The room I worked in as a VoIP administrator had a sound. Not the phones — the PBX cabinets. Fans, disks, the deep hum of hardware that had been running since before I was hired. When people say "the phone system," they used to mean that room.
Today the equivalent is a dashboard and an API. This essay is about what changed, what didn't, and why the second list matters more.
What actually changed
Where the intelligence lives. A PBX concentrated everything: call control, voicemail, IVR, billing records. A SIP trunk plus a cloud PBX splits those jobs across systems owned by different people. Integration — once a cable — became a contract.
Where failure lives. When the cabinet died, the failure was in the room. When a cloud telephony integration misbehaves, the failure can be in your config, the carrier's routing, a DNS record, or a certificate that expired at 2 a.m. on a long weekend. Failure didn't get rarer; it got distributed.
Where the leverage is. Provisioning a new extension once meant a ticket and a technician. Now it's an API call. When configuration became programmable, everything downstream of configuration became automatable.
What didn't change
Dial tone is a promise. Users don't care about your architecture. They pick up a phone and expect a tone, and their call is expected to connect. That expectation was old when the PBX was new, and it survives every platform shift since. Whatever you build, the promise is identical.
Capacity is seasonal. Call volumes spike at the same moments business spikes. The cloud made elasticity a purchase instead of a warehouse, but the discipline of planning for the busy hour is the same discipline — the math just runs in a different ledger.
Voice is unforgiving. Data can retry. A dropped word is gone. Jitter budgets, codec choices and QoS are the cloud era's version of a clean copper pair: invisible until wrong.
The seam I built a career on
I entered telecom at the moment voice was becoming software — VoIP administration in 2007 — and moved into cloud infrastructure just as voice finished becoming someone else's managed service. That seam, between the century of hardware telephony and the decade of cloud communications, is the most interesting place I've worked.
The engineers who only know the new stack trust it too much. The engineers who only know the old stack distrust it too much. The useful position is both: the trunk is a cable, and the cable is a contract.