AI for HVAC Contractors: Where to Start, What to Avoid, and What Works

ACHR NEWS: AI is a pretty broad spectrum — what are some of its specific uses for a contracting business? 

Sponsler: The variation in AI systems really is as plentiful as the applications. My favorite move is simple: Ask AI to act like a CFO and review a templated P&L against trade benchmarks so it flags what’s drifting — labor, materials, advertising, office salary — and drafts first-pass fixes. That turns into price book updates, mix changes, and truck-level profitability checks instead of vague “do better” goals. On the people side, I’ll run call transcripts against a clean rubric, so CSRs and sales get targeted coaching without anyone wearing out a pair of headphones. The scoreboard it affects is the one owners already track: booking rate, membership mentions, average ticket, and gross margin. Same idea in operations — capacity planning and dispatch nudges that improve on-time starts and reduce returns. This works with the stack they already use (ServiceTitan, Housecall Pro, FieldEdge, CallRail, whatever records the calls), so we’re not asking for a new platform. It’s just less drudgery, faster decisions, and a tighter loop between data and action. 

ACHR NEWS: While AI is powerful, it can spook employees. How do you introduce it without triggering panic about layoffs? Any phrasing or timetables you recommend?

Sponsler: I start with the Zone of Genius from The Big Leap: what you love and do well, what you’re good at but don’t love, and what drains you. We point AI at the draining work first. The message is plain: AI is a co-pilot, not a headcount plan. If AI saves you an hour, we reinvest that hour into your craft or the customer. I like a 90-day rollout because it keeps everyone honest — month one we pilot one or two high-impact workflows per team; month two we expand to adjacent tasks; month three we standardize what worked and publish the wins. When employees see their job getting more interesting — not smaller — the fear drops. Owners see it on the KPIs: booking rate, average ticket, same-day/next-day capacity, and fewer callbacks. 

ACHR NEWS: What are some of the best ways to get employees up to speed on how to effectively leverage AI? Are there any dos and don’ts?

Sponsler: Teach people to prompt, not just click around. A usable prompt tells the model who it is, what outcome we want, what constraints matter, the example to learn from, and how to format the output so it plugs right back into the work. I give each role a small set of proven prompts and improve them over time. Keep training anchored to one real task a week so the “before/after” is obvious. Don’t flood people with features, and don’t let AI guess off thin air — attach the P&L, the call rubric, or the SOP and require it to cite what it used. Also set the data rules on day one so nobody pastes something sensitive where it doesn’t belong. 

ACHR NEWS: AI hallucinates sometimes. How do you safeguard against that? What guardrails actually help?

Sponsler: Process first. Keep a human in the loop for anything customer-facing during the first 90 days of a new workflow. Bind the model to a source of truth — P&L export, call rubric, SOP — so it references facts instead of freelancing. Keep a small golden sample of known-answer cases; any prompt change has to pass those before we scale. Version prompts like code, give someone ownership, and keep a rollback plan. If the output contradicts a baseline you trust — like labor magically dropping to 10% in July — assume the prompt or the data path is wrong and fix that, not the person reading it. 

ACHR NEWS: Can you give us a specific example or success story of a company integrating AI?

Sponsler: We’ve had clients run inbound call transcripts against CEO Warrior’s Phone Warrior model. AI scores the greeting, empathy, discovery, options, membership mention, and the close, then gives each rep three targeted fixes. Managers stopped hunting for coaching moments; reps got precise direction they could use on the next call. The pattern has been higher booking rates, more consistent membership conversations, fewer price-only dead ends, and faster ramp for new CSRs. A quick field vignette: a tech finishes a job, snaps photos, and dumps a few rough notes. AI turns that into a clean, options-driven estimate and a plain-English summary the homeowner can trust. Less typing in the truck, fewer callbacks, better close, and a steadier average ticket — without asking the tech to become a copywriter.