For most of the telephone's history, talking to a business meant talking to a person — or to a machine that wanted you to press a number. The leap to systems that understand and reply in natural speech took far longer than anyone predicted. A quick tour of how we got here.
Operators and switchboards
Early telephony was human end to end: an operator connected your call and often knew the person you were calling. The service was personal, slow and impossible to scale.
Touch-tone and the IVR era
Tone dialling made it possible for machines to route calls, and the interactive voice response menu was born. It cut costs and created the most disliked customer experience of the twentieth century: 'press one for bookings'.
Speech recognition arrives
Early recognition could handle a few words in ideal conditions. Systems improved steadily but still forced callers to speak in the machine's terms — short commands, fixed vocabulary, no interruption.
Why conversation was hard
A conversation needs recognition, understanding, memory, decision-making and natural speech, all within a few hundred milliseconds and all robust to accents, noise and the unexpected. Each piece existed separately for years before they were fast and good enough to work together.
Where we are now
Assistants that listen while they speak, remember the whole call, consult your systems mid-sentence and sound like a person. The remaining frontier is not the conversation itself but connecting it deeply enough to every system a business runs.
The telephone made distance disappear. Conversational AI is finally making the queue disappear.
