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

Two years ago, AI in Maximo was a story about future potential. IBM talked about generative AI assistants, computer vision, and predictive analytics, but the production deployments were small, the features were limited, and most customers were still trying to figure out whether the underlying data was good enough to feed any kind of model. In 2026, the conversation has changed. Maximo Application Suite 9.x ships with AI capabilities that are operationally grounded, integrated into the daily workflow, and proven in production at a scale that would have seemed ambitious in 2024.

The release of Maximo Application Suite 9.2 in June 2026 is the most consequential step in that transition. IBM has embedded AI more deeply into the daily work of managing assets, teams, safety, and operations. The release expands AI across reliability insights, field execution, safety and compliance workflows, document-based information extraction, and orchestration across systems, and it introduces agentic workflows designed to help guide decisions and move work forward in operationally grounded ways. The phrase "operationally grounded" matters. The 9.2 AI is not a chatbot bolted onto the side of the product. It is a set of capabilities that are wired into the work order, the inspection, the asset record, and the reliability workflow.

This article walks through the AI capabilities in Maximo in 2026: Condition Insight, Visual Inspection, Maximo Assistant, the watsonx foundation, and the agentic shift that defines the 9.2 release. We close with practical implications for teams planning an AI roadmap and a bottom line on what the AI actually does in production.

Maximo Condition Insight: Prescriptive Maintenance at Last

The single most important AI capability in MAS 9.x is Maximo Condition Insight, which IBM introduced in late 2025 and which has matured into a flagship feature by mid-2026. Condition Insight is an agentic AI capability within Maximo Asset Performance Management (APM) that interprets asset data to explain asset condition, highlight emerging trends, and recommend corrective actions. It works in concert with the other MAS applications to deliver a unified, condition-based maintenance approach across the Maximo ecosystem.

The traditional challenge with condition-based maintenance is that it requires deep specialist expertise. A reliability engineer has to look at work order history, meter readings, time-series sensor data, failure mode and effects analysis (FMEA), and alert patterns, and synthesize all of that into a recommendation. The skill is rare, the time required is significant, and the output is not easily auditable. Condition Insight removes that barrier by analyzing asset data in seconds and returning a clear, explainable summary of condition, trends, and recommended actions. The result is condition-based maintenance that is practical for every maintenance team, not just the ones with a reliability engineer on staff.

The technical foundation is watsonx, IBM's enterprise AI platform, which provides the large language model and the orchestration layer. Condition Insight is not a black box. Every recommendation includes the underlying data, the trend analysis, the relevant work order history, and the FMEA context that drove the suggestion. The auditability is essential for regulated industries, where a recommendation to defer maintenance or to replace a component has to be defensible to a regulator.

A typical Condition Insight interaction in MAS 9.2 looks like this. A maintenance engineer opens an asset record for a critical pump and sees a Condition Insight panel at the top of the screen. The panel shows a current condition score, a trend line over the last 90 days, and a plain-language summary: "Vibration trends on bearing 2 have been increasing for 14 days, and the last three work orders on this asset involved seal leaks. Based on similar asset history, the probability of a bearing failure in the next 30 days is elevated. Recommended actions: inspect bearing 2 within 7 days, order replacement seal kit, and review PM frequency." The engineer can drill into the data, accept the recommendation (which creates a work order), or override it with a reason. The override is captured for future model improvement.

The practical impact is significant. Teams that previously had a single reliability engineer reviewing all critical assets can now review ten times as many assets in the same time, because the engineer is reviewing and validating AI recommendations rather than generating them from scratch. Teams that did not have a reliability engineer at all can now make condition-based decisions that were previously the exclusive province of specialists. The democratization of reliability is the most important practical consequence of Condition Insight, and it is the reason the feature is at the center of the MAS 9.x value proposition.

Maximo Visual Inspection: Computer Vision for Asset Health

Visual inspection is the second major AI capability in MAS 9.x, and it is the one that has improved the most in the last 12 months. Maximo Visual Inspection uses computer vision models to detect defects, corrosion, hotspots, and other visual indicators of asset condition from photos or video. The original Visual Inspection capability required custom model training for each use case, which made it expensive to deploy. The 9.x release introduces a low-code "Visual Prompting" capability that lets subject matter experts train models by highlighting examples in a few images, without writing code or working with labeled datasets.

The Visual Prompting workflow is the most disruptive change. A maintenance engineer with no machine learning background can train a model in an afternoon by uploading 20 to 30 images of a particular defect (corrosion on a pipe, a hotspot on a transformer, a crack in a tank weld), drawing a box around the defect in each image, and giving the model a name. The model trains in the background and is available for inference within hours. The same engineer can then use the model from a mobile device, pointing the camera at the asset and getting a real-time assessment of whether the defect is present. The inference can run on the device, on the edge, or in the cloud, depending on the connectivity and the latency requirements.

The 9.2 release adds local inference directly on the mobile device, which is a major improvement for field use. The previous cloud-only model required connectivity at the moment of inspection, which was a problem in substations, underground vaults, and remote sites. Local inference also reduces the cost of high-volume inspections, because the model runs on hardware the technician already carries rather than on cloud GPU time.

The production use cases for Visual Inspection are diverse. Utilities use it to inspect transmission towers, distribution poles, and substation equipment. Oil and gas operators use it to inspect pipelines, storage tanks, and flare stacks. Manufacturers use it on the production line to detect product defects and equipment wear. The common pattern is that visual inspection is a high-volume, repetitive task that is hard to do consistently with human inspectors and that produces large amounts of unstructured data that is hard to capture in a traditional CMMS. Visual Inspection turns the camera into a sensor, and it turns the inspection record into structured data that flows back into Maximo automatically.

A practical example illustrates the value. A utility was inspecting 10,000 distribution poles per year using a contracted inspection team. Each inspection took about 30 minutes, and the resulting report was a mix of photos and free-text notes that had to be manually transcribed into Maximo. With Visual Inspection, the same team now uses a mobile app that captures the photo, runs the model, and creates a Maximo inspection record with the defect type, severity, and location automatically populated. The inspection time dropped to 10 minutes, the data quality improved dramatically, and the team can now inspect 15,000 poles per year with the same headcount. The annual savings pay for the Visual Inspection deployment in less than 18 months.

Maximo Assistant: Natural Language for Asset Data

Maximo Assistant is the Gen AI capability that has been part of MAS since the 9.0 release, and it is the feature that most users interact with first. The assistant is a conversational interface that lets users ask questions in natural language and get answers from their Maximo data. It is built on watsonx.ai, which provides the underlying large language model, and it is integrated into the Maximo UI as a chat panel that follows the user through the application.

The typical use cases are the kind of questions that a maintenance manager or planner would otherwise have to look up manually: "Which work orders are missing job plans?", "Show me the sum of the total cost of work orders per site for the last quarter", "Which assets have had more than three corrective work orders in the last 30 days?". The assistant translates the question into a Maximo query, runs it, and returns a formatted answer with a chart or table where appropriate. The user can drill into the underlying records, save the query for later, or share it with a colleague.

The 9.2 release extends the assistant in three important ways. The first is Maximo Assistant on Mobile, which lets technicians use natural language to find asset information, review history, and complete work efficiently in the field. The second is AI-enabled conversational scheduling, which lets planners and schedulers explore changes such as increasing capacity or prioritizing critical work using plain language, with the assistant suggesting schedule changes based on real-time conditions, constraints, and resource availability. The third is what-if analysis, which lets managers explore scenarios like "What if I add a third crew to the Bedford site?" and see the projected impact on backlog, overtime, and cost.

The practical pattern that emerges from production deployments is that the assistant is most valuable for two specific user groups. The first is the executive who needs a quick answer without learning the Maximo query language. The second is the new hire who needs to find their way around an unfamiliar system. Both groups can get value from the assistant within minutes, which is a much shorter time to value than the typical Maximo training program. The assistant is not a replacement for Maximo expertise, but it is a genuine productivity multiplier for the people who occasionally need information from Maximo but do not live in the system every day.

The governance story is important. The assistant runs inside the Maximo security context, which means it can only see the data that the user is authorized to see. The queries it generates are logged and auditable, and the data it returns is subject to the same row-level security as any other Maximo query. This is a significant improvement over the typical "chat with your data" experience, where the security model is often an afterthought. The Maximo Assistant approach is the right template for enterprise AI: powerful for the user, governed by the same controls as the rest of the system.

Agentic Workflows and the 9.2 Release

The most important conceptual change in MAS 9.2 is the shift from AI-as-feature to AI-as-workflow. The previous releases embedded AI capabilities into specific screens and specific decisions. The 9.2 release introduces agentic workflows that span multiple applications and multiple decisions, with the AI acting as a coordinator that moves work forward across the system.

The canonical example is the condition-based maintenance workflow. In the old model, a reliability engineer would review asset data, identify a developing issue, write a recommendation, and hand it to a planner. The planner would review the recommendation, check resource availability, create a work order, and assign it to a crew. The crew would execute the work and close the loop. In the agentic model, the AI monitors the asset continuously, identifies the issue, generates the recommendation, checks resource availability, drafts the work order, and routes it to a human planner for review and approval. The planner becomes a reviewer rather than a creator, which is a meaningful shift in how the work gets done.

The agentic pattern is not unique to condition-based maintenance. The 9.2 release applies it to safety and compliance workflows, where the AI reviews inspection records, identifies gaps, and drafts the corrective actions. It applies it to field execution, where the AI routes work to technicians based on location, skill, and current workload. It applies to document processing, where the AI extracts structured data from P&IDs, equipment lists, and DCS I/O files, and creates the corresponding asset records in Maximo automatically. In each case, the AI is doing the routine work, and the human is doing the exception handling and the approval.

The honest assessment is that the agentic capabilities in 9.2 are still early. The workflows are well-designed, the integration is deep, and the underlying AI is powerful, but the production deployments are limited and the edge cases are still being discovered. Organizations adopting agentic workflows in 9.2 should plan for a learning curve, with a phased rollout that starts with the lowest-risk workflows and expands as the team gains confidence. The upside is significant, but the path to that upside is iterative rather than instant.

The watsonx Foundation and the Data Imperative

Every AI capability in MAS 9.x rests on the same foundation: watsonx, IBM's enterprise AI platform. The watsonx.ai component provides the large language models, the watsonx.data component provides the data lake, and the watsonx.governance component provides the tools to manage the model lifecycle, the bias detection, and the regulatory compliance. The Maximo team did not build their own foundation models. They built the application layer that uses the foundation models, which is the right architectural choice for a domain-specific product.

The data layer is where the practical reality of AI in Maximo becomes clear. Every AI feature in MAS 9.x depends on the quality of the underlying Maximo data. Condition Insight depends on accurate work order history, meter readings, and FMEA records. Visual Inspection depends on labeled image data and the asset hierarchy that the images map to. Maximo Assistant depends on the same data that any other Maximo query would use. The AI is not a substitute for data quality. The AI is a multiplier on top of data quality, and it amplifies both the good and the bad.

The teams that get the most value from AI in Maximo are the ones that have invested in data quality for years. The teams that are disappointed by AI in Maximo are typically the ones that are discovering, in real time, the data quality problems that they have been deferring for a decade. The AI features do not create new data quality requirements, but they expose existing data quality problems much more visibly than the previous generation of features. Investing in data quality is the single most important pre-requisite for any organization planning an AI roadmap in Maximo.

Practical Implications

The AI capabilities in MAS 9.x are real, mature, and ready for production. The most important practical decision is which capability to start with, because each one has different data prerequisites and different organizational impacts. Condition Insight is the best starting point for organizations with a reliability engineering function and a data foundation that includes time-series and FMEA data. Visual Inspection is the best starting point for organizations with a high volume of visual inspection work and a willingness to invest in image data. Maximo Assistant is the best starting point for organizations that want to drive broad adoption quickly and that have users who need occasional access to Maximo data rather than daily access.

The second practical decision is about change management. AI changes the nature of the work, and the people whose work is changing need to be involved in the design of the new workflow. The teams that have done this well have set up cross-functional working groups that include reliability engineers, planners, technicians, and data scientists. The teams that have done this poorly have handed the AI roadmap to the IT department, which has produced technically sound deployments that the operations team has rejected. The technology is the easy part. The organizational change is the hard part.

The third practical decision is about vendor lock-in. The AI features in MAS 9.x are tightly integrated with watsonx, which is IBM's platform. Organizations that have standardized on a different AI platform, such as Azure OpenAI or AWS Bedrock, will need to evaluate whether the tight integration is a feature or a constraint. For most Maximo customers, the integration is a feature, because it removes the complexity of managing multiple AI platforms. For organizations with a strong existing AI platform strategy, it is worth understanding the boundaries of the integration and whether the underlying models can be swapped.

The fourth practical decision is about cost. AI features are not free, and the cost model for Condition Insight, Visual Inspection, and Maximo Assistant varies. The most common cost drivers are the number of assets being monitored, the volume of inspections being processed, and the number of assistant queries being made. Organizations should plan for these costs explicitly in their AI roadmap, and they should expect the cost to scale with usage. The good news is that the value typically scales faster than the cost, especially for organizations that have been deferring reliability work because of specialist constraints.

Bottom Line

IBM has embedded watsonx into the heart of Maximo Application Suite 9.x, and the AI features have moved from experimental to operational. Condition Insight brings prescriptive maintenance to every team, not just the ones with reliability specialists. Visual Inspection turns the camera into a sensor and the inspection record into structured data. Maximo Assistant puts natural language on top of the asset data for the people who occasionally need it. The 9.2 release adds agentic workflows that span the system and move work forward across applications. The technology is real, the integration is deep, and the value is documented. The teams that get the most value are the ones that invest in data quality, involve the operations team in the design, and plan the AI roadmap as a multi-year program rather than a single deployment.