Automating Dispatch: How AI Forecasting Reduces Overtime | Domo BUILD 2026

Matt: Our final use case is presented by Mark Mullane. Mark is the Director of Marketing at TR Miller, and he's going to show one of our most common use cases: how he is using AI to forecast. Now, what's interesting about this is that we previously talked about the 500 Apps program, where we received over 150 requests from you at Domopalooza to build new applications. The number one use case requested was using AI to forecast, and that is exactly what Mark did. He is applying that AI forecasting to automate his company's dispatch process. So with that, Mark, you are our very last use case for the day. We give you the stage.

Mark Mullane: Thank you so much, Matt. I'm happy to be here. As Matt introduced, this project came out of the 500 Apps program with Domo. Our focus was really on how we can better prepare our dispatch system and forecast for the future.

To give you a quick introduction, my name is Mark Mullane, and I'm the Director of Marketing over at TR Miller Heating, Cooling, Plumbing, and Electrical. We are a residential home service company operating in the Chicago area. We really wanted to focus on finding an app that would help us solve some of the persistent challenges that have existed within our industry since time immemorial—namely, being able to accurately predict the future and the weather coming down the road.

While weather stations are relatively good at predicting the weather a week out, what we couldn't always get right was the specific impact that weather would have on our business. We were using old-school, manual methods to try and track that relationship.

In the blue-collar trades, there is a bit of a stereotype that we are low-tech or not particularly cutting-edge. I want to dispel that myth today. I do understand where the stereotype comes from; on my second day working at TR Miller seven years ago, I spent my time helping a team member log into their Apple account to set up their new phone. That just came with the territory.

However, what I quickly realized was that even if our team's daily usage of technology wasn't at a high IT level, they were still capturing a massive amount of high-quality data within our CRM and operating systems. There was simply a gap between capturing that data and actually utilizing it.

The company had already started utilizing data in some highly effective ways. For example, on our dispatch boards, we look at how we line up technicians for the day with smart routing. We analyze how we can get them from one job to the next with the least amount of "windshield time"—which is what we call the hours spent driving around instead of helping a customer. We also actively track inventory, measure technician productivity, and monitor our conversions, average tickets, and sales goals.

Yet, the one thing we hadn't fully figured out was how call volume swung so drastically with the weather. Whether it's a heatwave, extreme cold, or freeze breaks, every area of our business—plumbing, HVAC, and electrical—is heavily impacted by outdoor temperatures. We didn't know how to mathematically model and forecast those shifts to prepare for tomorrow.

This unpredictability cascades down the entire business. It impacts our pre-booking, causes sudden spikes and crashes in maintenance requests, and leads to staffing whiplash for our technicians. We care deeply about our technicians' work-life balance to make sure they are happy in their roles. The last thing we want to do is call someone on a Saturday morning asking them to come to work because we didn't prepare. That's a bad experience for them, which leads to a bad experience for our customers. Unpredictability also forces marketing to play catch-up rather than getting ahead of the curve to know where to push or pull back budget.

To solve this, we built an AI app to forecast demand before it hits so our team can plan rather than simply react. Looking at our Domo app, we have the upcoming week forecasted based on our historical data. We loaded eight years of data showing how many repair calls we ran on each calendar day and connected it to weather APIs to pull the historical weather for those exact days. We used this data to accurately map the direct impact of weather on a day-to-day basis.

We then use these insights to look at the weeks ahead. For example, during June in Chicago, the weather was surprisingly cold, staying in the 70s and even dropping into the 60s. This meant air conditioners weren't running, and if they aren't running, they aren't breaking. Using our forecasting tool, we saw that emergency repair demand was going to drop. Instead of leaving our technicians without work, we loaded up our schedule with preventative maintenance inspection calls.

By performing these inspections ahead of time, we took care of our customers before the summer heat wave arrived. This kept our technicians on the road, secured their working hours, kept them busy, and continued to drive revenue and positive business outcomes. When the heat wave finally arrived, those inspections were already done, and we could focus entirely on incoming emergency repair demand.

Weather is our major demand signal, and our extensive historical database allows us to generate highly accurate projections. The AI model accounts for factors we intuitively understood from years in the trade but could never mathematically model. For example, 80 degrees is not just 80 degrees; the business impact is completely different depending on the time of year it hits. An 80-degree day in May represents the first time homeowners turn on their air conditioning, causing a massive spike in start-up breakdowns. However, an 80-degree day in August usually has almost no impact on call volume. The AI model also accounts for weekends and holidays, which heavily influence how likely people are to call.

Ultimately, this data informs our staffing requirements and recommendations. If we see that a Saturday is going to be hot, we can project the expected call volume and schedule the necessary technicians on Monday. This allows us to have proactive conversations early in the week so our team can plan their personal lives, rather than calling them last-minute on Friday night.

The results of this improved forecasting have been incredible for our team members. One of our most important KPIs is technician retention, as experienced techs are extremely rare and valuable. In our HVAC service department, we currently have an active retention rate of over 80%. A huge portion of that success is due to improved work-life balance—keeping them busy when it's slow and avoiding overworking them when it is busy.

It has also had a major financial impact. The biggest contributor to managing our repair calls is our cancellation rate. A high cancellation rate indicates we were ill-equipped to handle incoming demand, whereas a low rate shows we were prepared. Last June, we had an 18% cancellation rate during a very hot, spiky month where we weren't prepared for the demand surges. This year, using our forecasting app, we brought that cancellation rate down to 14%. That 4% drop represents dozens of additional completed service calls and hundreds of thousands of dollars in captured revenue.

I am incredibly excited to take this app forward, expand it into our other trades, and track how we can apply these models to rainfall in plumbing or seasonal peaks in electrical. This will allow us to be even better prepared and set up for the future. Thank you, everyone, it has been a pleasure sharing what we've built.