Skip to content

AI systems

AI receptionist cost: a checklist for small-business buyers

There is no useful comparison until both quotes cover the same job, usage unit, connections, human review, and fallback. Build the comparison from your own volume and vendor terms.

Why one headline number cannot answer the question

A monthly figure is only comparable when two offers cover the same work. One may include an answering line; another may include setup, appointment requests, connection to a calendar, and a human handoff. Their billing units may also differ. Before comparing amounts, write down what the receptionist must do and what each vendor has included.

Published vendor prices vary by plan and usage rule. The examples below are dated snapshots, not an industry-wide average; compare your own call log against each provider’s current terms and the work each price includes.

Published US plan examples

Checked September 27, 2026, Goodcall lists AI phone-agent tiers at $79, $129, and $249 monthly per agent, with different unique-customer allowances and a $0.50 overage for each customer above the tier limit. Its published pricing page defines a unique customer by caller phone number. Smith.ai lists a free tier for 25 calls, then $3 for each additional call if overages are enabled; its paid AI plans list $150 and $500 monthly at different call levels. See its AI receptionist pricing. These are live vendor examples, not an industry average or a quote for your business. They count usage differently, so compare the same call sample against each provider’s current terms.

Worked example: 90 callers in a month

Suppose a repair shop logs 90 calls from 90 distinct numbers in a month. On Goodcall’s $79 Starter tier, that sample sits within the listed 100 unique-customer allowance. On Smith.ai’s free tier, 25 calls are included; if the shop enables the listed $3 overage and makes 90 calls, the additional 65 calls would add $195. That arithmetic excludes taxes, optional services, and staff time, and it does not make the plans equivalent: Goodcall counts unique customers while Smith.ai counts calls. The example shows why the same call log and the actual contract terms belong in the comparison.

List the one-time setup work

Ask what must be completed before the first customer can use the system. Setup may involve writing approved answers, cleaning source material, configuring a phone or messaging channel, connecting a calendar or customer record, and testing handoffs. Separate what the vendor does from what your team must supply and approve.

Setup lineQuestion for the quote
Scope and conversation designWho defines the allowed questions, answers, and escalation rules?
Source preparationWho checks that the policies and service details are current?
ConnectionsWhich accounts, permissions, and integrations are included?
Testing and revisionsHow are failed cases recorded and corrected before launch?
Training and handoverWhat does the team receive to operate and review the system?

A quote for initial setup that excludes these tasks is not necessarily lower; it may be moving work onto your staff. Record the internal hours alongside the vendor quote so both options use the same boundary.

Normalize the recurring and usage units

Get the billing unit in writing. A plan may count connected minutes, calls, conversations, messages, locations, or staff seats. Ask what happens when a conversation is transferred, a caller repeats information, a call runs long, or a connection retries. If the offer uses included usage, ask how excess usage is measured and reported.

Quote fieldRecord this before comparing
Base subscriptionWhat stays due when usage is low?
Usage unitWhat event is counted and when does its meter start or stop?
Included volumeWhat volume is included for the quoted period?
Overage treatmentWhat happens when included volume is exceeded?
Telephony or messagingAre carrier, number, or channel charges separate?
Multiple channelsDoes each line, account, or location change the bill?

Build a usage baseline from your own logs. Separate answered inquiries from spam, internal calls, duplicates, and transfers. Include busy periods rather than relying on an average day. Ask a vendor to show how that same sample would be counted under its contract.

Model the busy period, not only the average

An average can hide the exact time the system is meant to help. Split your sample by business hours, after-hours, and known busy windows. Ask how simultaneous inquiries are handled, whether a transfer continues to count as usage, and what the customer hears if the line or integration is unavailable. Get the answer in the proposal rather than relying on a demo.

Add the work that continues after launch

The operating bill can include more than software. Count the people who review exceptions, update approved answers, investigate failed connections, and check whether records reached the right place. Include any separate storage, analytics, or system administration your team needs. A system that requires regular attention still has an operating cost even when the invoice does not show those hours.

  • Human review time for uncertain answers and escalations.
  • Changes when policies, hours, services, or staff ownership change.
  • Monitoring, incident response, and access reviews.
  • Integration maintenance when a connected system changes.
  • Internal time spent reconciling records or correcting missed handoffs.

Use one total-cost equation

For each vendor, calculate the same categories over the same review period. Do not let one quote use a monthly bill while another spreads setup across an assumed contract term. Keep one-time and recurring totals visible before deciding whether to combine them.

Keep setup separate: configuration + source preparation + connection work + testing + training + your team’s time. If a vendor includes one of these, mark it included rather than treating it as zero-effort. You are trying to compare the same delivered scope, not the most attractive line on a quote.

Calculate a break-even point from your records

Do not assign a made-up dollar value to a recovered call. Use your own evidence. Start with the inquiries the team can verify were missed or delayed, then count which ones became completed requests or appointments after the new process. If the business cannot connect an inquiry to a completed outcome, keep it out of the revenue calculation.

  1. Count the work the current process leaves waiting.
  2. Measure how many of those cases the proposed system completes correctly.
  3. Use your own contribution margin or saved staff time, not a vendor’s generic value claim.
  4. Subtract the full recurring operating total and track setup separately.
  5. Repeat the calculation after a review period using recorded outcomes.

A simple model is: verified monthly value minus full recurring cost. If that result is not positive, or if the underlying value cannot be measured, do not call the purchase self-funding. A cost model is useful even when it tells you to keep the current process.

Count the cost of an exception

A routine answer is only one outcome. Count what happens when the assistant cannot answer, the calendar is unavailable, a customer asks for a person, or a request arrives with missing details. A safe handoff still takes staff time, and a failed handoff may send the team back to manual search. Record those cases separately from successful routine work.

For each exception, write down the person who receives it, the context they need, the customer-facing fallback, and how the team confirms resolution. Ask the vendor to show this route in the same test used to compare plans. A quote that omits exception work can understate the total even when its usage meter is accurate.

Compare quotes line by line

TopicQuote AQuote B
Same channel and job?Yes / no / detailYes / no / detail
Billing unit and included volumeWrite the exact termsWrite the exact terms
Setup and connection workIncluded or separateIncluded or separate
Human review and handoffOwner and processOwner and process
Failure alerts and supportHours and response pathHours and response path
Data access and exitExport and deletion termsExport and deletion terms

Ask both vendors to walk through the same real inquiry and the same failure case. Include the response, the record created, the notification, and the recovery path. A quote that cannot explain what happens when the assistant does not know the answer is not a complete operating scope.

Make the exit part of the cost

Before signing, ask how you can retrieve conversation records, configuration, source material, and customer data if you stop. Confirm which data can be exported, the format, the steps for revoking access, and the provider’s deletion process. Also ask what remains usable if an integration or vendor account is removed.

These questions are easier before implementation than after your workflow depends on a single service. Brand Wisdom’s AI systems service is scoped around the work, connections, tests, and handover; the page does not publish a price before those details are known.

A worksheet you can fill in

Your inputValue to collect
Current monthly inquiry sampleCount from your own call or inbox records
Vendor billing unitExact unit and counting rule
One-time setupQuote plus internal staff time
Recurring platform and usageQuote at your measured volume
Human review and maintenanceHours and responsible owner
Verified completed outcomesCount from records after review

Use the worksheet to ask better questions, not to predict a result the business has not measured. Keep the source for each input beside the number. If the quote changes its included volume, update the model rather than carrying forward an old estimate.

The short answer

The cost of an AI receptionist is the cost of the whole operating lane: setup, usage, connections, human review, upkeep, and the work that remains with your team. Compare complete scopes against your own inquiry sample. Then test the proposed workflow before expanding it.

QUESTIONS

Common questions

Why do AI receptionist quotes use different units?

Vendors define usage differently. A quote may count minutes, calls, conversations, messages, seats, or locations. Compare the counting rule and included volume before comparing the headline amount.

What is usually missing from a software subscription?

Check whether setup, source preparation, integrations, testing, human review, maintenance, phone or messaging charges, and internal staff time are included or separate.

How do I estimate whether an AI receptionist is worth it?

Use your own inquiry records, measured completed outcomes, contribution margin or saved staff time, and the full operating cost. Exclude outcomes you cannot verify and review the calculation after launch.

REFERENCES

Sources and further reading

← All field notes