Abigel.ai: a multi-tenant voice AI platform from idea to operations

The business problem

The same request kept coming up in our customer-service projects: part of the inbound calls, appointment booking, opening hours, simple status checks, leaving a message, should be handled automatically, around the clock and in several languages.

Speech recognition, large language models and speech synthesis made this technically possible. What was missing was a platform that connects them to the ordinary phone network and that a small business can set up itself, without a developer.

What we built

• A no-code flow editor: the call flow is assembled visually in the browser (greeting, questions, branches, data collection, handover to a human).
• A real-time STT–LLM–TTS pipeline, streaming at every step, with barge-in handling.
• SIP and PSTN integration on our own PSTN-bridge component: it takes and places calls on real phone numbers.
• A call log with transcripts and summaries, and CSV export into existing systems.
• A second product on the same voice pipeline: a conversational AI language teacher.

Multi-tenancy and billing

One infrastructure serves every customer, with data, configuration and billing isolated per tenant: a shared schema, row-level security and per-tenant limits.

Pricing is usage-based: no prepaid plan, customers pay for the calls actually handled. Billing reuses the usage-event model proven in Abisearch. Sign-up is self-service.

Launch and operations

A Docker-based environment, automated tests, separate staging and production. Call flows are also tested automatically with real SIP calls.

Recordings, transcripts and personal data are stored encrypted on servers inside the EU, with configurable retention. ProCat Solutions operates and develops the system.