If you work in home services, all this AI noise probably has you wondering: What does it actually look like to implement AI in your operations? And can you do it without hurting customer satisfaction?
That’s what we covered in this webinar, including what’s working, what’s not, and what’s next when it comes to putting AI on the phones.
Read on for the skim-friendly recap.
Speakers:
- Rich Jordan, CEO, High Ground Service Pros
- Tim Goodwin, CRO, Hatch
- Alec Stevanovski, Editor, HomePros (moderator)
- Kristen McCormick, Head of Marketing, Hatch (host)
The focus: how High Ground moved from skeptical to running all inbound calls through Hatch’s AI CSR—and what happened next.
TLDR: 5 hot takeaways
- AI for cost control. High Ground reduced labor costs by 36% by redeploying strong CSRs into higher-leverage roles, phasing out only a few low performers, and now being able to scale growth without ballooning headcount.
- AI for quality control. It's not just about costs. As CSR teams grow, quality diminishes because consistency gets harder. More people to train and scripts to follow.
- Metrics that matter. Abandoned call rate fell from 10–12% to 2%. 81% set rate on bookable leads, which is comparable to a high-performing human team.
- Transparency is the real differentiator. Hatch is the only tool Rich has used where his team can edit their AI’s knowledge base directly and see changes on the next call.
- AI matters even more for SMBs: Missed calls hit harder, CSRs multitask, and AI gives small teams big-company coverage without ballooning headcount.
Why High Ground bet on Voice AI in the first place
High Ground grew from 2 people to ~150 in five years. With that growth came a call center problem: CSR headcount was scaling linearly with revenue.
“If you grow 30X, you need ~30X the CSRs. That gets expensive.”
Even worse: after hiring more CSRs, the abandoned call rate stayed stuck at ~10%.
Other challenges included:
- Human inconsistency on scripts, fees, trades, after-hours rules
- CSR turnover and retraining
- Call center cost ballooning without performance gains
At first, Rich doubted the tech. “You think it’s going to sound like a bank IVR. But the tools are getting better, fast.” What ultimately moved him was:
- Newer voice AI sounded natural
- AI booking was reliable
- It was better than missing the call entirely
- The tech is improving weekly: “Voice AI today is the worst it will ever be.”
Step 1: Easing in with AI on overflow and after-hours
High Ground did not start by putting AI on the main line. They used Hatch to catch:
- Overflow calls after 15–20 seconds
- Calls that would’ve hit voicemail
- After-hours and weekend calls
“Even if AI’s not perfect, it’s better than missing the call entirely.”
This “shallow end” let the team:
- Build confidence
- Monitor every call
- Fix issues fast
- Push improvements into the knowledge base
And because High Ground already had a detailed Notion knowledge base, onboarding the AI was a cleaner lift.
They ran backend AI for about a year—and couldn’t imagine operating without it.
Step 2: Moving AI to the front line, gradually and intentionally
The jump happened once voice AI got good enough that Rich felt comfortable putting it on 100% of inbound calls.
Their rollout strategy:
One day a week, AI handles all calls
- CSRs off the phones
- AI handles all calls
- Team reviews every booking and non-booked lead
- Knowledge base updated constantly
Then two days a week
- Similar process, more volume, more refinement
Then five days a week
- AI now answers every inbound call
- If needed, AI escalates to CSRs with a live call summary so customers don’t repeat themselves
How the system flows
- Calls hit ServiceTitan first
- ST forwards to Hatch
- Hatch answers, books, or escalates
- High Ground created a Hatch AI employee in ST for reporting parity
An early fix: The “AI loop”
- Customers who asked for a human could accidentally hit AI twice if the call center missed the escalation.
- Solution: Escalations now route to a separate non-AI line, eliminating loops.
What changed: the numbers that matter
Abandoned calls
- Before: ~10%
- Now: ~2%
- “Just about every call is being answered.”
Booking rates
- Total calls booked: just under 20% (very close to human baseline of 20–24%)
- Bookable leads booked: 81% (top-tier human benchmark ~85%)
- Benchmark: Many call centers operate around 60–65%
Call time
- AI: ~3–3.5 minutes
- Humans: ~3–3.5 minutes
- No loops, no longwinded scripts.
Escalation rate & call volume
- 75% of calls handled fully by AI
- 25% escalate to humans
- Human call volume down by three-quarters
- High Ground handles ~10,000 calls/month
Headcount
- 36% reduction in CSR headcount
- High performers redeployed into:
- Dispatch support
- Permits and inspections
- Install coordination
- Purchasing
- Other admin-heavy areas
Culture impact: From fear to “don’t take this away”
Rich said putting AI on the backend was easy; putting it front and center was “disruptive and anxiety-inducing.”
CSRs feared displacement—but High Ground handled it with transparency:
- “We’re not nuking the call center.”
- “We need you for escalations, nuance, and ops work.”
- “We want a smaller team of elite CSRs—not endless hiring.”
They involved CSRs in the rollout process. And now…
“If I told them AI will be gone next Monday, they’d riot.”
Because the AI:
- Reduced call load by 75%
- Freed CSRs to do higher-value work
- Made their day less chaotic
What good AI looks like
Tim shared why the tech finally crossed the threshold from “interesting” to “operationally viable”:
- Underlying models improved massively in the last year
- Hatch spent ~6 months getting the product reliable for real call centers
- AI now handles:
- Interruptions
- Tone shifts
- Sentiment
- Complex branching without freezing
Rich’s favorite feature? Direct access to the AI’s knowledge engine. No waiting on support.
“A lot of voice AI tools are black boxes. With Hatch, you can see and edit your own knowledge base and watch changes appear on the next call.”
This transparency was a deal-maker.
Looking ahead: Refine, don’t reinvent
High Ground’s 2026 focus is:
- Refinement (driving escalation rate lower)
- Better routing logic
- More internal runway for CSRs (purchasing, permits, coordination, etc.)
- Deploying people strategically, not reactively
“We’ve culled the team down to strong people. Now we want to deploy them where they create leverage.”
Q&A
Metrics & performance
Q: What did your abandoned call rate normalize to after implementing AI?
A (Rich): It’s around 2% now. Before, even while adding CSRs, we were stuck around 10% abandoned.
Q: How has AI impacted your booking rates?
A (Rich):
- On total calls, Hatch books just under 20%.
- Historically, our humans booked 20–24% on lower call volume (since we were missing ~10% of calls).
- On bookable leads, a strong human call center should be around 85%; Hatch is at about 81% for us.
- Many call centers that aren’t dialed in sit in the 60-something percent range.
Q: What’s your average number of calls per week/month? Have you “load tested” AI in peak season?
A (Rich):
- We handle about 10,000 calls per month (~2,500 per week).
- We haven’t done a formal “load test,” but I don’t expect AI to have more issues with volume than my CSRs would.
Small business & use cases
Q: As a small business owner with two CSRs, is AI really necessary, or should I just consider it for peak seasons?
A (Rich): I actually think as a smaller team, AI can be more important:
- Your abandoned call rate is more vulnerable during peak times.
- You often need your CSRs to dual-hat as office managers, A/P, etc.
- AI can catch overflow and give your team breathing room.
You can absolutely start with peak season and overflow if you don’t want it on 24/7 right away.
Backend vs voicemail & capabilities
Q: If AI is on the backend, isn’t that basically the same as voicemail? Don’t CSRs still have to call customers back?
A (Rich):
No—this is the key difference. The AI can book the job live in your system. With voicemail, you’re hoping your team can call back fast enough before the customer calls a competitor. AI reduces that risk by handling it in real time.
Q: I haven’t considered running AI on the backend. What can the AI actually do there?
A (Rich):
- Book service calls directly into ServiceTitan.
- Book based on real-time capacity, even on weekends/after hours.
- For non-booking issues, it can take detailed notes from the customer and pass them to the admin team for action in the morning.
How AI hands off to humans
Q: What’s the process from front-end AI to a live agent?
A (Rich):
- The AI talks to the customer first.
- If it needs to escalate, it forwards the call to the call center.
- Hatch provides a heads-up summary on the CSR’s screen: what the customer said, what’s been tried, and why it’s escalating.
- That way, the CSR doesn’t have to ask the customer to repeat everything.
Q: Are calls being left “unassigned” in ServiceTitan when AI books them?
A (Rich):
No. We created a “Hatch AI” employee in ServiceTitan. That way:
- Calls show up on our call center scorecards as being handled by Hatch AI.
- We can track number of calls, booking rate, etc., just like we do with human CSRs.
- For technicians, we often let bookings go to unassigned and then use DispatchPro or our dispatchers to assign techs later.
Setup, training & definitions
Q: What’s the general training time to prepare and launch AI?
A (Tim):
- For a smaller, simpler operation: ~2–3 weeks is typical.
- For larger or more complex companies (multiple regions, territories, heavily segmented rules), it can take longer.
- Most of that time is making sure we handle edge cases correctly and that the AI is aligned with your SOPs.
Q: How do you define a “bookable lead”?
A (Rich & Tim):
- Rich: No different than any contractor: Did this person call for service, and did we book it?
- Tim: We also use AI to infer intent from transcripts and classify which calls represent an intent to book. Those are your “bookable” leads.
Financial impact & team structure
Q: In terms of savings as a percentage of SG&A, what have you seen since moving to 100% inbound AI?
A (Rich):
We’ve seen about a 36% reduction in CSR headcount in the call center.
- Some low performers were let go and not replaced.
- Many others moved into higher-leverage roles (dispatch support, permits, inspections, install coordination, purchasing, etc.).
Integrations & where AI fits
Q: What field service software does Hatch integrate with? Does it work with all FSMs?
A (Tim):
- We’re most deeply embedded with ServiceTitan today.
- We also connect to Jobber, and we’re working on expanding and deepening integrations with platforms like Housecall Pro, Jobber (more deeply), and Salesforce, among others.
- If you’re on one of these platforms, let us know—we can keep you in the loop as integrations deepen.
Q: Is this more for home service/residential contractors, or does it work for commercial too?
A (Tim):
- Today we’re primarily focused on residential/home services, where call volumes and patterns fit really well with AI CSRs.
- There may be interesting use cases in commercial, but many of those calls are more sales-oriented, where companies often want their sales team directly involved.
- We’re open to learning more about commercial needs, but our current sweet spot is residential.
Dispatch & technician assignment
Q: Can the AI agent prioritize certain techs for calls (e.g., high-ticket techs getting high-priority jobs)?
A (Rich):
We don’t use it that way. We actually prefer:
- Jobs to be booked as unassigned
- Then let DispatchPro or our dispatchers decide which tech to put on which job
A (Tim):
I’m ~90% sure we can support routing with tech prioritization logic, but I’ll confirm with our product team before making that a promise.
Want to learn more?
For past and upcoming Hatch webinars, check out our webinar page!