Concierge Bot Setup — Route Visitors to Specialized Bots

What Is a Concierge Bot?

A concierge bot is a main "receptionist" AI that greets visitors, understands their intent, and routes them to the right specialist. Instead of one bot trying to do everything, you have a coordinator plus focused specialists. This is the recommended setup for most businesses with Shopify, BigCommerce, or CRM integrations.

Why Use a Concierge Setup?

• Each bot stays focused and gives better answers in its domain
• Visitors get routed to exactly the right capability
• You can give each specialist only the skills it needs (reducing risk)
• Easier to train and update — change one bot without affecting others
• Natural escalation: the concierge can always route to a human agent

How the Concierge System Actually Works

The concierge system uses AI Configs (created in Bots → AI Configs) and a delegation call stack. Here is what happens:

1. Your workflow has one AI node pointing at the Receptionist AI Config

2. The Receptionist greets the visitor and identifies their intent

3. If the visitor needs order help, the Receptionist calls delegate_to_orders_specialist — a tool that is auto-generated from the specialist's AI Config name

4. The Orders Specialist AI takes over, uses its skills, then calls return_to_previous when done

5. The Receptionist resumes and wraps up the conversation

The visitor experiences a seamless conversation. They never see the handoff between specialists.

How to Set Up a Concierge Bot

Step 1 — Create Specialist AI Configs

Go to Bots → AI Configs and create a separate config for each specialist role. The name you give each config determines its delegation tool name:

| AI Config Name | Auto-generated delegation tool |

|---|---|

| Orders Specialist | delegate_to_orders_specialist |

| Returns Desk | delegate_to_returns_desk |

| Product Advisor | delegate_to_product_advisor |

Write a focused prompt for each specialist. Tell it what it handles AND what it should NOT handle (so it delegates back correctly).

Step 2 — Set Up the Receptionist AI Config

Create your Receptionist AI Config. In the prompt, tell it to greet visitors and route them. You do NOT need to list the delegation tools in the prompt — the system injects them automatically based on which specialists you wire up next.

Step 3 — Wire Up Specialists in the Workflow Node

In your workflow, add a Transfer to AI node pointing at the Receptionist AI Config. In the right panel, find the Can Delegate To section and add each specialist:

• Select the specialist AI Config
• Fill in the "When" field: describe the topics that specialist handles (e.g., "orders, shipping, returns, tracking, delivery status")

The system injects this hint into the Receptionist's prompt so it knows when to delegate.

Step 4 — Set Permitted Skills Per Specialist (Optional)

In the Permitted Skills section of the workflow node, restrict which skills each specialist can use. Leave empty to allow all site-enabled skills, or pick specific skills to lock down what that specialist can do.

Example: Shopify Store Concierge

Receptionist AI Config

Prompt: "You are a helpful assistant for [Store Name]. Greet the visitor, identify whether they need help with an order, a product question, or a return, then route them to the right specialist. Do not look up orders or products yourself."

Orders Specialist AI Config

Prompt: "You help customers track orders and check delivery status. Look up orders by email or order number. For refunds, let the visitor know you'll connect them with the Returns specialist."

When hint: "orders, shipping, tracking, delivery status, where is my order"

Returns Desk AI Config

Prompt: "You handle returns and refunds. Verify the order, confirm eligibility, then process the refund. Always confirm the amount before proceeding. Do not process refunds without explicit customer confirmation."

When hint: "returns, refunds, cancellations, send back, wrong item"

Product Advisor AI Config

Prompt: "You help visitors find the right product. Ask about their needs and recommend options. You do not handle orders or payments."

When hint: "product questions, recommendations, what should I buy, features, comparison"

Example: BigCommerce B2B Concierge

Receptionist delegates to:

• Quote Bot (when: "pricing, quotes, bulk orders, price lists")
• Account Bot (when: "my account, invoices, order history, billing")
• Procurement Bot (when: "products, stock, availability, specifications")

Example: CRM + Commerce Combined

Receptionist delegates to:

• Sales Bot (Shopify product skills + HubSpot deal creation; when: "interested in buying, product demo, pricing")
• Support Bot (Shopify order skills + HubSpot ticket creation; when: "problem, issue, broken, damaged, complaint")
• Account Bot (HubSpot contact/company lookup, read-only; when: "my account, contact info, company details")

How Many Specialists Should I Have?

Most businesses do well with 2–4 specialists:

• 1 Receptionist (routes, identifies, greets — no skills needed)
• 1–2 Commerce specialists (orders/returns split, or combined)
• 1 Support/CRM specialist
• Human escalation via Transfer to Agent

Avoid going beyond 5–6 specialists — visitors may bounce between them too many times before getting help.

What If a Specialist Cannot Resolve the Issue?

If a specialist cannot fully resolve the issue, it automatically returns to the Receptionist (via return_to_previous). From there, the Receptionist can try another specialist or escalate to a human agent. You can also configure each AI Config's Escalation Rules to transfer directly to an agent queue when needed.

Common Mistakes to Avoid

• Do not give the Receptionist any skills — it should route, not resolve
• Do not give the Returns bot discount-creation skills — separate permissions by role
• Make each specialist's prompt clear about what it does NOT handle
• Always test by typing common visitor phrases in the bot preview panel
• The "When" hint field is what the Receptionist reads to decide who to delegate to — write it like a tag list, not a sentence