# What is Telephony Integration?

Also called Phone Integration, SIP Integration, Programmable Voice.

Telephony integration is the connection between software and the telephone network so that a program can place calls, receive them, and control audio in real time. It typically uses a session protocol such as SIP or a provider API that abstracts it, together with phone numbers, call routing rules, and media streaming. This is the plumbing behind any automated phone line.

Signaling and media are separate concerns. Signaling sets up, modifies, and ends the call, while media carries the audio itself, usually as a stream of small packets. A provider platform hides most of this behind webhooks and a streaming endpoint, so an inbound call triggers an HTTP request to your application, and the response instructs the platform to play audio, gather input, bridge, or hang up.

Phone numbers are regulated inventory. Acquiring one usually requires an address or identity document in the relevant country, and rules about what may be sent from it differ by jurisdiction. Numbers are then mapped to routing logic so calls reach the right queue, application, or fallback, with explicit handling for busy, no answer, and out of hours conditions.

Real calls degrade in ways lab tests do not show: packet loss, echo, one way audio, and codecs that strip the frequencies speech recognition depends on. Transfers, hold, keypad tones, and voicemail detection each need deliberate treatment. Recording introduces a legal obligation as well, since consent requirements for recording calls differ by country and sometimes by region within one.

Because the phone network is old and standardized, its failure modes are well documented and largely predictable. That is an advantage. Call quality problems can be traced to a specific leg, a specific codec, or a specific carrier, and the diagnostic vocabulary already exists, which is rarely true of newer channels.

## Key points

- Signaling sets up the call; media carries the audio separately
- Inbound calls typically arrive as webhooks to your application
- Number acquisition and usage rules are jurisdiction specific
- Transfers, keypad tones, and voicemail need explicit handling
- Call recording consent requirements vary by region

## In practice

A property management company publishes one local number for maintenance requests. When a tenant calls, the telephony provider sends a webhook to the company's application, which answers, streams the audio to a voice agent, and logs the request against the property. If the description mentions a gas smell, the call is bridged to the emergency contractor's line and the transfer is written into the ticket.

## Related terms

- [Voice Agent](/en/glossary/voice-agent)
- [Interactive Voice Response](/en/glossary/interactive-voice-response)
- [Speech to Text](/en/glossary/speech-to-text)
- [Channel Routing](/en/glossary/channel-routing)
- [Voice Latency](/en/glossary/voice-latency)

[Back to the AI Glossary](/en/glossary)