Turning Data into Action with Irth Insights

AI Is Redefining Damage Prevention and Safety

Not all 811 tickets carry the same risk of damage. And not all damages are equal.

A minor scrape to a facility and a gas line strike that shuts down critical services may both be labeled as “damage,” but their real-world impact couldn’t be more different.

At the Irth User Summit, NiSource shared how it’s using Irth Insights to apply AI-driven impact modeling to millions of locate requests, helping teams focus on the incidents that matter most for safety and operational risk.

By combining deep historical data, predictive AI modeling, and automated intervention strategies, Irth Insights is enabling utilities like NiSource to turn insight into action, improving safety, efficiency, and decision-making at scale.

The Challenge: Millions of Tickets, Limited Resources

Utilities process millions of locate requests each year, many of them low risk. NiSource alone manages 2 million 811 tickets across gas and electric operations spanning six states. Even with strong damage prevention teams, it’s impossible to physically intervene on every ticket.

At the same time, everyone in the industry understands a hard truth:

• Some damages are minor
• Others are catastrophic

With millions of tickets, how do you identify the small percentage of tickets that carry the highest risk before something goes wrong?

Insights Drive Damage Prevention Actions  

NiSource’s journey began with Irth’s existing North American impact model, which already delivered measurable improvements for them:

• 2.9% improvement in identifying critical excavation safety events
• 2.6% improvement in predicting high-cost damage incidents

While safety was the main focus, cost was used to distinguish high-impact incidents and focus damage prevention efforts where they mattered most.

NiSource pairs AI-driven impact insights with automated and human-led interventions. Tickets are scored daily based on impact and risk, triggering thousands of highly targeted communications tailored to the person receiving them, whether an excavator or a homeowner. These automated emails provide timely best practices and reminders to help ensure safe digging before work begins.

For the highest-risk and highest-impact tickets, automation is only the first step. Damage prevention specialists proactively follow-up via phone calls or in-person visits, prioritizing locations with elevated consequences, such as hospitals and other critical facilities.

This approach is currently live in one operating region, delivering strong results and executive buy-in, and setting the stage for future expansion and additional intervention strategies.

These mitigation actions are tracked in Utilisphere via the platform’s integrated BI reporting.

Why This AI Model Succeeded

It’s often said that 95% of AI initiatives fail. This one didn’t and for a simple reason:

AI wasn’t treated as a “set it and forget it” tool.

Instead, it became part of an iterative loop where the team:

1. Analyzes results
2. Adjusts the model
3. Refines intervention strategies
4. Repeat

The result is a system that continuously learns and improves while producing real, operational results.

Why Data Quality was the Differentiator

Encouraged by the initial results, NiSource asked the next question:

How can we make this model even better and safer?

They understood that effective AI starts with quality data. NiSource provided Irth access to nearly 10 years of consistently collected incident and consequence data, including repair costs and safety outcomes.

As a result, Irth was able to train a bespoke model tailored specifically to NiSource’s environment. That historical depth allowed the AI to detect patterns that would otherwise remain hidden, especially when it comes to rare but critical events.

Financial Impact: Finding High-Cost Incidents Before They Occur

When NiSource’s incident data was analyzed, 4.9% of incidents were “high-cost” from a monetary perspective. Average costs were around $2,500, while the max was close to $500,000. Additionally,

• 15% of high-cost incidents were in the top 5% of locate requests, representing 9% of total financial exposure
• 42% were identified within the top 20%, representing 28% of financial exposure

This proved the hypothesis that it was possible to predict high-cost incidents before they occur.

Safety Impact: Predicting What Truly Matters

Financial impact is important but safety is paramount. Out of more than 15,000 incidents, about 2.2% involved:

• Evacuations
• Property damage
• Injury or loss of life

While these are rare events, they are the most critical.

As a result of NiSource’s comprehensive data, Irth Insights was able to predict almost half of these safety-critical incidents in the top 20% of locate requests.

When even one serious incident can change lives, that level of foresight is invaluable. Now that the bespoke model has been trained, NiSource’s next step is to roll it and the mitigation automations to its entire operations footprint, monitor mitigation tasks, iterate, adjust, and measure success.

The Takeaway

As demonstrated by NiSource at the Irth User Summit, AI-driven impact modeling helps damage prevention professionals focus on the right work at the right time. By combining AI impact modeling, high-quality data, and automated workflows, Irth Insights enables organizations to:

• Reduce damages
• Improve safety outcomes
• Increase operational efficiency

Most importantly, it transforms data into action where it can make the greatest difference.

Want to see how AI-driven impact modeling can help your organization prevent high-risk excavation incidents before they occur?