What a PBX Actually Is
A private branch exchange is a switch that sits between the telephones inside an organisation and the public telephone network outside it. It does two jobs, and it has done the same two jobs since the 1890s.
Internal switching
It connects one internal telephone to another without touching the public network. Extension 204 rings extension 211, and no external line is used and nothing is charged. This is the "private" and the "branch" in the name.
Shared external access
It lets many internal telephones share a smaller number of external paths. Forty staff, six simultaneous outside calls. Everything about how businesses buy phone service still follows from this ratio.
Everything else, added later
Transfer, hold, hunt groups, pickup, paging, voicemail, a receptionist console, call queues, then recording, then reporting, then automated attendants. All of it accreted on top of those first two functions.
The reason this matters in 2026 is that the function has not gone away — only the location has changed. When someone says "we do not have a PBX, we are on the cloud", what they mean is that the exchange is now software in a data centre rather than a cabinet in a comms room. Internal extensions still ring each other for nothing, and external calls still share a pool of capacity. The concepts you will be quoted against are the same concepts.
Terminology note that saves confusion later. "PBX" describes the switching function. It does not tell you where the switch runs, who owns it, or how you pay for it — and those three things are the entire commercial decision. Any sentence of the form "you need a PBX" or "PBX is obsolete" is missing the only part that matters.
The Vocabulary You Will Meet in Every Quote
These terms all predate the internet and all still appear on 2026 invoices. Knowing them makes quotes comparable.
| Term | What it means | Why it still matters |
|---|---|---|
| Extension | An internal number identifying one endpoint — a desk phone, a softphone, a mobile app, or a person across all three. | Frequently the billing unit. Ask whether a person with a desk phone, a laptop app and a mobile app counts as one extension or three. |
| Trunk | A path to the outside world. Historically a physical line; now almost always a SIP trunk, which is a logical channel over your internet or a provider's network. | Some quotes charge per trunk. Understanding this is how you avoid buying capacity for a peak you do not have. |
| Concurrent calls | How many conversations can be in progress at once. The modern equivalent of "how many lines do we have". | The number that determines whether callers hear engaged tone at 9am. Usually far lower than headcount — see the sizing note below. |
| DID / DDI | Direct inward dialling. An external number that rings a specific extension without going through reception. | How each staff member or department gets a public number. Often charged separately per number. |
| Hunt group / ring group | Several extensions rung together, in sequence or simultaneously. | The simplest form of call distribution, and often sufficient. A queue is the version with waiting, announcements and reporting. |
| Auto attendant / IVR | The recorded menu that answers and routes. | Still standard, increasingly replaced by an AI answer that skips the menu entirely. Our argument on that is here. |
| SIP | Session Initiation Protocol — the signalling standard modern systems use to set up and tear down calls. | The reason handsets from different manufacturers work on the same platform. Covered in the SIP trunking guide. |
Sizing concurrent calls, since everyone asks
For a general office where the phone is not the primary job, concurrent call requirements are typically a small fraction of headcount. For a team whose whole job is the phone — a support desk, a booking line, a collections team — plan closer to one path per person actively on calls, plus headroom. The mistake to avoid is sizing from headcount, which over-provisions a normal office badly and under-provisions a contact centre badly. Look at your own busiest fifteen minutes, not at a rule of thumb.
The Five Architectures
They form a sequence, and each one moved a specific responsibility from the customer to a provider. That is the whole story: not features, but who is on the hook for what.
1
Traditional PBX — your hardware, your comms room, physical lines.
2
IP PBX — your hardware, your comms room, calls over data.
3
Hosted PBX — the same software, in a provider's data centre.
4
Cloud PBX — a shared platform built to be run as a service.
5
UCaaS / CCaaS — exchange functions inside a wider platform.
1. Traditional PBX
A cabinet on your premises with cards in it, wired to physical extensions and to physical external lines — analogue, ISDN in the past, or digital. Programming is done on the box, by someone who knows that box.
| What you own | What you are responsible for | Where it fails |
|---|---|---|
| The hardware outright, usually as a capital purchase, plus the cabling and often the handsets. | Power, environment, cabling, spares, the maintenance contract, and finding somebody who still works on that model. | Parts and skills. As platforms age, both become scarce, and the underlying line types have been progressively withdrawn — ISDN in Australia ceased sale in 2018, disconnections began in 2019, and it was fully decommissioned by mid-2022. |
Traditional systems are still in service, and some still work perfectly for what they do. What has ended is the environment around them: the line types they were designed for, the technicians, and the spare parts. If you have one, the question is not whether it works — it is what happens the week it does not. The historical arc is set out in PSTN to ISDN to VoIP to voice over cloud.
2. IP PBX
Still on your premises — a server or an appliance in your comms room — but calls are switched as data over your network rather than over dedicated voice wiring, and external calls usually leave over SIP trunks instead of physical lines. Handsets are IP phones on the same network as your computers.
| What changes from traditional | What does not |
|---|---|
| One cabling system instead of two. Software features rather than hardware cards. Remote extensions become possible. Trunk capacity becomes a configuration change rather than a truck roll. | You still own and run the equipment, still maintain it, still patch it, still hold the security exposure, and still lose service if the site loses power, internet or the box itself. |
The part that gets underestimated
An IP PBX is an internet-facing server running voice software, and it is a standing target. Toll fraud against exposed systems is a persistent, automated business — attackers scan for reachable SIP services, brute-force weak extension credentials, and place expensive international traffic through them, often over a weekend. Owning the box means owning that risk: firmware patching, credential hygiene, firewall rules, international dialling restrictions and monitoring. That is real, ongoing work. See protecting a phone system from fraud.
3. Hosted PBX
The same category of software, moved off your premises into a provider's data centre. Frequently a dedicated instance for your business. You configure it through a web interface; the provider handles the hardware it runs on.
In practice this is the transitional model, and it is often the right one for a business with unusual requirements: a hosted instance can be configured in ways a shared platform cannot, and the migration path from an existing on-premises system is short because the concepts map one-to-one.
| Moves to the provider | Stays with you |
|---|---|
| Hardware, power, environment, backup, the data centre's connectivity, and usually operating-system patching. | Configuration, dial plan design, user administration, and — importantly — the internet connection at each of your sites, which is now the single path your calls take. |
The comparison against keeping the box is the subject of hosted PBX versus on-premise PBX, which goes into the cost modelling in more detail than belongs here.
4. Cloud PBX
A multi-tenant platform: one large system serving many businesses, each with its own isolated configuration. You are not given an instance to run — you are given a service. Provisioning is minutes, capacity is elastic, upgrades happen underneath you, and there is nothing at your site except handsets and apps.
What genuinely improves
No hardware, no patching, no maintenance contract. Add a person in minutes. Handsets provision themselves. Work from anywhere without a VPN. Numbers, apps, recording and reporting all in one place.
What you depend on instead
Your internet at every site, and the provider's network and platform. The dependency does not disappear; it moves — and it becomes somebody whose full-time job it is rather than yours.
What you give up
Deep, unusual customisation. A shared platform offers configuration, not modification. If your business genuinely requires behaviour the platform does not support, this is where you find out.
The question to ask
Whose network and whose data centres? A cloud PBX operated on infrastructure the provider runs is a different accountability proposition from one resold on top of somebody else's platform — the distinction we set out in not all providers are equal.
The general explainer for this model, aimed at readers who have not decided yet, is what is a cloud phone system.
5. UCaaS and CCaaS
These are not PBX types so much as platforms that contain one. Unified communications as a service bundles calling with messaging, presence, meetings, and integration into the applications people already work in. Contact centre as a service adds queueing, skills-based routing, workforce and quality tooling, and reporting built for a team whose job is the phone.
| What it adds beyond exchange functions | Who needs it | |
|---|---|---|
| UCaaS | Chat, presence, video meetings, file sharing, mobile and desktop apps, and CRM integration so calls arrive with context. | Almost every business with more than a handful of staff, because these are now the default expectations rather than extras. Explained in what is UCaaS. |
| CCaaS | Queues with announcements and callback, skills routing, real-time wallboards, service level and abandonment reporting, quality scoring, omnichannel handling. | Any team where callers wait in a queue and someone is accountable for how long. Covered in the CCaaS piece and the metrics guide. |
Why the naming still causes trouble. A vendor selling UCaaS will sometimes say "you do not need a PBX any more". What they mean is that you no longer need a box, because the exchange function is included in their platform. The function is not optional — extensions still have to ring each other and calls still have to reach the outside world. So the question is never PBX or cloud. It is where does the exchange run, who operates it, and what else comes with it.
Side by Side
| Traditional | IP PBX | Hosted | Cloud | |
|---|---|---|---|---|
| Where it runs | Your comms room | Your comms room | Provider data centre, your instance | Provider platform, shared |
| Who owns it | You | You | Provider owns hardware, you hold the instance | Provider |
| Up-front cost | High capital | Moderate capital | Low | Very low — handsets and porting |
| Ongoing model | Maintenance contract plus line rental | Support plus SIP trunks | Per extension or per instance | Per user or per extension |
| Who patches it | Your technician | You or your IT provider | Provider, usually | Provider, invisibly |
| Adding a person | Hardware and a site visit | Licence and configuration | Configuration | Minutes, self-service |
| If the site loses internet | Calls continue on physical lines | External calls stop; internal may continue | Site is offline; calls divert in the cloud | Site is offline; calls divert in the cloud, apps keep working on mobile data |
| If the site loses power | Depends on line type and battery | Everything stops | Site stops; diversion still works | Site stops; diversion still works, mobile apps unaffected |
| Remote and hybrid work | Not practically | Possible, with effort | Yes | Native |
| Multi-site | A system per site, tied together | Possible, complex | Straightforward | One system, many sites |
Read the two outage rows together
They are the most misused rows in every comparison. A traditional system genuinely does keep working through an internet outage — on lines that in most cases no longer exist. A cloud system genuinely does stop working at a site that has lost its internet, and simultaneously continues to answer, divert to mobiles and take messages, because the exchange is not at the site. Neither is "more reliable" in the abstract; they fail differently, and the honest comparison is between an outage that takes your calls off the air and an outage that reroutes them. That is the substance of how phone redundancy actually works.
How the Pricing Units Differ
Comparing quotes across architectures fails when the units differ, which they always do. Normalise before comparing.
| Unit | Where you see it | What to watch |
|---|---|---|
| Per extension | Hosted and some cloud platforms. | Whether one person's desk phone, laptop app and mobile app are one extension or three. This single question can double a quote. |
| Per user | Cloud, UCaaS. | What counts as a user. A shared warehouse phone, a lobby handset, a conference room and a door intercom are endpoints, not people — ask how each is charged. |
| Per concurrent call | SIP trunking, some hosted models. | Cheap if sized from real data, expensive if sized from headcount, and a source of engaged tone if sized from optimism. |
| Per number | All models. | DIDs, 1300/1800 numbers and inbound minutes are often separate lines on the quote. A 1300 number's inbound cost depends on where the caller is calling from. |
| Capital plus maintenance | Traditional and IP PBX. | The comparison needs the hardware refresh cycle in it. A five-year total with a replacement at year six flatters the on-premises case. |
The dollar figures, with worked examples for several business sizes, are in how much a business phone system costs in Australia.
Who Is Responsible for What
This is the practical difference between the models, and it is worth writing down explicitly before you sign anything.
| Responsibility | On-premises | Hosted | Cloud |
|---|---|---|---|
| Hardware failure | You, via a maintenance contract you must have | Provider | Provider |
| Security patching of the voice platform | You or your IT provider | Provider, usually — confirm it | Provider |
| Toll fraud exposure | Yours to prevent and, commonly, yours to pay for | Shared; ask where the line is | Provider monitors; you still set international dialling policy |
| Emergency call address accuracy | Yours, as configured | Yours to maintain, provider to deliver | Yours to maintain, provider to deliver — and it matters more when staff are mobile |
| Numbers and rights of use | Held with your carriage provider | Depends on who ordered them | Depends on the provider — ask directly |
| Getting your configuration out | Yours, and legible | Usually exportable | Varies enormously — ask before you need it |
Two rows worth pressing on
Emergency calling. A cloud extension can be anywhere, and the address associated with a number is only as accurate as the record you maintain. This is a real obligation with real consequences, and it is covered in how Triple Zero works from a cloud phone. Rights of use over numbers. Whether your main number is yours or your provider's decides whether you can leave. Ask before you sign, not at the point of leaving — see who owns your 1300 number.
When On-Premises Still Makes Sense
It would be convenient to claim there is no remaining case for equipment in a comms room. There is, and it is worth stating honestly, because a business in one of these situations will be poorly served by a recommendation that ignores it.
| Situation | Why on-premises may still fit | What to check first |
|---|---|---|
| No usable internet | A site with genuinely inadequate or unreliable connectivity and no fixed-wireless or satellite option cannot depend on a cloud platform for internal calling. | Whether that is still true. Fibre availability and business-grade fixed wireless have changed a great deal, and the answer from three years ago is frequently out of date. |
| Integration with site systems | Nurse call, industrial paging, lift phones, gate intercoms, legacy analogue devices, and equipment with a serial or contact-closure interface. | Most of this can be bridged with a gateway. Some cannot. Get a device-by-device answer, not a general assurance. |
| Regulatory or contractual constraints | A specific obligation about where a system physically resides, occasionally in defence, critical infrastructure or under a customer contract. | Whether the obligation is about location or about control. Australian-hosted infrastructure satisfies more of these than people assume — see the data sovereignty piece. |
| Recent capital investment | A system bought two years ago with life left in it. Replacing working equipment early to save monthly cost usually does not add up. | The end-of-support date, the parts position, and whether the line types it depends on are still available. Plan the transition; do not force it. |
Outside those four, the case has become hard to make for most businesses under a few hundred staff — not because cloud is fashionable, but because the responsibilities in the table above are real work, and few organisations of that size want to be in the business of patching a voice server.
Eight Questions That Identify a Quote
Ask these and the architecture behind any proposal becomes obvious within a few minutes.
| Question | What the answer reveals |
|---|---|
| 1. What equipment will be on my premises when this is finished? | Handsets and a router means cloud or hosted. Anything else means you still own a system. |
| 2. Is the platform shared with other customers, or an instance for me? | Separates cloud from hosted. Both are legitimate; they behave differently on upgrades and customisation. |
| 3. Who operates the network the calls traverse? | Separates an operator from a reseller. Decides who can actually investigate a fault rather than escalate it. |
| 4. What exactly is the billing unit, and how are shared devices counted? | Makes quotes comparable. Ask specifically about a person with three devices, and about lobby, warehouse and meeting-room phones. |
| 5. What happens to inbound calls when my office internet goes down? | A good answer describes automatic diversion and mobile apps continuing. A vague answer means nobody has configured it. |
| 6. Who will hold the rights of use over my numbers? | Decides whether you can leave. Get it in writing. |
| 7. Who patches the platform, and who is liable for toll fraud? | The two responsibilities most often left undefined, and the two most expensive to get wrong. |
| 8. If I leave in two years, what do I get back? | Numbers, recordings, call history, configuration. Ask now; the answer at the point of leaving is worse. |
Choosing, in Practice
For most Australian businesses the decision has collapsed into a narrow set of choices, and it is decided by size and by how central the phone is to the work.
| If you are | The usual answer | Why |
|---|---|---|
| A sole trader or a team of two to five | Cloud, with mobile apps and one shared number | Nothing to maintain, no capital, and a business number that is separate from a personal mobile. Nothing else justifies its overhead at this size. |
| Five to fifty staff, one or two sites | Cloud, with queues and a proper after-hours path | The exchange functions are all present, hybrid work is native, and the administrative work is a web page rather than a technician. |
| Fifty to two hundred, multiple sites | Cloud, or hosted if you have genuinely unusual requirements | One system across sites removes an enormous amount of duplication. Hosted remains the answer where a shared platform cannot express what you need. |
| A team whose entire job is the phone | Cloud with contact centre capability | Queueing, routing and reporting are the product at that point, not an add-on. Exchange functions are assumed. |
| A site with no viable internet, or heavy legacy site integration | On-premises or a hybrid, deliberately | The four situations in the section above. Decide it explicitly and revisit it, rather than defaulting into it. |
The summary
A PBX is the switch that lets internal extensions call each other and share a smaller number of external paths. That function is in every business phone system; what changed is where it runs and who is responsible for it. Traditional and IP PBX are yours to own and maintain; hosted moves the hardware to a provider; cloud makes it a service; UCaaS and CCaaS wrap it in a broader platform. Compare on responsibility and pricing unit, not on feature lists — the features converged years ago. Ask what is left on your premises, who operates the network, what the billing unit counts, what happens when your internet drops, who holds your numbers, and what you get back if you leave. On-premises is still defensible in four specific situations, and outside them it is mostly inertia with a maintenance contract attached.
Related reading: hosted versus on-premise PBX, what is a cloud phone system, SIP trunking explained, what it all costs, and what UCaaS means.