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-Natural-Language-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

About this skill

Overview

Prerequisites

Procedure

Step 1: Identify Catalog Items for Conversational Ordering

Step 2: Design the Conversation Flow

Step 3: Create the Virtual Agent Topic

Step 4: Configure Topic Details for Item Discovery

Step 5: Configure Variable Collection Nodes

Step 6: Add Order Confirmation and Submission

Step 7: Test and Publish the Conversation Flow

Tool Usage

MCP Tools Reference

REST API Reference

Best Practices

Troubleshooting

Virtual Agent Not Recognizing Catalog Intent

Variables Not Being Saved Between Turns

Cart Submission Fails

Item Search Returns Wrong Results

Examples

Example 1: Simple Hardware Order Conversation

Example 2: Software Request with Disambiguation

Related Skills