happy-technologies-llc/multi-turn-ordering

Multi-Turn Catalog Ordering

Overview

This skill configures conversational multi-turn ordering flows where users can browse, select, and order catalog items through a guided dialogue. It covers:

• Setting up Virtual Agent topics for catalog ordering conversations using sys_cb_topic
• Configuring conversational item discovery and selection from sc_cat_item and sc_category
• Designing multi-turn variable collection flows that gather item_option_new values step by step
• Implementing cart management and order placement via the Service Catalog API
• Handling disambiguation when multiple items match a user's request
• Managing conversation context across turns for complex ordering scenarios

When to use: When organizations want employees to order catalog items through a chat-based interface (Virtual Agent, Slack, Teams) rather than navigating the traditional catalog portal, or when items require guided, step-by-step variable collection.

Value proposition: Multi-turn ordering reduces catalog abandonment rates, improves requester experience by guiding users through complex forms, and enables ordering from messaging platforms without portal access.

Prerequisites

• Plugins: com.glide.cs.chatbot (Virtual Agent), com.glideapp.servicecatalog (Service Catalog)
• Roles: admin, virtual_agent_admin, or catalog_admin
• Access: Read/write access to sc_cat_item, sc_category, item_option_new, sys_cb_topic, and sys_cb_topic_detail
• Knowledge: Understanding of Virtual Agent topic design, NLU models, and Service Catalog API

Procedure

Step 1: Identify Catalog Items for Conversational Ordering

Determine which items are suitable for multi-turn ordering. Items with 3-8 variables and clear selection paths work best.

Using MCP (Claude Code/Desktop):

Tool: SN-Query-Table
Parameters:
  table_name: sc_cat_item
  query: active=true^type=item^sc_catalogs.titleLIKEService Catalog
  fields: sys_id,name,short_description,category,sc_catalogs,price,order,availability
  limit: 50
  order_by: category,name

Using REST API:

GET /api/now/table/sc_cat_item?sysparm_query=active=true^type=item&sysparm_fields=sys_id,name,short_description,category,sc_catalogs,price,order&sysparm_limit=50&sysparm_display_value=true

Check variable count per item to assess complexity:

Tool: SN-Execute-Background-Script
Parameters:
  description: Count variables per catalog item for ordering suitability
  script: |
    var items = new GlideRecord('sc_cat_item');
    items.addQuery('active', true);
    items.addQuery('type', 'item');
    items.orderBy('category');
    items.query();

    var results = [];
    while (items.next()) {
      var vars = new GlideAggregate('item_option_new');
      vars.addQuery('cat_item', items.sys_id);
      vars.addQuery('active', true);
      vars.addAggregate('COUNT');
      vars.query();

      var varCount = 0;
      if (vars.next()) varCount = parseInt(vars.getAggregate('COUNT'));

      var mandatoryVars = new GlideAggregate('item_option_new');
      mandatoryVars.addQuery('cat_item', items.sys_id);
      mandatoryVars.addQuery('active', true);
      mandatoryVars.addQuery('mandatory', true);
      mandatoryVars.addAggregate('COUNT');
      mandatoryVars.query();

      var mandCount = 0;
      if (mandatoryVars.next()) mandCount = parseInt(mandatoryVars.getAggregate('COUNT'));

      if (varCount > 0) {
        results.push({
          name: items.name.toString(),
          category: items.category.getDisplayValue(),
          total_variables: varCount,
          mandatory_variables: mandCount,
          suitability: varCount <= 8 ? 'Good' : 'Complex'
        });
      }
    }

    gs.info(JSON.stringify(results, null, 2));

Step 2: Design the Conversation Flow

Map the ordering process into conversation turns:

| Turn | Bot Action | User Response | Data Captured | |------|-----------|---------------|---------------| | 1 | "What would you like to order?" | "I need a new laptop" | Intent: catalog_order, keyword: laptop | | 2 | "I found these options: [list]" | "The performance model" | Item selection | | 3 | "What RAM size do you need?" | "16GB" | Variable: ram_size | | 4 | "What storage capacity?" | "512GB SSD" | Variable: storage_size | | 5 | "Business justification?" | "Current laptop failing" | Variable: justification | | 6 | "Confirm order: [summary]" | "Yes, submit" | Cart submission |

Step 3: Create the Virtual Agent Topic

Configure the Virtual Agent topic for catalog ordering.

Using MCP:

Tool: SN-Create-Record
Parameters:
  table_name: sys_cb_topic
  fields:
    name: "Order Catalog Item"
    description: "Conversational flow for browsing and ordering service catalog items with guided variable collection"
    category: service_catalog
    active: true
    enabled: true
    nlu_intent: catalog_order
    greeting_message: "I can help you order from the service catalog. What are you looking for?"
    fallback_message: "I couldn't find a matching item. Could you describe what you need differently?"
    end_message: "Your order has been submitted! You'll receive a confirmation email shortly."
    topic_type: standard

Using REST API:

POST /api/now/table/sys_cb_topic
Content-Type: application/json

{
  "name": "Order Catalog Item",
  "description": "Conversational flow for browsing and ordering service catalog items",
  "category": "service_catalog",
  "active": "true",
  "enabled": "true",
  "nlu_intent": "catalog_order",
  "greeting_message": "I can help you order from the service catalog. What are you looking for?",
  "topic_type": "standard"
}

Step 4: Configure Topic Details for Item Discovery

Add topic detail nodes that handle item search and selection.

Create search node:

Tool: SN-Create-Record
Parameters:
  table_name: sys_cb_topic_detail
  fields:
    topic: [topic_sys_id]
    name: "Search Catalog Items"
    node_type: script
    order: 100
    script: |
      (function() {
        var keyword = vaSystem.getLastUserMessage();
        var items = [];

        var gr = new GlideRecord('sc_cat_item');
        gr.addQuery('active', true);
        gr.addQuery('type', 'item');
        gr.addQuery('nameLIKE' + keyword)
          .addOrCondition('short_descriptionLIKE' + keyword);
        gr.setLimit(5);
        gr.query();

        while (gr.next()) {
          items.push({
            sys_id: gr.sys_id.toString(),
            name: gr.name.toString(),
            description: gr.short_description.toString(),
            price: gr.price.toString()
          });
        }

        vaVars.items = JSON.stringify(items);
        return items.length > 0 ? 'found' : 'not_found';
      })();

Create disambiguation node for multiple matches:

Tool: SN-Create-Record
Parameters:
  table_name: sys_cb_topic_detail
  fields:
    topic: [topic_sys_id]
    name: "Disambiguate Items"
    node_type: user_input
    order: 200
    prompt_message: "I found multiple items matching your request. Which one did you mean?"
    input_type: picker
    picker_source_variable: items

Step 5: Configure Variable Collection Nodes

Create conversational nodes that collect each required variable.

Using MCP:

Tool: SN-Execute-Background-Script
Parameters:
  description: Create variable collection nodes for catalog item ordering
  script: |
    var topicId = '[topic_sys_id]';
    var itemId = '[cat_item_sys_id]';

    // Get mandatory variables for the item
    var vars = new GlideRecord('item_option_new');
    vars.addQuery('cat_item', itemId);
    vars.addQuery('active', true);
    vars.addQuery('mandatory', true);
    vars.orderBy('order');
    vars.query();

    var nodeOrder = 300;
    while (vars.next()) {
      var detail = new GlideRecord('sys_cb_topic_detail');
      detail.initialize();
      detail.topic = topicId;
      detail.name = 'Collect: ' + vars.question_text.toString();
      detail.node_type = 'user_input';
      detail.order = nodeOrder;
      detail.prompt_message = vars.question_text.toString();

      // Map variable type to input type
      var varType = parseInt(vars.type);
      if (varType === 3) {
        detail.input_type = 'picker';
      } else if (varType === 9) {
        detail.input_type = 'date';
      } else if (varType === 2) {
        detail.input_type = 'text_area';
      } else {
        detail.input_type = 'text';
      }

      detail.mapped_variable = vars.name.toString();
      detail.insert();
      nodeOrder += 100;
    }

    gs.info('Created ' + ((nodeOrder - 300) / 100) + ' variable collection nodes');

Step 6: Add Order Confirmation and Submission

Create the confirmation and cart submission nodes.

Create confirmation node:

Tool: SN-Create-Record
Parameters:
  table_name: sys_cb_topic_detail
  fields:
    topic: [topic_sys_id]
    name: "Confirm Order"
    node_type: user_input
    order: 900
    prompt_message: "Here's your order summary:\n\nItem: {{selected_item_name}}\n{{variable_summary}}\n\nWould you like to submit this order?"
    input_type: yes_no

Create submission node using the Service Catalog API:

Tool: SN-Create-Record
Parameters:
  table_name: sys_cb_topic_detail
  fields:
    topic: [topic_sys_id]
    name: "Submit Order"
    node_type: script
    order: 1000
    script: |
      (function() {
        var itemId = vaVars.selected_item_id;
        var variables = JSON.parse(vaVars.collected_variables || '{}');

        // Create cart item
        var cart = new sn_sc.CatalogOrderHelper();
        cart.setRequestedFor(vaSystem.getUserSysId());

        var cartItem = cart.addToCart(itemId);
        for (var key in variables) {
          cartItem.setVariable(key, variables[key]);
        }

        var request = cart.submitOrder();
        vaVars.request_number = request.number;

        return 'submitted';
      })();

Step 7: Test and Publish the Conversation Flow

Validate the multi-turn flow end to end.

Using MCP:

Tool: SN-Query-Table
Parameters:
  table_name: sys_cb_topic_detail
  query: topic=[topic_sys_id]
  fields: sys_id,name,node_type,order,prompt_message,input_type,mapped_variable
  limit: 20
  order_by: order

Using REST API:

GET /api/now/table/sys_cb_topic_detail?sysparm_query=topic=[topic_sys_id]^ORDERBYorder&sysparm_fields=sys_id,name,node_type,order,prompt_message,input_type,mapped_variable&sysparm_limit=20&sysparm_display_value=true

Activate the topic:

Tool: SN-Update-Record
Parameters:
  table_name: sys_cb_topic
  sys_id: [topic_sys_id]
  fields:
    active: true
    enabled: true
    published: true

Tool Usage

MCP Tools Reference

| Tool | When to Use | |------|-------------| | SN-Query-Table | Query catalog items, variables, categories, topics | | SN-Create-Record | Create topics, topic details, and catalog configurations | | SN-Update-Record | Activate and publish topics, update item settings | | SN-NL-Search | Find catalog items matching natural language queries | | SN-Execute-Background-Script | Batch-create conversation nodes and test flows | | SN-Discover-Table-Schema | Explore Virtual Agent and catalog table schemas |

REST API Reference

| Endpoint | Method | Purpose | |----------|--------|---------| | /api/now/table/sc_cat_item | GET | Query catalog items for ordering | | /api/now/table/item_option_new | GET | Retrieve item variables for collection | | /api/now/table/sc_category | GET | Browse catalog categories | | /api/now/table/sys_cb_topic | GET/POST/PATCH | Manage Virtual Agent topics | | /api/now/table/sys_cb_topic_detail | GET/POST | Configure conversation flow nodes | | /api/sn_sc/servicecatalog/items | GET | Service Catalog API for item details | | /api/sn_sc/servicecatalog/cart | POST | Cart management and order submission |

Best Practices

• Keep conversations short: Limit to 5-7 turns maximum; combine related variables into single prompts when possible
• Offer quick picks: Present the 3-5 most popular items first before searching the full catalog
• Provide context at each turn: Remind users what they are ordering and what information is still needed
• Handle backtracking: Allow users to say "go back" or "change my selection" to revise previous answers
• Use rich messages: Display item images, prices, and descriptions using card-style responses
• Collect mandatory variables first: Ask for required fields before optional ones to minimize abandonment
• Confirm before submission: Always show a complete order summary and ask for explicit confirmation
• Graceful error handling: When item search finds no matches, suggest browsing by category or rephrasing the query

Troubleshooting

Virtual Agent Not Recognizing Catalog Intent

Cause: NLU model is not trained with sufficient utterances for the catalog ordering intent Solution: Add training utterances to the catalog_order intent such as "I need to order," "request a new," "can I get a," "I want to buy." Retrain the NLU model after adding utterances.

Variables Not Being Saved Between Turns

Cause: Conversation variables (vaVars) are not correctly mapped to catalog variables Solution: Verify mapped_variable on each topic detail node matches the name field in item_option_new. Check that the variable names in the submission script match exactly.

Cart Submission Fails

Cause: Missing mandatory variables, invalid variable values, or user session issues Solution: Validate all mandatory variables are collected before the submission node. Add a pre-submission script node that checks for missing values and loops back if needed. Verify CatalogOrderHelper is available in the instance.

Item Search Returns Wrong Results

Cause: Search query too broad or catalog items lack descriptive names and keywords Solution: Improve item short_description and description fields with searchable keywords. Consider adding a catalog keyword field or using NLU entity extraction to refine the search.

Examples

Example 1: Simple Hardware Order Conversation

Bot: "What would you like to order?"
User: "I need a new monitor"
Bot: "I found 2 monitors available:
      1. Standard Monitor (24" Dell) - $350
      2. Ultra-Wide Monitor (34" LG) - $750
      Which one would you like?"
User: "The ultra-wide"
Bot: "What is your desk location?"
User: "Building A, Floor 3, Desk 312"
Bot: "Do you need any mounting accessories? (desk mount, wall mount, or none)"
User: "Desk mount please"
Bot: "Order Summary:
      Ultra-Wide Monitor (34" LG) - $750
      Location: Building A, Floor 3, Desk 312
      Mounting: Desk mount
      Shall I submit this order?"
User: "Yes"
Bot: "Order submitted! Request REQ0045123 created. Expected delivery: 5-7 business days."

Example 2: Software Request with Disambiguation

Bot: "What would you like to order?"
User: "I need Adobe"
Bot: "I found several Adobe products:
      1. Adobe Creative Cloud (Full Suite) - $55/mo
      2. Adobe Acrobat Pro - $15/mo
      3. Adobe Photoshop (Single App) - $22/mo
      Which one do you need?"
User: "Just Acrobat"
Bot: "How many licenses do you need?"
User: "3"
Bot: "Business justification for the licenses?"
User: "Legal team needs PDF editing for contract reviews"
Bot: "I'll submit this for approval since it's 3+ licenses. Confirm?
      Adobe Acrobat Pro x3 - $45/mo
      Justification: Legal team PDF editing for contract reviews"
User: "Confirm"
Bot: "Submitted! REQ0045124 is pending manager approval."

Related Skills

• catalog/catalog-item-generation - Generate catalog items from descriptions
• catalog/variable-management - Advanced variable configuration
• catalog/item-creation - Standard catalog item setup
• catalog/request-fulfillment - Post-order fulfillment workflows
• catalog/approval-workflows - Approval routing for orders
• genai/playbook-generation - Generate Virtual Agent playbooks