Apisrom
Client
AI TRANSFORMATION
Apisrom
Voice-capable AI concierge for conversational commerce on WooCommerce.
eCommerce
Industry
Conversational AI & ERP Integration
Services
Romania
Region
2025-present
Year
Commerce that talks back.
Apisrom sells premium honey and bee products through a physical storefront and a WooCommerce site. They wanted conversational commerce on the online store. Not a chatbot. A voice-capable agent that could take an order in natural Romanian, the way a customer would ask for a jar of honey at a market stall. For a brand built on provenance and craft, voice reinforces the positioning.
The desired outcome: a customer lands on the site, clicks the chat bubble, and either types or talks to place an order end to end, in their own language, without ever leaving the storefront.
Voice-first
A chatbot would not fit
Provenance
Brand built on craft
No handoff
Orders end to end on-site
Conversational
Natural Romanian, spoken
One knowledge base, ready for phone and WhatsApp next.
A voice-capable AI concierge drops into the WooCommerce site as a single embeddable widget. The customer can speak to it and the agent speaks back, or they can type. They can interrupt mid-sentence and edit input on the fly. The agent retrieves from a dedicated OpenAI vector store built on Apisrom's own catalogue: every variant, current pricing, delivery rules, and brand tone.
The agent does real commercial work, not just Q&A. It collects the order, calculates the total live with delivery fees applied, and on confirmation fires two templated emails from one structured payload: a notification to the shop and a confirmation to the customer, both in their native language, driven by per-tenant YAML templates.
A drop-in widget that lets customers type or talk.
01.
Two-way voice
Streaming speech-to-text and voice output via ElevenLabs, with a Romanian pronunciation normaliser so weight units and price formats are spoken correctly.
02.
Owned knowledge base
Retrieves from an Apisrom-owned OpenAI vector store: every product variant, price, delivery rule, and the brand's tone of voice. One source of truth.
03.
End-to-end ordering
Collects customer type, products, addresses, and contact, calculates the total live with delivery rules, and submits on confirmation.
04.
Order notifications
Two templated emails from one payload, shop notification and customer confirmation, both localised, driven by per-tenant YAML.
Let your customers order by just talking
We gave Apisrom a voice assistant that takes orders in natural Romanian, right on their store. If you want AI that serves your customers, not a basic chatbot, this is the kind of thing we build.
Starts with a free 30-minute discovery call.
Next case study
An AI surface for a dental group: a phone receptionist, a self-improving knowledge engine, and a vision model the clinic owns.