Episode Overview
Three office staff. Sixty-five electricians. One voice agent named Sam.
That's the setup at ARC Electric, a Des Moines-based electrical contractor that's grown from one guy and a Suburban to a multi-state operation in just a few years. In this episode of the Mastering ServiceTitan podcast, host Josh Lu sits down with Liz Larson, Office & Service Coordinator at ARC Electric, to walk through how her team implemented AI Voice Agent — what the setup actually required, how they configured it for their specific call mix, and what it's like to run it live with a lean office team.
The short version: it's like adding a fourth person to the office.
Read on for the podcast highlights.
Life Before AI Voice Agent: Inconsistent Messaging and a Live Agent That Couldn't Keep Up
Before AI Voice Agent, ARC Electric relied on a third-party live answering service alongside their internal Dialpad setup. With three people juggling everything from scheduling to inspections to field coordination, any gap in coverage meant calls slipped. But the bigger issue wasn't volume — it was control.
"We just had a lot less control over how our message was sent out to customers when they were calling in," Larson explains. The live agents were answering for multiple companies at once, and the consistency ARC needed — especially across calls ranging from routine estimates to emotionally charged situations like post-house-fire inspections — was hard to maintain.
Their previous service worked for five years. It just stopped being the right fit. When AI Voice Agent came up through their account rep, the team was already intrigued by the concept and had been approached by a third-party voice agent vendor. The ServiceTitan integration made the decision straightforward.
Setting Up AI Voice Agent: Adaptive Capacity, Scripts, and Intentional Constraints
Larson is upfront: the setup took weeks, and she wouldn't have it any other way. The front-end work is what makes the agent effective once it's live.
A key prerequisite — and the part that required the most mental energy — was configuring Adaptive Capacity. ARC Electric runs four technicians through ServiceTitan for their residential service work, and the range of job types made booking rules essential. An estimate and a full installation aren't interchangeable, and the agent needed clear guardrails to reflect that.
The decision Larson landed on: Sam books estimates only. Every estimate slot is two hours — 8–10, 10–12, 12–2, 2–4. Anything beyond that — installs, larger scopes, jobs that might take six hours or two days — the agent takes a message and the team follows up.
"There's no way for us to differentiate when booking those install appointments versus estimate how much time they need because it's not a cut-and-dry situation," Larson says.
For the agent's knowledge base, Larson built a simple Excel sheet with two columns:
Question (e.g., "Can you come install a light fixture tomorrow?")
Answer (framed as a solution, not a no — "I'd be happy to set up an estimate")
That document gets uploaded directly into the voice agent settings, giving Sam a library of responses she can draw from on live calls. The goal wasn't just accuracy — it was tone. ARC Electric wanted every caller to feel like they were talking to someone who actually understood their situation.
On escalation, Larson took a different path than the default. Rather than routing unanswered escalations back to a live agent — which would create a loop for a three-person office already on the other line — the agent offers a callback instead. The team gets an email notification, follows up with context, and the caller gets a real person who already knows why they called.
Going Live and Early Results
Once the setup was complete, Larson didn't soft-launch. She emailed their implementation contact, got on a Zoom call, and was live within 15 minutes.
She used the free trial month deliberately, intentionally missing some calls to stress-test the system — identifying gaps in the Q&A, tuning call length settings, and observing how Sam handled different scenarios before optimizing. That context matters for reading the numbers.
From February 1 through the day of recording, ARC Electric's AI Voice Agent booking rate sat at 20% across 370 calls. That figure reflects their specific setup: Sam only books estimates, and a meaningful portion of ARC's inbound volume isn't bookable — bill pay calls, people trying to reach their utility provider, commercial project inquiries. Larson estimates roughly 20% of their calls are actually estimate requests, making that booking rate track closely with how she'd perform on the same call mix herself.
"Before, they weren't necessarily booking anything. They were just kind of taking a message," she says. "So we're extremely happy with that number."
The reporting side has also become part of the daily workflow. Inside Voice Agent settings, the team can review every call — classified as lead, booked, or unbooked — listen to recordings, read transcripts, and reclassify calls as needed. A reschedule call, for example, isn't a lead, but it's not unimportant either. Being able to associate it with an existing job keeps the data clean and the numbers meaningful.
One feature that took some back-and-forth to get right: email notifications for missed calls. It wasn't available at launch, and the ARC team flagged it as something they genuinely needed. ServiceTitan's team moved quickly — the feature shipped not long after. "To take that feedback and make it a reality so quickly," Larson says, "it just makes you want to dive into it and use it even more."
Advice for Teams Implementing AI Voice Agent
Larson's guidance for anyone getting started is practical and earned.
Don't put a hard deadline on yourself. The setup requires real time — understanding Adaptive Capacity, building out your Q&A, getting the settings right — and rushing it creates problems you'll be cleaning up after launch. Watch the webinars. Review the support materials before touching the settings. Get into the groove of what's needed before you start clicking around.
"I can't stress enough how happy I am with AI Voice Agent and how user-friendly it is compared to programs we were using previously," Larson says. "The lift in the beginning is quite a bit of work, but once you're ready to go live, it's smooth sailing."
Her parting advice: when you do launch, go all in. "I don't think there's an easy way to take it slow when you're introducing a voice agent. Really embrace it and work hard to make it work for you quickly, because it is such an asset." The teams that work through the early hiccups fast are the ones that end up with something that runs reliably — and feels like another person on the team.
Liz Larson recently joined Josh Lu on the "Mastering ServiceTitan" podcast to discuss:
[3:53] Life before AI and the challenges of handling calls
[13:47] Adaptive capacity and defining what AI can book
[23:43] ARC's AI experience so far and advice for implementing AI successfully
Check out these resources mentioned during the podcast:
You can find this interview and many more by subscribing to Mastering ServiceTitan on Apple Podcasts, on Spotify, or here.
Related Videos
About ServiceTitan
ServiceTitan is a comprehensive software solution built specifically to help home service companies streamline their operations, boost revenue, and substantially elevate the trajectory of their business.
Our comprehensive, cloud-based platform is used by thousands of electrical, HVAC, plumbing, garage door, and chimney sweep shops across the country—and has increased their revenue by an average of 25% in just their first year with us.
Ready to learn more about what ServiceTitan can do for your business?
