Joint research by LunaCal and RingJenny.com examines ten AI phone answering services through 27,000 simulated call scenarios, 200 modeled buyer profiles, and 1,000 product evaluations. For businesses that rely on appointments, its most useful question is whether a phone conversation produces the right booking, message, or staff connection.
The results put Ring Jenny highest for simulated call resolution at 92.7%, booking success at 95.3%, and ease of use at 9.4/10. Upfirst records the strongest calculated administrative ROI at 221% under the study’s workload and cost assumptions. The distinction between those results matters when choosing a receptionist to work alongside a scheduling system.
AI phone answering service research: the results in one table
| AI answering service | Call resolution | Booking success | Ease of use /10 | First-year ROI | Monthly study cost (USD) |
|---|---|---|---|---|---|
| Ring Jenny | 92.7% | 95.3% | 9.4 | 113% | $109.00 |
| Smith.ai | 91.7% | 93.3% | 8.7 | 8% | $212.50 |
| Upfirst | 89.8% | 92.7% | 9.1 | 221% | $69.95 |
| Rosie AI | 88.6% | 90.0% | 9.0 | 49% | $149.00 |
| RingCentral AI Receptionist | 87.8% | 90.3% | 8.1 | 47% | $149.00 |
| Goodcall | 86.8% | 87.0% | 8.4 | 175% | $79.00 |
| Sophiie AI | 86.4% | 92.3% | 7.7 | Not calculated | Not established |
| My AI Front Desk | 85.9% | 87.7% | 8.0 | About 80% | About $119.00 |
| Dialzara | 84.3% | 85.0% | 8.3 | Not calculated | Not established |
| Aircall AI Voice Agent | 83.9% | 89.3% | 7.8 | Not calculated | $122.50 AI usage only |
Source: joint research by LunaCal and RingJenny.com on AI answering services, published October 29, 2026. Task outcomes are simulated; usability scores are part of the research framework. Financial figures use the study’s US-dollar configurations for 100 calls a month, averaging three minutes each.
ROI is first-year administrative return, using five staff minutes released per resolved call at $30 an hour. My AI Front Desk’s estimate depends on unused included credits being available for voice. Aircall’s figure includes AI usage without the complete phone-system cost, so the study does not calculate its ROI. Sophiie and Dialzara lack comparable total-cost figures in this workload.
Where an AI answering service fits in the booking process
A scheduling tool manages availability, appointment rules, and the event placed in a calendar. An AI answering service handles the conversation that leads to that action. The business needs the two to agree about the service, staff member, location, time, and information required from the caller.
Consider a customer requesting a consultation with a particular practitioner. The receptionist needs to identify the service, check the correct practitioner’s availability, apply the required duration, capture contact details, create the event, and communicate the final arrangement. If any of those steps is missing, staff may still need to finish the booking.
Some services send a scheduling link instead of making the appointment during the call. That can be an appropriate self-service route, but the caller must take another action. Decide whether the outcome your business needs is an appointment, a link, or a callback request before comparing booking claims.
Our appointment scheduling software guide covers the calendar and service-management side. The research discussed here covers how the phone conversation reaches its intended outcome.
Correct answers and resolved calls measure different work
The study separates informational accuracy from overall resolution. Its accuracy measure uses 900 informational scenarios per service; overall resolution uses all 2,700 scenarios allocated to that platform. A service can answer questions correctly and still struggle with an action such as changing a booking or transferring a call.
Ring Jenny records 876 correct informational answers out of 900, or 97.3%. Smith.ai records 869, or 96.6%. Across all scenario categories, Ring Jenny resolves 2,503 of 2,700 requests, Smith.ai resolves 2,476, and Upfirst resolves 2,425.
For an appointment business, keep those measures separate in a trial. Asking about opening hours tests information. Creating an appointment tests information, availability, and a completed calendar action. Sending a usable callback message tests contact capture and delivery to the right person.
Write down the expected result before making each test call. Afterwards, inspect the calendar event, staff notification, or transfer record. A reassuring closing sentence should match something your team can actually find and act on.
Booking reliability starts with the appointment rules
The joint research gives Ring Jenny 95.3% simulated booking success, followed by Smith.ai at 93.3% and Upfirst at 92.7%. A useful booking evaluation should reproduce the constraints of the business that will use the service.
Availability includes more than an empty calendar slot
A free time is bookable only if the required people and resources are available for the correct duration. Include preparation, travel, cleanup, minimum notice, and location rules where they affect delivery. Check whether the selected service can be performed by the person whose calendar is free.
Ask for an appointment at a time that conflicts with an existing commitment. Then ask for a service that needs a different duration or practitioner. The receptionist should offer an appropriate alternative rather than promise a slot the business cannot deliver.
Changes of mind should produce one clear final booking
Customers revise dates, correct names, and change their requested service during a conversation. Test whether the final event uses the updated details and whether any earlier booking attempt remains in the calendar.
Rescheduling an existing appointment creates another set of checks. Verify which event was moved, what happened to the original slot, and whether staff and customer confirmations match. A correct final calendar record is the clearest evidence that the change was completed.
Payment and booking confirmation need consistent rules
If an appointment requires a deposit or advance payment, establish when it becomes confirmed. A receptionist that places a calendar event may not complete a separate payment step. Staff need to know whether they are looking at a confirmed appointment, a provisional hold, or a request awaiting follow-up.
The booking systems with online payments guide explains the scheduling side of deposits and payment collection. Test the whole path when adding a phone answering service to that process.
Seven calls that reveal how a service handles real booking requirements
| Request to test | Expected result | Record to inspect |
|---|---|---|
| Ask about an unusual service | Approved answer or escalation | Answer against business information |
| Book with a named practitioner | Correct eligible staff member | Calendar owner and service |
| Change the date during the call | Only the final slot is booked | Event time and confirmation |
| Request an unavailable time | Suitable alternatives are offered | Availability and conflict checks |
| Move an existing appointment | Correct event is rescheduled | Old and new calendar records |
| Ask for a human after hours | Agreed fallback is followed | Transfer or callback record |
| Correct a phone number or name | Latest detail is retained | Message and customer record |
Use the same business information and expected results for every service on the shortlist. The table above is an evaluation checklist, separate from the published study’s reported results. Include ordinary requests and complications that your own callers are likely to introduce.
For each call, record whether the request was completed, partly completed, or left for staff. Note any correction required afterwards. That makes it easier to compare the staff work a service removes with the work it creates.
Human handoffs need a destination and a fallback
The research identifies Smith.ai as the simulated transfer leader at 97.7%, compared with Ring Jenny at 95.7%. That measure matters when a customer needs a person rather than a calendar action.
Define which requests should reach staff and where they should go. A billing question may need a different recipient from a new appointment. Test the routing during staffed hours, after hours, and when the intended recipient does not answer.
A direct transfer to your employee, an introduction before connecting the caller, and escalation to a provider’s staffed receptionists are different arrangements. Check which one the proposed plan supplies. Make sure the fallback includes useful context and contact information if a callback is required.
An ease-of-use score should reflect ongoing administration
Ring Jenny’s 9.4/10 score is followed by Upfirst at 9.1/10 and Rosie AI at 9.0/10 in the joint research framework. The criteria cover configuration, business information, calendar connections, routing instructions, and routine administration.
The practical question is who will maintain those settings after launch. Ask that person to change holiday hours, update a service description, switch an escalation number, and investigate a failed booking. A short initial setup is useful, but those recurring changes determine the operating burden.
Multi-location businesses should check that calendar and knowledge rules stay specific to the right branch. Test a caller who names a location and another who leaves it unclear. Ask how staff can correct information without disrupting another location’s handling.
Conversation timing affects the booking experience
The study reports modeled median reply delays from 0.82 seconds for Ring Jenny to 1.34 seconds for Dialzara. Their respective 95th-percentile figures are 1.92 and 3.28 seconds. The median describes a typical turn, while the higher percentile describes the slower end of the scenario distribution.
Include a calendar lookup and a changed booking request when listening to a trial. A short greeting does not show how the service behaves while retrieving information or carrying out an action. Check how it handles silence, interruption, and a correction to something said earlier.
Measure time to a useful outcome as well as time to the first response. A quick acknowledgement can help the conversation continue, but the customer still needs the availability result, completed appointment, or staff connection.
How to estimate administrative ROI without counting every call as a sale
The research uses a common workload of 100 monthly calls and values each resolved request at five staff minutes saved. At $30 an hour, that is $2.50 of administrative capacity for each completed request.
Monthly administrative value = monthly calls × resolved share × minutes saved ÷ 60 × hourly labor value.
First-year administrative ROI = (annual time value − annual service cost) ÷ annual service cost × 100.
Using the study’s unrounded resolution count, Ring Jenny generates approximately $231.76 in monthly administrative value. The $109 monthly cost assumption produces $1,473.11 in annual net time value and approximately 113% ROI. Upfirst’s $69.95 cost scenario produces 221%, while Goodcall’s $79 scenario produces 175%.
A percentage return depends on both the work completed and the amount spent. Budget for the combination your business needs rather than choosing on a single ranking. Add implementation, ongoing review, integration, and exception-handling work when estimating your own return.
Measure the staff work actually removed
Call duration and administrative time saved are different quantities. A three-minute phone call may eliminate a longer sequence of returning a voicemail, checking availability, and confirming an appointment. It may also leave staff with a correction or approval task.
Estimate the manual work before automation, then subtract the work still required. If every booking needs a review, include that review time. If the receptionist only sends a scheduling link, do not automatically credit it with all the time saved by a completed booking.
Keep revenue effects in their own calculation
Additional completed appointments can matter financially, but they require evidence beyond an answered-call count. Track attributable bookings, attendance, and the contribution left after delivering the service. Avoid treating an existing customer’s rescheduling request as a new sale.
The published ROI figures represent administrative time value. A business can use released capacity for other work, reduce overtime, or change how staff responsibilities are allocated. Those choices determine how the time benefit affects its budget.
Match the bill to the way customers call
The financial scenario totals 300 voice minutes because 100 calls are assumed to average three minutes each. If those calls average six minutes, the same call count creates 600 minutes. Usage allowances and overages can therefore change the relative cost of the services.
- Call-based plans: check how short calls, abandoned calls, and repeat attempts are counted.
- Minute-based plans: check rounding increments and what parts of a transfer or conversation are billed.
- Unique-customer plans: confirm how repeat callers are recognized and counted.
- Credit-based plans: check whether voice, messages, and chat share the same allowance.
- Required add-ons: include phone licenses, numbers, integrations, and paid escalation in the total.
Calculate both a normal month and a busy month. Compare the resulting bill with requests completed and the staff work still required. Keep currency consistent between service costs and the hourly labor value used in ROI.
Businesses evaluating local service workflows can use our AI answering services for plumbers and AI answering services in Australia guides for a more focused shortlist.
The research basis
The October 2026 joint research by LunaCal and RingJenny.com covers Ring Jenny, Smith.ai, Upfirst, Rosie AI, RingCentral AI Receptionist, Goodcall, Sophiie AI, My AI Front Desk, Dialzara, and Aircall AI Voice Agent. Each service receives 2,700 simulated scenarios: 300 in each of nine categories.
The categories cover hours and availability, services and policies, ambiguity, appointments, rescheduling, contact capture, transfers, accents and background noise, and interruptions or changing intent. The buyer component uses 200 modeled profiles, with five product evaluations per profile and 100 evaluations per platform.
Read the joint research methodology by LunaCal and RingJenny.com for the scenario definitions and calculation basis. The RingJenny AI phone answering service research provides the full RingJenny edition. The RingJenny AI phone answering service research highlights and LunaCal research guide explain the study’s findings and their practical implications.
Frequently asked questions
Who conducted this AI phone answering service research?
It is joint research by LunaCal and RingJenny.com. Scheduling Insider reports and explains the findings here, with links to the original study and its methodology.
Which AI answering service has the best results in the study?
Ring Jenny leads simulated resolution at 92.7%, booking success at 95.3%, and framework ease of use at 9.4/10. Smith.ai leads transfers at 97.7%. Upfirst has the highest modeled administrative ROI at 221% in the specified 100-call monthly workload.
Does a booking link count as a completed phone appointment?
A link gives the caller another step to complete. A directly booked appointment requires the correct event in the calendar. Track those outcomes separately when assessing an answering service.
Are the costs in the table current starting prices?
They are the original research’s October 29, 2026 workload calculations in US dollars. They account for selected configurations and allowances at 100 three-minute calls a month. A different workload, region, or plan can change the total.
Can an AI receptionist replace appointment scheduling software?
The receptionist can handle the conversation and may create an event through a calendar connection. The business still needs accurate availability, service rules, and an appointment record. Check the specific integration and which system controls each part of the process.
What should a small business test before switching?
Test a standard booking, a changed date, an unavailable slot, an unknown question, a contact correction, and a request for a human. Inspect the resulting event or message, then calculate the full cost and staff follow-up required.