3 min read
If your team is managing locate tickets by the hundreds of thousands — or even millions — you already know the math doesn’t work. There aren’t enough hours in the day to manually review every photo attached to every ticket. And yet, accuracy isn’t optional. Every missed detail carries real-world consequences, not just compliance issues, but safety risks and costly damages.
That’s the challenge Irth set out to solve with Irth Insights — a suite of AI models built directly into Utilisphere to advance damage prevention. By automating photo validation, flagging compliance issues in real time, and helping field teams catch mistakes before they become damages, Irth Insights enables proactive damage prevention programs.
Locate Photo Ticket Audits: The Right Photo, The Right Place, Every Time
Traditionally, auditing locate tickets has been a sampling exercise. There’s simply no practical way to manually review every photo attached to every ticket at scale.
Irth Insights changes that.
At the heart of the locate photo audit model is a deceptively simple question: Was this photo taken at the right place, with the right markings?
Using AI, the system validates:
• Paint and flag colors to confirm proper marks
• Location accuracy to ensure photos were taken within the accepted distance of the dig site
The pass/fail criteria are fully configurable. Companies can set their own thresholds for how far from the dig site a photo can be taken, what percentage of certainty counts as a valid attachment, and how many valid attachments a ticket requires before it can move forward. If a locator submits photos that don’t meet the threshold, the system can prevent them from completing their response entirely — catching the error at the time of the locate, not weeks later during an audit or after a damage has already occurred.
The system also flags duplicate photos — cases where the same image is reused across tickets — which helps surface patterns of non-compliance that would be nearly impossible to catch through manual review.
Currently, the models support yellow and orange flag detection, with more colors in development. Training is built on real customer photos using Microsoft Custom Vision, and the models are regularly retrained as new data comes in. The more photos, the better — a minimum of 100 passing and 100 failing images is recommended, though accuracy continues to improve with more data.
OCR: Adding a Layer of Location Intelligence
Color validation is only part of the equation. Irth Insights also applies optical character recognition (OCR) to extract GPS coordinates, timestamps, and other metadata directly from images.
By cross-referencing photo metadata with the actual dig site location, the system confirms that work happened where it was supposed to happen. If a photo falls outside a configurable buffer zone, the ticket is automatically flagged.
This works regardless of whether teams use the Irth mobile app or third-party photo tools — the system reads the same underlying data.
Some organizations have configured alerts that notify locators immediately if their photos fall outside the dig area — and prevent them from completing the ticket until the issue is corrected. What used to be discovered after the fact is now addressed in the moment.
Meter Inspections: AI on the Front Lines of PHMSA Compliance
The same photo AI approach extends to meter inspections. Field technicians take photos of the meter during their inspection, and the model automatically classifies the condition — detecting corrosion, rust, buried meters, and vegetation interference without requiring any additional input from the user.
Each condition is assigned a confidence score, enabling organizations to define thresholds that:
• Trigger corrective actions
• Automatically close compliant inspections
• Schedule follow-up reviews for emerging issues
The forms included out of the box are the meter inspection and corrective action form, though these can be customized or added to existing workflows.
This creates a more consistent and scalable approach to compliance and PHMSA regulations.
For utilities navigating PHMSA regulations, this kind of automated documentation provides both a compliance record and a proactive maintenance signal — catching vegetation overgrowth or early corrosion before it becomes a regulatory finding or a field emergency.
From Automation to Operational Intelligence
AI handles validation at scale, applying consistent standards across thousands — or millions — of decisions. Humans step in where judgment is required.
The result is a move away from reactive auditing toward proactive risk management.
And this model doesn’t stop at locate tickets or meter inspections. As workflows evolve, the same approach can extend into adjacent areas like leak surveys — where automated systems cover large areas, and AI-driven validation ensures accuracy at the edges.
Across use cases, the pattern is the same:
• Automation manages volume
• AI enforces consistency
• People focus on exceptions and decision-making
Built Into the Work You Already Do
For teams already using Utilisphere, these capabilities are built into the system they’re already working in. No new tool to learn. Just smarter data, surfaced at the right moment.
See how teams are applying this in the field. Watch the full session from our annual User Summit.
Integrating Geospatial Intelligence for Proactive Geohazard Management
Understanding the interaction between pipelines and their surrounding environment is essential for effective integrity management. Slope instability,...
Using Machine Learning to Manage Data Gaps and Improve Decision-Making
Machine learning is often known for building prediction models using training data, i.e., identifying your susceptibility to an internal corrosion...