How AI Is Making Visual Condition Assessments Faster, Safer, and More Scalable

By Antoine Adderley, P.Eng., Project Professional – Asset Management

Condition assessments have always been essential to good asset management. Municipalities and utilities know these assessments help understand asset health, identify risks, and enable us to make better decisions about maintenance, renewal, and capital planning.

But, in practice, these assessments come with familiar challenges. Visual inspections are labour intensive, often require multiple disciplines on site, and can be difficult to coordinate around outages, access limitations, and safety constraints. Even after field work is complete, reporting can still take significant time.

But visual condition assessments are at an important turning point. Advances in artificial intelligence, particularly computer vision combined with large language models, are enabling a different kind of assessment workflow. Not one that replaces engineering judgment, but one that helps teams document conditions more consistently, process information more quickly, and focus effort where it adds the most value.

The question is no longer whether digital tools belong in the process. It is: what will visual condition assessments look like in the near future and how can organizations prepare now?

From Manual Documentation to Structured, Photo-Based Intelligence

Traditional assessments follow a standard sequence: Inspect the asset, take notes, organize photos, assign condition ratings, and draft comments for reporting.

That process works, but it is slow and highly dependent on individual documentation style.

Where the industry is heading is a more structured model. In this new model, site photos are no longer just records, they are inputs into a workflow where they are uploaded, linked to specific assets, and analyzed to detect visible issues such as corrosion, cracking, or wear. From there, the photos can help teams:

• Identify assets
• Detect visible deficiencies
• Assign draft condition ratings
• Generate structured technical commentary

The real shift is consistency. When condition language, rating logic, and documentation structure become standardized, organizations gain repeatability across sites, reviewers, and inspection cycles. That consistency ultimately improves how condition data is used—not just how it is collected.

Leaner Field Teams, Smarter Use of Expertise

One of the most immediate impacts of AI-assisted workflows is efficiency in the field.

This is not about doing less work, but rather, reallocating effort.

Instead of relying on larger teams for first-pass observation and documentation, organizations can shift toward:

• Exception-based review
• Technical validation
• Decision-making

This results in experienced staff spending less time capturing information and more time interpreting it. That matters in an environment where many municipalities are already operating with limited internal resources.

The early takeaway is straightforward: your field effort can be reduced without lowering the technical standard (provided that structured review remains in place).

Improved Safety and Access Flexibility

This shift is particularly relevant in water and wastewater environments.

Facilities such as treatment plants, pump stations, wet wells, and MCC rooms introduce real constraints around access and safety. Often, the complexity of an inspection is driven as much by logistics as by the assessment itself.

AI-assisted workflows do not remove these constraints, but they allow teams to navigate them more efficiently. Here’s an immediate and practical example:

With fewer personnel required on site and greater flexibility to use mobile or remote capture methods–such as pole cameras or drones where appropriate– organizations can reduce their exposure and simplify their planning.

Human Expertise Is Non-Negotiable

AI is never a replacement for engineering judgment.

Visual evidence has limits. A photograph can show corrosion, cracking, or deterioration but it cannot always explain cause, internal condition, or performance impact.

That is why the credible model is human-in-the-loop review.

In this model:

• AI performs the initial organization and interpretation
• Qualified professionals validate observations
• Engineers confirm condition ratings and recommendations

The result is not reduced oversight but more focused oversight. Professionals spend less time on repetitive documentation and more time on decisions that require judgment.

From Inspection Outputs to Asset Strategy

Faster reporting is useful, but it is not the primary value.

The larger opportunity is what happens after the inspection.

When condition data is structured, photo-linked, and tied to asset records, it becomes easier to translate field observations into:

• Remaining useful life estimates
• Priority rankings
• Replacement timing
• Multi-year capital plans

This creates a direct pathway from inspection to decision-making.

Figure 1 – AI-Assisted Condition Assessment Workflow

Figure 1 illustrates how structured visual data flows from field capture through AI-assisted analysis and into capital planning outputs.

For municipalities managing large portfolios, this shift matters. The goal is not just to inspect faster, but to generate condition data that is more usable, scalable, and defensible.

That is where the long-term value sits.

A Proof of Concept—And What It Signals

GEI recently explored this approach through a proof of concept combining computer vision, large language models, and human QA/QC within a multidisciplinary assessment workflow.

While the pilot was completed at a municipal facility, the model is transferable across water and wastewater systems.

The results suggested:

• Field staffing could be reduced by approximately 50–60%
• Reporting timelines could be shortened
• Technical oversight could be maintained through structured review

The proof of concept is not the conclusion but an early indicator of what is becoming possible.

Looking Ahead

Visual condition assessments are not being replaced. They are being reshaped.

The near future will be defined by hybrid workflows that combine structured data capture, AI-assisted analysis, and professional judgment.

Organizations that begin exploring these approaches now will be better positioned to improve efficiency, enhance safety, and strengthen asset management decision-making.

The tools are still evolving, and the workflows are still being refined. But the direction is clear.

Visual condition assessments are moving toward a model that is faster, safer, and more tightly integrated with how infrastructure decisions are made.

And in many ways, this is just the beginning.

Learn More

We’re still in the early stages of AI-driven visual condition assessments and the pace of change is accelerating.

I’ll be sharing what we’ve learned, what’s already possible, and what’s coming next at the upcoming WEAO Technical Symposium. Find me there to continue the conversation.