happy-technologies-llc/now-assist-qa

Now Assist Q&A Configuration

Overview

This skill covers configuring ServiceNow Now Assist Q&A for conversational AI assistance:

• Setting up Now Assist skills for Q&A across different contexts (agent, employee, developer)
• Configuring knowledge sources that feed conversational responses
• Tuning response quality through prompt engineering and LLM parameters
• Managing conversation context and multi-turn dialogue
• Deploying Now Assist panels in Workspace, Portal, and Mobile
• Monitoring response quality through feedback analysis
• Integrating with AI Search for retrieval-augmented answers

When to use: When enabling AI-powered Q&A for agents, employees, or customers through Now Assist, configuring knowledge-grounded responses, or tuning conversational AI quality.

Prerequisites

• Roles: now_assist_admin, admin
• Plugins: sn_now_assist (Now Assist), com.snc.generative_ai_controller (Generative AI Controller), sn_gen_ai (Generative AI)
• Access: sn_now_assist_skill, sn_now_assist_skill_config, sn_gen_ai_config tables
• Knowledge: Understanding of LLM concepts (prompts, temperature, tokens), knowledge management
• Related Skills: genai/ai-search-rag for search configuration, genai/skill-kit-custom for custom skills

Procedure

Step 1: Assess Current Now Assist Configuration

Query existing Now Assist skills and their status.

MCP Approach:

Use SN-Query-Table on sn_now_assist_skill:
  - query: active=true
  - fields: sys_id,name,description,skill_type,context,active,status
  - limit: 20

Check existing skill configurations:

Use SN-Query-Table on sn_now_assist_skill_config:
  - query: active=true
  - fields: sys_id,skill,name,config_type,value
  - limit: 50

REST Approach:

GET /api/now/table/sn_now_assist_skill
  ?sysparm_query=active=true
  &sysparm_fields=sys_id,name,description,skill_type,context,active
  &sysparm_limit=20

Step 2: Configure Now Assist Q&A Skill

Agent-Facing Q&A Skill:

MCP Approach:

Use SN-Create-Record on sn_now_assist_skill:
  - name: "IT Agent Q&A"
  - description: "Provides AI-powered answers to IT agent questions from knowledge bases and resolved incidents"
  - skill_type: "qa"
  - context: "agent_workspace"
  - active: true
  - status: "draft"
  - knowledge_sources: "<kb_sys_id_1>,<kb_sys_id_2>"
  - enable_rag: true
  - response_mode: "generative"

REST Approach:

POST /api/now/table/sn_now_assist_skill
Body: {
  "name": "IT Agent Q&A",
  "description": "Provides AI-powered answers to IT agent questions from knowledge bases",
  "skill_type": "qa",
  "context": "agent_workspace",
  "active": true,
  "status": "draft",
  "enable_rag": true,
  "response_mode": "generative"
}

Employee-Facing Q&A Skill:

Use SN-Create-Record on sn_now_assist_skill:
  - name: "Employee Self-Service Q&A"
  - description: "AI answers for common employee IT and HR questions"
  - skill_type: "qa"
  - context: "employee_portal"
  - active: true
  - status: "draft"
  - knowledge_sources: "<it_kb_sys_id>,<hr_kb_sys_id>"
  - enable_rag: true
  - response_mode: "generative"
  - fallback_action: "create_incident"

Context options for Now Assist skills:

| Context | Value | Description | |---------|-------|-------------| | Agent Workspace | agent_workspace | IT agents in workspace UI | | Employee Portal | employee_portal | Self-service employee portal | | Customer Portal | customer_portal | External customer-facing | | Mobile | mobile | ServiceNow mobile app | | Virtual Agent | virtual_agent | VA conversation channel | | Platform | platform | Platform-wide availability |

Step 3: Configure Knowledge Sources

Link knowledge bases to the Q&A skill for grounded responses.

MCP Approach:

Use SN-Query-Table on kb_knowledge_base:
  - query: active=true
  - fields: sys_id,title,description,kb_version
  - limit: 20

Associate knowledge bases with the skill:

Use SN-Create-Record on sn_now_assist_skill_config:
  - skill: "<qa_skill_sys_id>"
  - name: "knowledge_source_1"
  - config_type: "knowledge_base"
  - value: "<kb_sys_id>"
  - active: true
  - order: 100

Use SN-Create-Record on sn_now_assist_skill_config:
  - skill: "<qa_skill_sys_id>"
  - name: "knowledge_source_2"
  - config_type: "knowledge_base"
  - value: "<hr_kb_sys_id>"
  - active: true
  - order: 200

Configure knowledge filtering:

Use SN-Create-Record on sn_now_assist_skill_config:
  - skill: "<qa_skill_sys_id>"
  - name: "kb_filter"
  - config_type: "knowledge_filter"
  - value: "workflow_state=published^valid_to>=javascript:gs.now()"
  - active: true

Step 4: Configure LLM Parameters

Set up the generative AI parameters for response generation.

MCP Approach:

Use SN-Create-Record on sn_gen_ai_config:
  - name: "Now Assist Q&A LLM Config"
  - description: "LLM parameters for Now Assist Q&A responses"
  - active: true
  - llm_provider: "now_llm"
  - model: "default"
  - temperature: 0.3
  - max_tokens: 400
  - top_p: 0.9
  - frequency_penalty: 0.1
  - presence_penalty: 0.1

REST Approach:

POST /api/now/table/sn_gen_ai_config
Body: {
  "name": "Now Assist Q&A LLM Config",
  "active": true,
  "llm_provider": "now_llm",
  "temperature": 0.3,
  "max_tokens": 400,
  "top_p": 0.9
}

Link LLM config to skill:

Use SN-Create-Record on sn_now_assist_skill_config:
  - skill: "<qa_skill_sys_id>"
  - name: "llm_config"
  - config_type: "gen_ai_config"
  - value: "<gen_ai_config_sys_id>"
  - active: true

Step 5: Configure System Prompts

System prompts guide LLM behavior for Q&A responses.

MCP Approach:

Use SN-Create-Record on sn_now_assist_skill_config:
  - skill: "<qa_skill_sys_id>"
  - name: "system_prompt"
  - config_type: "prompt"
  - value: |
      You are an IT support assistant for {company_name}. Your role is to help agents
      find answers to technical questions using the provided knowledge base content.

      Guidelines:
      - Only answer based on the provided context. Do not make up information.
      - If the context does not contain a clear answer, say "I don't have enough
        information to answer this question" and suggest creating a knowledge article.
      - Always cite the source knowledge article by title.
      - Keep answers concise and actionable, using bullet points for steps.
      - If the question involves a specific incident, reference relevant fields.
      - Never reveal internal system details, credentials, or sensitive data.
  - active: true

Prompt template variables:

| Variable | Description | Example | |----------|-------------|---------| | {company_name} | Customer organization name | Acme Corp | | {user_name} | Current user's name | John Smith | | {user_role} | Current user's primary role | ITIL Agent | | {context_record} | Current record being viewed | INC0012345 | | {knowledge_context} | Retrieved knowledge content | (injected by RAG) | | {conversation_history} | Previous turns in dialogue | (auto-populated) |

Step 6: Configure Now Assist Panel

Set up where and how the Now Assist panel appears.

MCP Approach:

Use SN-Create-Record on sn_now_assist_panel_config:
  - name: "Agent Workspace Q&A Panel"
  - description: "Now Assist Q&A panel for IT agent workspace"
  - active: true
  - context: "agent_workspace"
  - skills: "<qa_skill_sys_id>"
  - position: "right_sidebar"
  - auto_open: false
  - show_feedback: true
  - show_sources: true
  - max_conversation_turns: 10
  - greeting_message: "Hi! I can help you find answers from our knowledge base. What would you like to know?"

REST Approach:

POST /api/now/table/sn_now_assist_panel_config
Body: {
  "name": "Agent Workspace Q&A Panel",
  "active": true,
  "context": "agent_workspace",
  "skills": "<qa_skill_sys_id>",
  "position": "right_sidebar",
  "show_feedback": true,
  "show_sources": true
}

Step 7: Configure Conversation Context

Set up how conversation context is managed for multi-turn dialogues.

MCP Approach:

Use SN-Create-Record on sn_now_assist_skill_config:
  - skill: "<qa_skill_sys_id>"
  - name: "context_config"
  - config_type: "context"
  - value: '{
      "max_history_turns": 5,
      "include_record_context": true,
      "record_fields": ["short_description", "description", "priority", "category", "assignment_group"],
      "context_window_tokens": 3000,
      "clear_on_record_change": true
    }'
  - active: true

Step 8: Set Up Response Quality Guardrails

MCP Approach:

Use SN-Create-Record on sn_now_assist_skill_config:
  - skill: "<qa_skill_sys_id>"
  - name: "guardrails"
  - config_type: "safety"
  - value: '{
      "require_knowledge_grounding": true,
      "min_confidence_score": 0.65,
      "block_pii_in_response": true,
      "max_response_length": 500,
      "prohibited_topics": ["salary", "termination", "legal_advice"],
      "fallback_message": "I am not able to answer this question. Please contact your IT service desk for assistance.",
      "enable_content_filtering": true
    }'
  - active: true

Step 9: Activate and Deploy

MCP Approach:

Use SN-Update-Record on sn_now_assist_skill:
  - sys_id: "<qa_skill_sys_id>"
  - status: "published"
  - active: true

Verify configuration is complete:

Use SN-Query-Table on sn_now_assist_skill_config:
  - query: skill=<qa_skill_sys_id>^active=true
  - fields: name,config_type,active
  - limit: 20

Verify panel is active:

Use SN-Query-Table on sn_now_assist_panel_config:
  - query: skills=<qa_skill_sys_id>^active=true
  - fields: name,context,position,active

Step 10: Monitor and Tune Response Quality

Query feedback data:

Use SN-Query-Table on sn_now_assist_feedback:
  - query: skill=<qa_skill_sys_id>^sys_created_on>javascript:gs.daysAgo(7)
  - fields: sys_id,query,response,rating,feedback_text,user
  - limit: 50
  - orderBy: sys_created_on
  - orderDirection: desc

Analyze negative feedback patterns:

Use SN-Query-Table on sn_now_assist_feedback:
  - query: skill=<qa_skill_sys_id>^rating<3
  - fields: query,response,rating,feedback_text
  - limit: 20

Adjust based on feedback:

Use SN-Update-Record on sn_gen_ai_config:
  - sys_id: "<gen_ai_config_sys_id>"
  - temperature: 0.2
  - max_tokens: 500

Tool Usage

| Tool | Purpose | When to Use | |------|---------|-------------| | SN-Query-Table | Find existing skills, configs, feedback | Discovery and monitoring | | SN-Create-Record | Create skills, configs, panel settings | Initial setup | | SN-Update-Record | Tune LLM params, activate skills | Optimization and deployment | | SN-Get-Table-Schema | Discover config table fields | Understanding available settings |

Best Practices

1. Start with a single knowledge base and expand sources incrementally as quality is validated
2. Use low temperature (0.2-0.4) for Q&A to keep responses factual and grounded
3. Write specific system prompts that define the assistant's role, boundaries, and response format
4. Enable citation display so users can verify answers against source material
5. Set confidence thresholds conservatively (0.65+) and adjust based on feedback
6. Include record context when the Q&A skill is used alongside incidents or cases
7. Limit conversation history to 5-7 turns to manage token usage and context relevance
8. Block PII in responses to prevent accidental exposure of sensitive data
9. Monitor feedback weekly and address patterns of negative ratings promptly
10. Test with real user queries from support tickets to validate response quality before launch

Troubleshooting

| Issue | Cause | Resolution | |-------|-------|------------| | Now Assist panel not visible | Panel config inactive or wrong context | Verify sn_now_assist_panel_config.active=true and context matches UI | | Responses are generic/unhelpful | Knowledge sources not linked or empty | Check skill_config knowledge_base entries and verify KB has published articles | | Hallucinated answers | Temperature too high or no grounding requirement | Lower temperature, set require_knowledge_grounding=true | | Response cut off mid-sentence | max_tokens too low | Increase max_tokens in sn_gen_ai_config (400-600 recommended) | | Slow response time | Too many knowledge sources or large context window | Reduce search sources, lower context_window_tokens | | Skill not triggering | Skill status is draft or inactive | Publish and activate the skill | | Feedback not collected | show_feedback disabled on panel | Set show_feedback=true in panel config | | Wrong knowledge base searched | KB filter conditions too broad or missing | Add specific kb_knowledge_base filter in knowledge_filter config |

Examples

Example 1: IT Agent Q&A in Workspace

Configure Now Assist Q&A for ITSM agents:

• Skill: Agent workspace context, linked to IT Knowledge Base and resolved incidents
• System Prompt: IT support assistant, answers from KB only, cites sources, suggests escalation paths
• LLM Config: Temperature 0.3, max 400 tokens, now_llm provider
• Panel: Right sidebar, auto-open disabled, feedback enabled, shows 3 source citations
• Context: Include incident short_description, description, priority, category from viewed record
• Guardrails: Require knowledge grounding, block PII, min confidence 0.7

Example 2: Employee Self-Service Q&A

Configure Now Assist for employee portal self-service:

• Skill: Employee portal context, linked to IT FAQ KB and HR Policy KB
• System Prompt: Friendly employee assistant, provides step-by-step guidance, offers catalog links
• LLM Config: Temperature 0.4, max 500 tokens for more detailed self-service instructions
• Panel: Embedded in portal search, greeting message, max 5 conversation turns
• Fallback: If confidence below 0.6, suggest creating an incident or browsing catalog
• Guardrails: Block salary/termination topics, require KB grounding, content filtering enabled

Example 3: CSM Agent Q&A for Customer Support

Configure Now Assist for customer service management agents:

• Skill: Agent workspace context for CSM, linked to product KB and customer solutions KB
• System Prompt: Customer support specialist, professional tone, includes troubleshooting steps
• LLM Config: Temperature 0.2 for precise technical answers, max 400 tokens
• Context: Include case short_description, product, account, priority from current case
• Panel: Right sidebar with auto-open on case creation, feedback and sources visible
• Guardrails: Prohibit sharing internal pricing, SLA details, or customer data from other accounts

Related Skills

• genai/ai-search-rag - AI Search configuration that powers Q&A knowledge retrieval
• genai/skill-kit-custom - Creating custom Now Assist skills beyond Q&A
• genai/flow-generation - Flows that can be triggered from Q&A fallback actions
• knowledge/article-management - Managing knowledge content quality for better Q&A