Best AI receptionist for home services: test the booking, not the demo
No supplier is “best” in the abstract. The useful comparison is verified bookings per eligible call on your own call mix, with service-area and capacity rules enforced.
Independent AI voice-agent research
For the people accountable for the call.Voice agents. Real decisions.
The call
sounded good.
Did the
booking stick?
Research the outcome, not just the voice.
Independent research and practical tools for US operations, finance and procurement teams buying AI voice agents.
Free tools. No sign-up. Your inputs stay in your browser.
THE OUTCOME MATTERS.From the sample call00:52.7
00:40.4 Booking requested00:42.1 Record confirmed
Read the evidence.
Make your own call.
- research studies
- 12
- buyer tools
- 10
- provider profiles
- 15
No paid placement. No referral fees. ↗
01 / The listening room
Listen past
the voice.
A convincing conversation is only one part of the job. Follow a call through to the booking record—or the person who receives the handoff.
Illustrative call — not a supplier benchmark
Inbound plumbing booking · water heater leak
+1 (312) 555-0187 · Inbound · plumbing dispatch · Phoenix service area (synthetic)
- Duration
- 00:52.7
- Usage (assumed)
- $0.16
- Outcome
Watch for the difference between a booking request and a confirmed system write. Illustrative cases; not executed against any supplier.
Build your own trial pack02
One queue. A clear boundary.
Start with the workload, not a vendor logo. Each queue is bounded: written exclusions, a destination system and a named fallback.
- New service requests inside the service area
- Arrival-window selection from live availability
- Confirmation text after the job record exists
- Gas smells, flooding or electrical hazards
- Invoice disputes and warranty claims
- Quotes for work not on the price book
Field-service job system (write access to jobs and arrival windows)
Dispatch supervisor, staffed 07:00–18:00; on-call technician phone after hours
- Missed and abandoned calls by hour
- Booked jobs per answered call
- Repeat calls for the same job within 48 hours
03
The price of a minute. The cost of an outcome.
A per-minute quote leaves out setup, failed attempts and human intervention. Compare the full cost of completing the task, then change the assumptions to see what matters.
Direct AI usage only
$1.24AI agent, all-in
$4.37Answering service
$3.47Staff only
$5.30
Change the assumptions
Your question. Your starting point.
A desk for
the decision-maker.
Will containment hide unresolved calls?
Start with Trial builder05
Research you can put to work.
Open a model, change its assumptions and export a working file. These previews use synthetic data. Downloads are free; no email is required.
Attachments · 6 working files
True Cost per Completed Call
Scenarios on the same denominator
| Scenario | Unique tasks | Completed tasks | Monthly total ($) | Steady monthly ($) | Cost per completed task ($) | Steady cost per task ($) |
|---|---|---|---|---|---|---|
| Low | 1056 | 587 | 3300.74 | 2550.74 | 5.62 | 4.34 |
| Base | 1056 | 697 | 3045.15 | 2295.15 | 4.37 | 3.29 |
| High | 1056 | 807 | 2789.56 | 2039.56 | 3.46 | 2.53 |
06 / Before you commit
The questions
behind the purchase.
Technology is only part of the decision. Check responsibilities, review the controls and understand which layer you are buying.
01Who owns the failed call?Compare Tool, Build, Expert and Keep.
04
Choose who owns the failed call.
Pick an ownership model before you shortlist suppliers. The ledger shows what stays visibly with you on each route.
Find my ownership routeBounded workflow, supported integrations, tolerable configuration limits.
Stays with you
- Testing on your calls
- Caller policies
- Fallback staffing
- Monitoring
- Contract review
02What needs legal and security review?Explore recording, consent and data responsibilities.
06
A working agent is not permission to make the call.
Bring legal and security approvers in before procurement. This lists review responsibilities. It does not assess legality and gives no clearance.
| Control | Why it applies | Status | Evidence to request |
|---|---|---|---|
| All-party recording consent (CA) | Calls are recorded | Needs evidence | Disclosure script, timing and placement; state-by-state review |
| Automation disclosure | Caller speaks with an AI agent | Needs evidence | Script showing the caller is told they are speaking with an automated assistant |
| Retention and deletion | Recordings and transcripts stored | Needs evidence | Retention schedule (90 days), deletion proof, backup handling |
| Subprocessors | Audio and text flow to third parties | Open | Named subprocessors, locations, change-notice terms |
| Model training on your calls | Supplier data-use terms | Open | Opt-out confirmation or contract exclusion |
Where the data goes in this deployment
Scroll advances one connection at a time.
01
Caller
Voice and phone number. Disclosure happens here, before anything is captured.
02
Telephony
Carrier and region. Audio may be recorded at this layer.
03
Agent
Transcripts and prompts. Subprocessors, retention and training use apply.
04
Operational record
Booking or job system. Least-privilege write access and an audit log.
03Which layer are you buying?Explore the supplier directory by deployment responsibility.
09
Know which layer you are buying.
A speech component, a developer platform and a managed service are not interchangeable. Each sits at a different layer of the call and leaves you different work.
Compare by deployment responsibilityContact-centre suite
Vertical or enterprise specialist
Voice-agent platform
Speech/model infrastructure
Telephony/network
A more informed first step
Start with
one queue.
Know what has to be true before you automate. The readiness scorecard helps you decide: pilot now, prepare first or hold off.
Check your first workflow Or choose an ownership route ↗Free. Runs locally. Saves to this device only.