Concierge

What you get

A chat icon on your website that can handle real questions, hold your brand voice, and call your business systems through governed connections. Optional voice mode where visitors can talk to it instead of type. Optional mobile SDK so the same agent lives in your app. One persona, many surfaces, all under your control.

What makes it different

Concierge isn't a generic chatbot you bolt on. It reads from your data, talks like your brand, calls your tools, and routes every conversation that doesn't end in a clean answer — a ticket gets created, an email lands in the right inbox, a human is paged with the full context ready. When you swap the AI model behind it for something better next quarter, your team changes a config; the customer-facing experience stays the same.

What it can do for you

Answer questions grounded in your actual product documentation, return policy, or knowledge base. Look up an order, an account, a reservation when the visitor is signed in. Take an action — create a return, book an appointment, escalate to your team — when it has the scope and approval to. Speak the language of the site it's on. Stay polite when it's pushed and refuse cleanly when something is out of scope.

• Customizable chat widget — your colors, your logo, your tone
• Optional voice mode in the browser, with reactive Orb visualization
• Grounded in your data — no generic answers, no hallucinated promises
• Connects to your business tools through governed connections
• Creates a ticket / sends an email / hands off to a human when needed
• Knows the visitor when they're logged in
• Multi-language with native tone, not literal translation
• Per-session cost budgets and a kill switch you control

Where it shows up

On your website as a script tag or React/Vue/Svelte component. On your mobile app via iOS, Android, or React Native SDKs. As a Web Component for any framework. In a sandboxed iframe for strict security policies. The same approved persona and capability set across every surface.

Who runs it

You do. There's an admin console where your team reviews proposed capabilities, approves what goes live, sets routing rules, watches dashboards, and pauses the whole thing with one click if something looks wrong. We set it up; you operate it; we're available when you want to expand.

How we ship it

A typical first deployment is 8–12 weeks: connect your knowledge sources, install the widget, approve the first batch of capabilities, ship behind a feature flag, expand. Voice and additional channels come in v2. We stay close as the system goes live.

Behind the scenes

Concierge is the packaged, customer-facing form of our Embedded Agents service. It runs on the same Foundations and Components that power every Group e-media engagement — Data Foundations, AI Platform, MCP Tool Registry, Agent Runtime, Synthetic Personality, Conversation Intelligence. You don't have to know that to use it. But when you want to know how it works underneath, the answer is documented and the architecture is yours.