A local services business (e.g. a solicitor's office) needed calls covered outside normal business hours and during overflow periods, without hiring extra reception staff, and without it feeling impersonal to callers.
- 01Framing choice
Deliberately positioned to the client and their callers as an 'AI call assistant,' not 'voice AI', a framing decision that matters for trust with a non technical, price sensitive audience.
- 02Scheduled coverage
The assistant only picks up during defined windows (e.g. weekends, after hours), with routing options considered for the existing phone number: scheduled VoIP routing rules, manual call divert codes, or full number porting, chosen based on the client's existing phone setup.
- 03Call handling
The assistant answers, takes the caller's details and reason for calling, and logs everything.
- 04Daily reporting
A summary email of the day's calls is sent automatically, plus a simple dashboard view.
- 05Commercial model
Priced as a low one off setup fee plus a per minute usage rate, calculated with a healthy margin over the underlying telephony and AI costs.
After hours and overflow calls get answered and logged without any addition to headcount, at a cost structure that scales with actual usage.
Originally built for the local services sector, the same architecture applies to any team facing a similar volume, compliance, or coordination problem. The stack (ElevenLabs, Twilio, Automated email reporting, and others) is stable, self hosted where it matters, and designed to be operated by a small team without a platform crew.
- AI call assistant for small business
- after hours call answering AI
- weekend call coverage automation
- AI receptionist for local business
- ElevenLabsSpeech synthesis for the voice agent. Chosen for latency and voice quality on production calls.
- TwilioTelephony carrier. Handles inbound and outbound PSTN, SMS, and recording.
- Automated email reportingPart of the production stack for this build.
- Simple dashboardPart of the production stack for this build.
If I rebuilt this today I would move error handling and observability in from day one rather than bolt them on once the system was in production. Everything else was fine, but that would have paid back within weeks.
Why we don't call it 'voice AI' to every client. A positioning and marketing savvy piece about framing AI tools for non technical audiences.
How is this priced?+
A low one off setup fee plus a per minute usage rate, calculated with margin over the underlying telephony and AI costs. Usage based means the client only pays for actual call volume.