With over 45 operating companies and 170 locations, standardizing a single AI solution isn’t an overnight project. Lawrence and his FIX Center team are in what he calls the “very early stages” of deployment. The tactic has been a careful one: the Voice AI is being put head-to-head against human troubleshooters. “It’s become sort of a competition,” Lawrence admits, noting the energy this creates internally. The next phase is already in the works: a field pilot with a handpicked group of technicians. Any hurdles encountered, Lawrence frames as “learning opportunities” rather than proof against Voice AI. Previous success with the FIX Center hotline, where techs regularly call for help, adds confidence that adoption won’t be a foreign concept for most.

Teaching Mark: When Manufacturer Manuals Aren’t Enough

“Mark,” the name given to their AI, was trained first on an obvious source – manufacturer manuals and internal HVAC knowledge. This was deliberate: Lawrence points out that “we do this to avoid any hallucinations that could come from a third-party source that could be a troubleshooting ‘opinion’ vs. a manufacturer’s fact.”

Early results weren’t perfect. Mark at first offered “fragmented or incomplete solutions”, but as Comfort fed in HVAC training documents and basic concepts, the AI’s answers improved. “We noticed a significant improvement,” Lawrence says, and now the team is moving to bring decades of service history into the mix – a step they hope will let the system spot trends and suggest fixes based on nationwide data.

Field testing Mark brought its own surprises. “We were happy to see that the system rarely struggled with the vernacular or style of questioning,” Lawrence mentions. Still, any AI is only as good as its sources, and Mark would blank on questions outside what was included in the manuals or initial training docs. Their “fail fast” mentality means that whenever the AI can’t answer, the call is quickly routed to a live FIX Center agent – who now gets all the relevant context captured from the AI chat.

Buy-in from experienced technicians remains a question mark at this stage. With the company’s structure giving local autonomy to its operating companies, Comfort expects some natural pushback. The strategy is simple: share wins transparently, offer flexibility in how techs use the solution, and highlight progress to drive adoption.

Is It Working? It’s Still Early.

Hard data on success – metrics like call time reductions or customer satisfaction – isn’t there yet. “It’s too early to provide metrics,” Lawrence says. But the thinking is proactive: as the number of data centers and complex builds rises, Voice AI “allows us to have a proactive solution to limiting the learning curve of field technicians working on pieces of equipment that they may not be completely familiar with.” The aim is fewer phone calls, fewer escalations, improved technician experience.

IVRs and 'Crazy Ideas'

One of Comfort’s boldest projects is their IVR (interactive voice response) replacement. The payoff, according to Lawrence, isn’t just technological. “The process of gathering the information to map out this solution has provided us with the ability to think of how we can improve our processes and offerings,” he observes. The real confidence booster? Structured, consistent processes already in place, and a culture that encourages ambitious “crazy ideas” and trial-and-error learning.

Culture, Not Just Technology

At the heart of all these moves is a technician-first philosophy. “An improved experience at the field technician level will lead to a better experience at the service operations level, which will lead to a better end-user customer experience,” Lawrence emphasizes. Comfort aspires to a culture that skilled workers choose “not for an extra dollar or two an hour, but for a group that believes in them and is constantly seeking out ways to improve.”

Looking at AI’s next evolution, Lawrence quotes a recent conference speaker: “Because of AI, technology will grow faster in the next 5 years than it has in the last 50.” Comfort’s view is pragmatic: don’t rush, but don’t be left behind either. The focus is on “making sure we focus on the technologies coming out and find the best fit... before releasing them to our customers.”

Perspective from the AI Side: Aquant’s Indresh Satyanarayana on MEP Challenges

If Comfort represents the end user’s balancing act, Aquant’s Indresh Satyanarayana, VP of Product & Labs, sees the technical puzzle up close. The MEP contracting world is “anything but standardized”, he noted. Technicians jump between HVAC, electrical, plumbing, automation, across “dozens of manufacturers, hundreds of models, generations, and control systems.” That means Call Assist needs to “navigate this long-tail reality” – recognizing specific vocabulary, failure modes, and workflows from a bewildering array of sources.

Up-to-date recommendations are a balancing act. Aquant’s system pulls in new manuals and technical bulletins automatically; learns from real-world technician interactions (“it observes those conversations and outcomes”); and incorporates structured feedback from the field. “This gives us both manual oversight and automated learning,” Satyanarayana says.

Asked about what works – and what slows transformation – he’s unequivocal: real impact starts with focused, “high-impact voice-based workflows”, not with trying to hand over every service process to AI at once. Early, continuous human involvement is critical. “Their feedback shapes the product, accelerates adoption, and ensures the AI actually solves practical problems, not abstract ones.”

Companies most often trip up by “underestimating the importance of user adoption and change management”, Satyanarayana says. When techs aren’t brought in early or the implementation is overpromised, resistance is all but guaranteed. “By combining a strong focus on human factors with clear, pragmatic goal-setting, companies can avoid these common traps and maximize the impact of their Voice AI investments.”

The Road Ahead

For Comfort, the rollout is about more than cutting tech support call times or gleaning insights from data. It’s about who will be left to fix America's critical infrastructure as experienced techs age out – and what the next generation expects in support and empowerment. The tools might be cutting-edge, but the goal, as Lawrence puts it, is “to fix things as quickly as possible.” In the age of AI, that’s a mission just beginning.