Streaming voice
Speech recognition and voice generation stream during the conversation. Response time depends on the configured services and connection.
AI voice agents for incoming calls and follow-ups.
Answer calls and call customers back in their language. Capture requirements, confirm part numbers and save a summary to the customer record.
Available capabilityA site foreman calls outside office hours. Terramine Desk introduces itself as an AI assistant, provides the recording notice and captures the machine, part number, required date and delivery location.
Speech recognition and voice responses are streamed to support a natural conversation. After the call, a summary is saved to the customer record.
Included capabilities
Turn-taking and interruptions
The voice pipeline distinguishes short acknowledgements from interruptions and tracks which agent words were played. Review the recording when checking recognition or timing errors.
Acknowledgements while processing
Short acknowledgements recorded in advance, and the opening line ready while the phone is still ringing. If the model is slow, the agent says so instead of going silent.
Silence and hold handling
Silence prompts and hold handling help manage pauses. Timing depends on call state and the configured voice settings.
Voicemail and switchboards
Detect supported voicemail and switchboard patterns, navigate an outbound phone menu and leave a message when enabled. Detected menu-only calls are separated from successful conversations.
Call completion detection
Conversation and silence checks help decide when to end a call. Operators can also end a call from the dashboard.
Call summary
After a conversation, AI prepares an outcome, summary and extracted details for review. Processing requires the configured AI services; the recording and transcript remain the sources to check.
Returning caller context
A customer who rings back gets the same agent in the same voice, and it already knows what was said last time.
Voice and language settings
Terramine Desk's default voice, the reception desk's and the answering machine's are picked from your voice library by name, with a sample to listen to; the pickup greeting, the voicemail message, the languages the recogniser expects and the brand names it should know are on the same page, in Admin.
The guided demo illustrates a selected workflow. Availability depends on the enabled configuration; beta items are labelled above.