Forty Seconds, From the Other Side
Follow one caller. She has a burst pipe under the house, she found you on Google, and she has three tabs open with your competitors in them.
| Elapsed | What she hears | What she is doing |
|---|---|---|
| 0:00 | "Thank you for calling. Please listen carefully as our menu options have changed." | Nothing yet. Mild resignation |
| 0:08 | "For sales and new enquiries, press 1." | Is this sales? It is not really sales. It is a burst pipe |
| 0:14 | "For accounts and billing, press 2." | No |
| 0:20 | "For service and maintenance, press 3." | Maybe. Is a burst pipe maintenance? It feels more urgent than maintenance |
| 0:27 | "For all other enquiries, press 4." | She presses 3, because she has to press something |
| 0:31 | Hold music | Switches to speakerphone. Looks at the second tab |
| 0:52 | A person: "Service, this is Dan." | Explains the burst pipe |
| 1:20 | "Ah, that's emergency — I'll put you through, can I get your name and address?" | Gives them. Again |
| 1:31 | Hold music | Rings the second tab on her mobile while she waits |
The most urgent call your business received that day spent its first thirty seconds being asked to choose a department, and its next forty being transferred. Nothing in your reporting will show that as a failure — it will show two answered calls.
Nothing in that sequence involved anyone doing anything wrong. Dan was helpful. The menu was clearly worded. The business is competent. The failure is in the format, which is why fixing the wording never fixes the outcome.
Nobody Chose Your Menu
Ask whoever set up your phone system how the menu structure was decided. In almost every business the honest answer is some version of: we listed the departments.
| How menus actually get built | What that produces |
|---|---|
| Someone listed the teams on the org chart | A menu describing your internal structure — a document your customers have never seen |
| Options were ordered by seniority or habit, not by call volume | The thing most people ring about is frequently option four |
| It was recorded once, in an afternoon | The business has since changed. The menu has not |
| Nobody has listened to it since | Including, usually, the owner |
Why it survived this long, in fairness
The menu was a good answer to a real constraint. When you had four physical lines and one receptionist, answering was genuinely scarce, and sorting callers before a human touched them spent that scarce resource well. On a cloud platform there is no fixed line count — ten callers can be answered on the first ring, simultaneously. The menu now rations something that has become free, while spending the thing that has not: your customer's patience.
The Mismatch Underneath Everything
Every problem with phone menus comes from one root cause, and once you see it you cannot unsee it.
A menu is organised around your company. Callers think in problems.
Nobody rings "accounts". They ring because an invoice looks wrong. Nobody rings "service". They ring because water is coming through the ceiling. The menu asks the caller to perform a translation from their problem into your structure — using knowledge of your business they do not have, under time pressure, while a recording talks at them.
Everything downstream follows from that one demand:
They choose wrong
Not because they are careless, but because the categories are yours and the problem is theirs. So they reach the wrong team, get transferred, and the wait restarts. This is the expensive failure and it is invisible in reporting — it looks like two answered calls.
They wait for options they do not need
The list is linear and unskippable at the speed of the recording. The regular caller who knows they want option four hears one, two and three anyway, every single time, forever.
They have a problem you did not list
A menu can only contain what someone thought of in an afternoon. Everything else becomes "press 4 for all other enquiries", which is a shrug with a keypress attached.
They give information twice
"Please enter your account number", then a human asks for it. This is the single most reliably infuriating thing a phone system does, and it happens because the menu and the person are not joined up.
What Replaces It
The replacement is not a shorter menu. It is the removal of the translation demand entirely.
| Menu | Conversational answering | |
|---|---|---|
| First thing the caller hears | "Please listen carefully to the following options" | "What can I help you with?" |
| Who classifies the problem | The caller, into your org chart | The system, into your routing |
| The burst pipe at the top of this page | Presses 3, waits, explains, is transferred, explains again | Says "I've got a burst pipe" — recognised as urgent and put straight through with the details already captured |
| "Are you open Saturday?" | Occupies a menu branch and usually a person | Answered in eight seconds. Call over, caller satisfied, nobody interrupted |
| Ten people ring at once | All hear the recording, then queue | All ten answered and handled in parallel |
| Sunday, 9pm | Menu, then voicemail | Books the job into Monday |
| What you learn afterwards | Which buttons were pressed | What customers actually said, in their words |
One rule, and it decides whether customers accept the change
Anyone who asks for a person gets one — immediately, unconditionally, without having to justify it. The menu-era equivalent was pressing zero, and the businesses customers complain about are exactly the ones that hid it. Do not rebuild that mistake in a new medium. Counter-intuitively, an obvious exit reduces how often people take it, because it is being trapped that makes people fight the system rather than use it.
What You Get That a Menu Never Could
Questions actually answered
A menu can only move a call. It cannot tell someone you close at four on Saturdays — the highest-volume, lowest-value question most businesses field, and one that never needed a person.
Appointments booked
Checks real availability and writes into the actual diary, at 9pm on a Sunday. This is where the measurable return concentrates for most businesses.
Urgency recognised
A menu treats a burst pipe and a billing query identically because it cannot hear either one. Conversational routing escalates the first immediately.
Context that travels
A keypress is one digit. A conversation produces the name, the address, the job and the urgency — arriving with the call, so nobody asks twice.
Vague requests handled
"I spoke to someone yesterday about the thing at the Newcastle site" has no keypress. It is also exactly how people talk.
What customers actually want
Transcripts tell you what people said, what confused them and what a competitor quoted. The best customer research your business generates — and a keypress throws every word of it away.
For the Sceptic: Measure It Yourself
If you search this topic you will meet a wall of confident statistics — three quarters of customers frustrated by menus, two thirds abandoning inside ninety seconds. We are not going to quote them, and you should be wary of anyone who does.
| What those figures look like | What they usually are |
|---|---|
| The same percentages on dozens of pages | Published by companies selling AI voice agents, citing each other rather than a study |
| No sample size, year, country or method | Unverifiable, and possibly not about Australian businesses at all |
The argument does not need them, and your own numbers will convince a sceptical colleague far better than a vendor's will. Four to ask your platform for:
- Drop-off per menu prompt. How many callers hang up while the menu is playing, and at which option. Almost nobody has ever looked at this, and it is the most damning number available.
- Zero-out rate. The share who bail to an operator instead of choosing. High means your menu is not routing anyone — it is an obstacle people have learned to go around.
- Mis-routes. Calls transferred within sixty seconds of being answered. Those people chose wrong, which means the menu sent them wrong.
- Time to first human word. From answer to a person actually speaking. Use a stopwatch if the report will not give it to you.
Then do the four-minute version
Ring your own main number from your mobile at 9am on a Monday, as if you were a new customer with an awkward problem. Do not take the path you designed — take the path a confused first-timer would. This costs nothing and it has settled this question for more business owners than any report, because hearing your own menu is a different experience from thinking about it.
For context that is verifiable: ACXPA's 2026 self-reported Australian figures put voice abandonment at an average of 9% and a median of 5%, and its 2026 Best Practice Report found speed of answer ranging from a 227-second average in utilities to a 79-second median in banking and finance. Not menu statistics, but a real picture of the surrounding conditions.
The Callers Your Menu Hangs Up On
This one rarely appears in overseas guides and it matters here, because it affects a group of Australian callers who are being disconnected by a system that looks entirely fine to everyone else.
Callers using the National Relay Service communicate through a relay officer, which adds significant time to every turn. A menu that allows the usual few seconds for a keypress does not allow enough time for a relay-mediated reply.
| What your menu does | What that caller experiences |
|---|---|
| Plays the options, waits about five seconds, repeats once, disconnects | The relay officer is still conveying option two when the timeout fires. The call ends |
| Logs it as an abandoned call | They abandoned nothing. A timer cut them off |
| Looks perfect in testing | Because nobody tests through the relay service |
Two fixes
Keeping the menu: lengthen timeouts substantially, allow repeats, and never treat a slow reply as no reply. Replacing it: conversational answering removes the timed-keypress failure at the root, because there is no countdown to lose — but apply the same principle and make sure a pause is read as thinking rather than as silence.
The Fortnight Plan
The mistake is trying to recreate every branch of the old tree. You are not migrating a menu — you are replacing what it was there to do.
| When | What happens | Why |
|---|---|---|
| Days 1–3 | Pull a month of menu data: which options were pressed, and how many of those calls were transferred again within a minute | The transfers identify which branches were actively wrong. Usually one or two carry most of the damage |
| Days 1–3 | Separately, have whoever answers tally what callers actually ask for in their own words | This list rarely matches the menu. The gap between the two is the problem, finally visible |
| Days 4–5 | Write the answers to the ten most common questions, the way you would say them out loud | This is the configuration. A writing job for whoever knows the business |
| Days 4–5 | Write the escalation rules: what is urgent, who it reaches, what triggers an instant transfer | Build the human path first, before the automated one |
| Days 6–8 | Connect the diary. Set durations, buffers, and what must never be booked | Booking is where the return concentrates |
| Days 9–10 | Test it badly on purpose — mumble, interrupt, ask something odd, demand a person immediately, and test through the National Relay Service | Test the awkward paths, not the designed ones |
| Day 11 | Go live after hours only. The menu keeps running in business hours | Lowest-risk launch there is — after hours you are only competing with voicemail |
| Week 3 | Read the transcripts, fix the three worst answers, then extend into business hours | Extend on evidence, not on a schedule |
| Week 4 | Switch the menu off | In our experience nobody asks for it back |
4
Numbers to pull first
10
Answers to write
4 min
To ring your own menu
0
Branches worth copying over
When Keeping It Is the Right Call
We sell the alternative, so here is the honest version. Four situations where a short menu is still correct.
| Situation | Why the menu wins |
|---|---|
| Exactly two destinations, low volume | A simple sales-or-service split is fast and predictable. Replacing it buys you very little |
| Under about ten calls a day | The setup effort outweighs the gain. A good greeting and a missed-call text will do more for you |
| A legally scripted disclosure | Where an exact statement must reach every caller verbatim, a recording is the right tool. Play it, then hand to a conversation |
| A hard-wired emergency route | Keep one deterministic path to a human — not because AI cannot detect urgency, but because that path should not depend on anything understanding anything |
And the way this goes wrong
Replacing a menu people had learned to defeat with an assistant they cannot escape. Callers who have spent years pressing zero on reflex will try the equivalent in the first three seconds, and what happens then decides whether they accept the change or tell people about it. Make the human path obvious and instant, or you will have modernised your way to a worse experience.
The short version
Menus made sense when answering was scarce, and they worked by making the customer do your filing. Neither condition holds any more. Pull four numbers, ring your own line once from a mobile, and let that decide it — and if you are small with two destinations, keeping your menu is a perfectly good answer that nobody should talk you out of.
Related: can AI answer my business calls for the honest capability picture, which calls to automate for the categorisation method, and the five numbers your phone system should report for the measurement side.
Frequently Asked Questions
Why are phone menus so frustrating for customers?
Because of one root cause that produces every other problem: a menu is organised around your company while callers think in problems. Nobody rings accounts — they ring because an invoice looks wrong. Nobody rings service — they ring because water is coming through the ceiling. The menu asks the caller to perform a translation from their problem into your internal structure, using knowledge of your business they do not have, under time pressure, while a recording talks at them. Four failures follow. They choose wrong, not through carelessness but because the categories are yours and the problem is theirs, so they reach the wrong team and the wait restarts. They wait through options they do not need, because the list is linear and unskippable at the speed of the recording, so the regular caller who knows they want option four hears one, two and three every time. They have a problem nobody listed, so they land on all other enquiries, which is a shrug with a keypress attached. And they give information twice when the menu collects an account number that a person then asks for again. Reordering or re-recording the options fixes none of this, because the problem is the format.
Is it worth replacing an IVR with AI answering?
For most businesses with meaningful call volume, yes, and the reason is historical rather than fashionable. Menus were built to ration a scarce resource: with a fixed number of physical lines and one receptionist, sorting callers into buckets before a human touched them spent that scarcity well. On a cloud platform there is no fixed line count and ten callers can be answered on the first ring simultaneously, so the menu now rations something that has become effectively free while spending the thing that has not, which is your customer's patience. What replaces it is not a shorter or smarter menu but the removal of the translation demand: the caller says what they want in their own words, and the system answers routine questions outright, books appointments into a real diary, recognises urgency and escalates it, passes captured details forward with the call so nobody asks twice, and handles ten callers in parallel rather than queueing them behind one recording. The exception is a business with very low volume or genuinely two destinations, where a short menu remains fast, predictable and cheap.
Should I trust the IVR statistics I find online?
Generally no, and this is worth knowing before you build a business case on them. Search the topic and you will meet a set of very confident figures — roughly three quarters of customers frustrated by menus, two thirds abandoning within ninety seconds, a third of those never ringing back. Those same percentages appear across dozens of pages, almost all published by companies selling AI voice agents, and they cite each other rather than any named study. There is typically no sample size, no year, no country and no methodology, which means the figure cannot be checked and may not describe an Australian business at all. The argument against phone menus does not need them, and your own numbers will convince a sceptical colleague far more effectively. Where verifiable Australian data does exist it is worth having: ACXPA's 2026 self-reported figures put Australian voice abandonment at an average of 9% with a median of 5%, and its 2026 Best Practice Report found speed of answer ranging from a 227-second average in utilities to a 79-second median in banking and finance. Those are not menu statistics but they describe the real surrounding conditions.
How do I measure whether my phone menu is hurting me?
Ask your platform for four numbers, then do a four-minute test yourself. Drop-off per menu prompt tells you how many callers hang up while the menu is playing and at which option — almost nobody has ever looked at this and it is the most damning figure available. Zero-out rate is the share of callers who bail to an operator rather than choosing an option, and a high rate means your menu is not routing anyone, it is an obstacle people have learned to go around. Mis-routes, counted as calls transferred within sixty seconds of being answered, identify callers who chose wrong, which means the menu sent them wrong. Time to first human word, from answer to a person actually speaking, can be stopwatched if the report will not produce it. Then the test that settles it for most owners: ring your own main number from your mobile at nine on a Monday morning, as though you were a new customer with an awkward problem, and take the path a confused first-timer would take rather than the path you designed. Hearing your own menu is a materially different experience from thinking about it.
What happens to callers using the National Relay Service?
They are frequently disconnected by menus that look perfectly fine to everyone else, and it is one of the least-discussed problems with touch-tone menus in Australia. Callers using the National Relay Service communicate through a relay officer, which adds significant time to every turn of a conversation, so a menu allowing the usual few seconds for a keypress does not allow enough time for a relay-mediated response. In practice the menu plays its options, waits about five seconds, repeats once and then disconnects, while the relay officer is still conveying option two — so the call simply ends. The system then logs it as an abandoned call, when the caller abandoned nothing and was cut off by a timer. It looks perfect in testing because nobody tests through the relay service. There are two fixes. If you are keeping the menu, lengthen your timeouts substantially, allow repeats, and never treat a slow reply as no reply. If you are replacing it, conversational answering removes the timed-keypress failure at its root because there is no countdown to lose, though the same principle applies — a pause should be read as the caller thinking rather than as silence.
How long does it take to replace a phone menu with AI?
About a fortnight, and the mistake to avoid is trying to recreate every branch of the old tree, because you are replacing what the menu was for rather than migrating the menu itself. Days one to three, pull a month of menu data showing which options were pressed and how many of those calls were transferred again within a minute, since the transfers identify which branches were actively wrong. In parallel, have whoever answers the phone tally what callers actually ask for in their own words — that list rarely matches the menu, and the gap between them is the problem made visible. Days four and five, write the answers to your ten most common questions the way you would say them aloud, which is the configuration and is a writing job rather than a technical one, and write the escalation rules covering what is urgent, who it reaches and what triggers an instant transfer. Days six to eight, connect the diary with durations, buffers and what must never be booked. Days nine and ten, test badly on purpose, including through the National Relay Service. Day eleven, go live after hours only. Week three, read transcripts and extend into business hours. Week four, switch the menu off.
When should a business keep its phone menu?
Four situations, and this is the honest answer from someone selling the alternative. Keep it if you have exactly two destinations and low volume, because a simple sales-or-service split is fast and predictable and replacing it buys very little. Keep it if you take under about ten calls a day, since the setup effort outweighs the gain and a good greeting plus a missed-call text will do more for you. Keep a recording where a legally scripted disclosure must reach every caller verbatim, because a recording is the right instrument for that — play it, then hand to a conversation. And keep one hard-wired deterministic route to a human for emergencies, not because AI cannot detect urgency but because that particular path should not depend on anything understanding anything. There is also a specific way the transition goes wrong that is worth guarding against: replacing a menu people had learned to defeat with an assistant they cannot escape. Callers who have spent years pressing zero on reflex will try the equivalent within three seconds, and what happens then decides whether they accept the change, so the human path must be obvious and instant.
