Giizo AI is not a tool for assembling empty chatbots. It puts a sales agent that knows your products and is ready with the right tools live in about 5 minutes. The flow is three steps; the platform supplies the rest.
1. Create your agent
Sign in to the Giizo AI panel and pick a persona — the E-Commerce Sales Agent that sells your products, or the Customer Support Agent that runs everything after the sale.
- Behavior rules, tone and system prompt come ready
- The MCP tool set e-commerce selling needs is pre-selected
- Knowledge-base templates and proactive scenarios arrive with the persona
You are not configuring a blank assistant; you are hiring a digital worker that is ready to sell.
2. Add your data
Connect your own data so the agent knows your business:
- Website URLs, PDFs and business rules → Knowledge Base (RAG)
- Product/service catalog (including variants, stock and price)
- Optionally connect order or CRM systems via MCP
Answers are grounded in your data, not generic model knowledge.
3. Go live
The same agent runs across six channels at once:
- WhatsApp — text customer communication via WhatsApp Business API
- Instagram — text sales and support from DMs
- Facebook Messenger — text messages to your business page
- Web Widget — one-line embed;text or voice
- Web Page (Standalone) — shareable URL / QR;text or voice
- Robot —voice AI on physical devices in a store, showroom or office (Raspberry Pi and compatible devices)
WhatsApp, Instagram and Messenger are text-based; voice conversation is available on Web Widget, Web Page and Robot. You do not build a separate agent per channel — one persona, consistent everywhere.
The agent runs work (it does not only reply)
Once live, the digital worker:
- Receives the customer message and builds context from the knowledge base / catalog
- Takes real action in external systems via MCP when needed (when the store is connected: query an order and track a shipment; log a return request)
- Plans multi-step requests with task orchestration and executes steps in sequence
- Acts on its own through proactive triggers (cart abandonment, silent customers, and similar events)
- Hands off to live support in one click when it cannot resolve — context is preserved
Track performance
In the panel you see conversation quality, channel mix and persona-specific outcomes (sales, returns, tickets). The agent learns; the knowledge base grows under admin control.
In short: Pick a persona → connect your data → open a channel. In 5 minutes a digital worker that knows your products starts selling.