Most auto attendants are bad because the person who built them was tired of the project halfway through. Seven options, three of them stale, hold music that loops every 14 seconds, and a default route that sends callers to a voicemail nobody checks. Here is how we set them up.
The three-decision rule
If a caller has to make more than three decisions to get to a human, they will hang up. That is the practical ceiling. Build your tree backwards from there.
- Greeting (5-8 seconds). Company name, then straight to options. Skip "thank you for calling" if you can. The caller knows where they called.
- Top-level menu (3-5 options max). Press 1 for sales, 2 for support, 3 for billing, 9 for the directory. Anything past 5 starts to fade in a caller's working memory.
- Sub-menus only when the volume justifies it. If "support" gets 200 calls a week split across three product lines, a sub-menu earns its keep. If it gets 12, just ring the support hunt group.
Why three is the ceiling
Working memory research is not the reason. The reason is that callers do not want to be on the phone. They picked up the device to solve a problem and they are now in a holding pattern listening to a list of options. Every decision after the third feels like punishment. Watch your own behavior next time you call a big-box retailer; the moment you hit the third sub-menu, you start mashing zero. Build your tree assuming your callers behave the same way.
Things to actually get right
The voice
Pay a real voice talent once. It is $150 to $400 on Voices.com or Voice123, and the recordings will outlast three phone systems. Synthesized voices are fine for menu options that change weekly; for the main greeting, the difference is audible and it sets the tone. We also see customers using ElevenLabs or PlayHT to generate prompts; those work for routine menu changes, but for the brand-defining main greeting, a human reading the script still sounds better than a model trying to sound human.
The after-hours flow
This is where most attendants leak business. After hours should not just say "we are closed." Give the caller a way to leave a voicemail that actually gets routed to a real inbox in the morning, and on our system, voicemail-to-email transcription is a paid add-on for a reason: when you turn it on, people read voicemails instead of ignoring them. Emergency support? Route it to an on-call rotation, not to the same voicemail.
A well-designed after-hours flow has at least three branches: a voicemail box for routine matters, an emergency path for urgent matters, and an informational option that plays your hours, address, and website. The informational option deflects calls that did not need a human at all ("are you open Sunday" is the most common one).
Holiday and weather routing
Set up a separate "holiday" greeting on a time-of-day rule and forget about it for the year. Hurricane in Central Florida? We flip your routing to a backup destination from our side in minutes. Do not depend on someone being in the office to change the greeting.
Build the holiday calendar once a year, in December, with all your scheduled closures (New Year, Memorial Day, July 4th, Labor Day, Thanksgiving plus the day after, Christmas Eve, Christmas Day, plus any company-specific days). The system handles the rest automatically.
Dial-by-name and direct extensions
If your callers know who they want, do not make them sit through a menu. Publish direct extensions on your website, and put dial-by-name on option 9. You will be surprised how many people use it.
Real menu builds for specific offices
A four-doctor dental practice
The wrong menu: "Press 1 for new patients, 2 for existing, 3 for billing, 4 for insurance, 5 for emergencies, 6 for pediatric, 7 for orthodontics, 0 for the directory."
The right menu: "For appointments, press 1. For dental emergencies, press 2. For billing or insurance, press 3. To speak with the front desk, stay on the line." Four decisions max, and most callers hit 1 or stay on the line. The detail on "new vs. existing" gets handled by the receptionist or the AI Receptionist with $49/mo HIPAA add-on, not by a menu. The dental practice phone system page goes deeper on the dental-specific routing patterns.
A 12-attorney law firm
The wrong menu: "Press 1 for family law, 2 for personal injury, 3 for real estate, 4 for estate planning, 5 for criminal defense, 6 for our directory." By the third option the caller has forgotten what was on option 1.
The right menu: "For an existing matter, press 1. For a new consultation, press 2. For billing, press 3. For the firm directory, press 9." Practice area routing happens once they identify whether they have a matter open, which is what the routing actually needs to know. See legal firm phone system for how this plays with Clio and call recording.
A 30-agent real estate brokerage
Most calls are for a specific agent. So lead with dial-by-name: "If you know your agent's extension or name, press 1. For our front desk, stay on the line. For new buyer inquiries, press 2." Lead routing for option 2 goes into the Follow Up Boss queue with round-robin distribution.
A multi-location property management company
Three top-level options. "For maintenance requests, press 1. To reach a leasing agent, press 2. To speak with our office, press 3." Maintenance goes straight into a queue that ties to AppFolio or Buildium tickets. Leasing routes by property zip code or by source if the lead came from Zillow vs. a direct call.
An HVAC shop
"For service or repairs, press 1. For a new system quote, press 2. For billing, press 3." Option 1 routes directly into the ServiceTitan dispatch queue with caller history pre-loaded. Option 2 goes to sales. Option 3 to bookkeeping. After-hours emergencies have their own path to the on-call rotation.
Where auto attendants fall short
An IVR menu is a routing tool, not a receptionist. It cannot tell a confused caller what your hours are, take a service request, screen out spam, or qualify a sales lead. For practices and offices that want a real conversation on the first ring, the AI Receptionist handles that.
Pricing is flat per office, not per minute: $99/mo Starter, $199/mo Pro, $299/mo Enterprise. HIPAA-compliant deployment is a $49/mo add-on for medical and dental practices. It answers in your business name, books appointments to your calendar, captures lead info into your CRM, and texts the caller a follow-up. For a single-location dental or wellness practice, $199 is less than two missed-call appointments per month, which is the actual math you should be doing.
When the menu wins anyway
If 90% of your inbound is routine routing ("transfer me to billing"), an IVR is faster and cheaper than an AI receptionist. The AI shines when callers want to have a real conversation. Many of our customers run both: AI Receptionist on the main line, a traditional IVR on a secondary line for known caller use cases.
If you are sticking with a traditional attendant
Our Phone Service includes a full IVR builder at every tier. You can record greetings from the portal, set time-of-day rules, route between queues, and pull call analytics to see where callers drop. The setup work is the same for a 10-seat office as for a 100-seat one: it is the thinking that takes the time, not the clicks.
Time-of-day rules in practice
Define your business hours once. Define your lunch hour separately if you have one. Define holidays as overrides. Test each rule before you publish; the most common bug is a missing edge case (Friday at 5pm-and-1-second goes to after-hours, which works, but Friday at 4:59pm on a holiday-eve does not, which sometimes does not). The portal shows live state per route, so you can see what the system thinks the current time block is at any moment.
Common mistakes we fix on day one
- Hold music with sharp loop points. A 14-second jingle that audibly restarts every loop signals "low effort" within 30 seconds. Use longer beds, or licensed instrumental tracks.
- No timeout fallback. If a caller does not press anything within 8-10 seconds, the system should route them somewhere, not replay the menu three times and hang up.
- Voicemail boxes that nobody owns. A general "info" voicemail with no assigned recipient is where messages die. Every box needs a human who checks it.
- Day-of-week rules with no holiday overlay. Your Monday-Friday schedule does not know about July 4th unless you tell it. Build a calendar of closures once a year.
- Recording new greetings from a cell phone. The audio quality is jarring next to the rest of the prompts. Either re-record everything or use a real mic for the new prompt.
- Routing emergency calls to voicemail after hours. If your business has emergencies (medical, plumbing, security), there must be an on-call path that pages a real person.
- Recording the menu options in a different voice than the greeting. Either re-record everything in the new voice or commit to the old voice. Mixed voices sound amateurish.
- Forgetting to disable the menu option that no longer applies. The "press 4 for the Tampa office" prompt that points to a closed location is the kind of detail callers notice.
What to ask before publishing a new menu
- Can a first-time caller reach a person in under three button presses?
- Where does a caller go if they say nothing for 10 seconds?
- Where does a caller go at 2am on a Saturday?
- Who owns each voicemail box, and do they check it daily?
- What happens if option 4 leads to a department that does not exist anymore?
- If the building loses power, where do calls go?
- Are emergency paths actually paging a person, or just leaving a message?
- Does the greeting still match the company name and branding from this year?
Measuring whether the menu actually works
The Phone Service portal reports drop-off per menu option, time spent in each branch, and after-hours volume. Pull those numbers monthly for the first quarter after a redesign. If 22% of callers hang up at the top-level menu, the menu is the problem, not the callers. If 60% of callers press 0 to bypass the menu entirely, you have your answer about how much menu they want.
The metrics that actually tell you something
- Top-level abandonment rate. Callers who hang up before pressing anything. Under 5% is healthy. Over 10% means your greeting is too long or your voice is wrong.
- Mid-menu abandonment. Callers who press one option then hang up before reaching a human. Usually means the wait time at the destination is too long or the prompts inside the branch are confusing.
- Zero-bypass rate. Percentage of callers who press 0 to skip the menu. Over 30% suggests callers do not want to interact with the menu at all; consider routing the main number to a queue directly.
- After-hours voicemail return rate. Percentage of after-hours voicemails that get returned within one business day. Below 80% means the boxes are not being checked.
Pricing for the parts that matter
The auto attendant builder is included with every Phone Service seat at $15 or $32 per user per month. Recording prompts in the portal is free; if you want professional voice talent we point you to Voices.com or Voice123. Voicemail transcription is a paid add-on, billed per minute of transcribed audio. The AI Receptionist runs $99/$199/$299 per office per month with a $49/mo HIPAA add-on where it applies. Per-number porting is $15 each way. Toll-free numbers and DIDs are on the pricing page.
Where to start
If your current menu has more than five top-level options or your after-hours flow ends at a dead-end voicemail, the fix is one conversation. Send us your current call flow and we will redraw it. If you would rather skip the menu entirely and have a real voice answer, AI Receptionist is the page to look at. For pricing on every piece of the stack, see pricing. Comparing us against the providers you already know? Start at vs.