineers and maintenance leaders, the practical implication is that the AI capabilities inside MAS 9.2 are mature enough to drive real value, but the value depends on the data foundation. Organizations with clean asset hierarchies, complete meter history, consistent work order records, and well-maintained FMEA data will see faster returns than organizations still working through data cleanup. The recommendation is to invest in the data foundation first, then layer the AI capabilities on top.

For architects and IT leaders, the practical implication is that the AI capabilities inside Maximo require the watsonx platform. Organizations that have not yet deployed watsonx should plan for that as part of their AI rollout. Organizations that have already deployed watsonx for other use cases (customer service, document processing, IT operations) will find that the Maximo integration is straightforward.

For data scientists and ML engineers, the practical implication is that Maximo Predict and the watsonx integration provide a managed path to deploy models in production without building the surrounding infrastructure from scratch. The custom notebook extension points in Predict allow specialized models to be deployed within the Maximo environment rather than as a separate analytics platform.

Bottom Line

AI inside Maximo in 2026 is asset-first, workflow-native, and grounded in the watsonx platform. Condition Insight provides agentic condition-based maintenance recommendations that reliability engineers can review and act on. Predict provides machine-learning-driven failure forecasting for asset classes with rich data history. Monitor provides AI-enabled remote monitoring at scale. Visual Inspection provides computer vision on mobile devices, with local inference for environments without reliable connectivity. And Maximo Assistant on Mobile gives technicians a natural-language interface to the asset and work order data. The capabilities are mature, but the value depends on the data foundation. Organizations that invest in data quality first, then layer the AI capabilities in the recommended sequence, will see measurable returns within the first year. Organizations that try to deploy everything at once on a weak data foundation will struggle. The AI is real. The data work is what makes it valuable.