AI Agent Advanced Scenarios & Runtime Issues Playbook - Support and Troubleshooting

AI Agent Advanced Scenarios & Runtime Issues Playbook Summary

:root { --ink: #1a1f36; --ink-light: #4a5068; --ink-muted: #8892a7; --surface: #f8f9fc; --card: #ffffff; --border: #e2e5ee; --accent: #2563eb; --accent-soft: #eff4ff; --accent-dark: #1d4ed8; --green: #059669; --green-soft: #ecfdf5; --amber: #d97706; --amber-soft: #fffbeb; --red: #dc2626; --red-soft: #fef2f2; --purple: #7c3aed; --purple-soft: #f5f3ff; --teal: #0891b2; --teal-soft: #ecfeff; --radius: 10px; --shadow-sm: 0 1px 3px rgba(26,31,54,0.06); --shadow-md: 0 4px 16px rgba(26,31,54,0.08); } *, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; } body { font-family: 'DM Sans', -apple-system, sans-serif; background: var(--surface); color: var(--ink); line-height: 1.65; -webkit-font-smoothing: antialiased; } .hero { background: linear-gradient(135deg, #1a1f36 0%, #2a3158 50%, #1e3a5f 100%); color: white; padding: 56px 32px 48px; position: relative; overflow: hidden; } .hero-overlay { position: absolute; inset: 0; background: radial-gradient(ellipse at 70% 20%, rgba(37,99,235,0.2) 0%, transparent 60%), radial-gradient(ellipse at 20% 80%, rgba(8,145,178,0.15) 0%, transparent 50%); pointer-events: none; } .hero-inner { max-width: 960px; margin: 0 auto; position: relative; z-index: 1; } .hero-badge { display: inline-flex; align-items: center; gap: 6px; background: rgba(255,255,255,0.1); border: 1px solid rgba(255,255,255,0.15); padding: 5px 14px; border-radius: 20px; font-size: 12px; font-weight: 600; letter-spacing: 0.5px; text-transform: uppercase; margin-bottom: 20px; } .hero-badge .dot { width: 6px; height: 6px; border-radius: 50%; background: #34d399; } .hero h1 { font-size: 32px; font-weight: 700; line-height: 1.25; margin-bottom: 12px; } .hero p { font-size: 16px; color: rgba(255,255,255,0.75); max-width: 640px; line-height: 1.6; } .hero-meta { display: flex; flex-wrap: wrap; gap: 24px; margin-top: 24px; font-size: 13px; color: rgba(255,255,255,0.55); } .container { max-width: 960px; margin: 0 auto; padding: 0 24px; } .content { padding: 40px 0 80px; } .use-case:hover { box-shadow: 0 4px 16px rgba(26,31,54,0.08); } .toc a:hover { color: #2563eb; } .prop-table tr:nth-child(even) td { background: #fafbfd; } @media (max-width: 640px) { .hero { padding: 36px 20px 32px; } .hero h1 { font-size: 24px; } .toc-grid { grid-template-columns: 1fr; } .uc-header, .uc-body { padding-left: 20px; padding-right: 20px; } .prop-table { font-size: 12px; } .prop-table th, .prop-table td { padding: 8px 10px; } } A supplementary troubleshooting playbook covering advanced runtime scenarios including plugin installation failures, AI Agent Studio security constraints, Knowledge Graph tool errors, agent naming conflicts, provider configuration, trigger misfires, and domain-specific agent issues across ITSM, HR, CMDB, and ITOM.

📂 Category: AI Agent — Advanced Scenarios & Runtime ⚡ Applies to: AI Agent Studio · ITSM · HR · CMDB · ITOM · Knowledge Graph · A2A · Now Assist Admin Use Cases in This Playbook 1 Plugin Dependency Chain Failure 2 SKU Lockout — High-Frequency Issue 3 One-Character Typo Crashes ITSM 4 State Desync — Terminated vs Ongoing 5 Emergency Stop — Runaway Agents 6 Vanishing OOB Workflows 7 Post-Upgrade Agent Amnesia 8 Name Change Orphan 9 Provider Ghost — Azure AI Missing 10 Knowledge Graph Dead End 11 CMDB Timeout — Large Table Queries 12 Phantom Triggers — Zero Executions 13 Probabilistic Drift in Predictions 14 False "Excellent" Ratings 15 Sticky Skill Won't Remove 16 17-Second BGE Reranker Lag 17 A2A Config Matrix — Async & Grants 18 Unwanted Translation in VA 19 Post-Clone — Unique Key Violations 20 Stale Topics — All Workflows Stop 21 Worknotes Display "Null" User 🧭 Decision Tree — Which Fix Do I Need? 📚 Consolidated Reference Links 📦 Use Case 1: Plugin Dependency Chain Failure — "This App Can Only Be Installed as a Dependency" Plugin Config Problem When attempting to install Now Assist plugins, the installation fails with errors such as "This app can only be installed as a dependency" (for Generative AI Controller) or dependency conflicts prevent installation of Now Assist Skill Kit, OpenAI Generative AI Spoke, or IntegrationHub Action Step - Server Sent Events.

Root Cause The Generative AI Controller ( sn_generative_ai ) is included as a dependency for all Now Assist applications. Per product documentation, it is not recommended to install it individually . It must be installed as part of a Now Assist application suite (e.g., Now Assist for ITSM). Additionally, some plugins require entitlement verification before installation.

Resolution 1 Do not attempt to install Generative AI Controller individually. Instead, install the Now Assist application suites: Now Assist for Platform , Now Assist for ITSM , Now Assist for CSM , Now Assist for FSO — the controller will be included as a dependency. 2 Before installing, verify the customer has entitlement for the Now Assist suites by checking with the Admin support team or reviewing the contract. 3 If Now Assist applications are not visible in the App Manager, run the update checker script to refresh the instance's connection with the ServiceNow Store. 4 Use the pre-validated suite versions list to ensure compatibility: see KB2666962. ℹ️ Installation Order: Install Now Assist suite apps (which pull in dependencies automatically). Do NOT install Generative AI Controller, Platform AI Agents, or Now Assist Skill Kit as standalone first — install them via the suite applications.

📘 KB2666962 — Pre-Validated Suite Versions 🔐 Use Case 2: SKU Lockout — "Security Constraints" Blocks AI Agent Studio Access (High-Frequency Issue) Config Licensing Problem Users encounter a "Security Constraints" banner in AI Agent Studio, or see "Access is limited due to license restrictions," preventing them from configuring or testing agents. The issue is caused by the ACL now.agent-studio.*/ux_route failing at the script level.

Root Cause Changes to the NAA (Now Assist Admin) Fulfiller SKU merged into recent branches affect instances with recent upgrades. The full ACL chain: ACL now.agent-studio.*/ux_route (sys_id: 669da2f1370caa10130b6c6655c06728 ) → calls sn_aia.AiAgentRuntimeUtil (sys_id: 206d92309f9312109a61bfbc2b0a1ceb ) → calls sn_nowassist_admin.NowAssistSkillConfig.checkAccessForAssetSubscription → checks if "Now Assist for Fulfiller subscription" SKU is provisioned. The Now Assist Admin Console version beyond 6.6.4 is impacted — missing SKU entitlements trigger access restrictions via the script: answer=(new sn_aia.AiAgentRuntimeUtil.hasValidLicense);

Resolution 1 Set system property sn_nowassist_admin.ignoreFulfillerSubscriptionCheck to true as a workaround. 2 Verify entitlements using background script: gs.info(new sn_entitlement.GlideEntitlementAPI.hasLicenseForArtifact("family_id", "sn_gen_ai_platform_fulfiller", false, true)); — returns "true" if SKU present, "false" if missing. 3 Verify that the customer's contract includes the required Now Assist SKU entitlements mapped to both production and sub-production instances. Property sys_id: d39fc7c7ff057210b97effffffffff75 . 4 If entitlements are confirmed but the error persists, engage the Pricing Helpdesk team to verify backend mapping. ⚠️ Important: The ignoreFulfillerSubscriptionCheck property is a workaround, not a permanent solution. The root cause is missing SKU entitlements that should be resolved through proper licensing.

💥 Use Case 3: One-Character Typo Crashes All ITSM Agents — journalGR vs journalGr Plugin Execution Problem After installing a newer version of the IT Service Management AI Agent Collection plugin, AI agents that use the "ITSM incident resolution plan investigation AI agent - Fetch all data" tool crash with the generic "Sorry, there was a problem" error.

Root Cause A typo in the script include sn_itsm_aia.InvestigatingAndResolvingIncidentsUtilSNC introduced in the plugin update. On line 53, journalGR.orderBy("sys_created_on") uses an incorrect variable name (capital R instead of lowercase r).

Resolution 1 Revert the script include sn_itsm_aia.InvestigatingAndResolvingIncidentsUtilSNC to the latest OOB version. 2 Then fix the typo on line 53: change journalGR.orderBy("sys_created_on"); to journalGr.orderBy("sys_created_on"); (lowercase 'r'). 3 Save the record and re-test the agent. 🔴 PRB1998233 — Script include typo in ITSM AI Agent Collection ⏹️ Use Case 4: State Desync — Agent "Terminated" While Tool Still Shows "Ongoing" Execution Problem In the AI Agent playground, the agent flow shows a "Terminated" status, but the tool state remains as "Ongoing" despite the tool having completed execution successfully. This creates confusion about the actual execution state.

Resolution 1 Check the execution plan and individual execution tasks to verify whether the tool actually completed. 2 Review the Gen AI logs for the conversation to identify if the issue is a state display bug versus an actual execution failure. 3 If the tool completed successfully but the agent still shows terminated, this may be a UI state synchronization issue — check for known defects in the current plugin version and upgrade if a fix is available. 🛑 Use Case 5: Emergency Stop — Halting Runaway In-Progress Agent Executions Execution Config Problem AI Agents or Agentic Workflows were unintentionally triggered (e.g., by auto-triggers on record creation) and are now in an in-progress state, consuming assists and resources. The administrator needs to stop them.

Resolution 1 Navigate to the sn_aia_execution_plan table and filter for in-progress execution plans. 2 Cancel the execution plans that were unintentionally triggered by updating their state. 3 Review and disable the auto-triggers that caused the unintended execution in AI Agent Studio → Workflows → Triggers. 4 Check syslog for rate limiting: filter source = sn_aia.AiAgentRuntimeUtilSNC AND message CONTAINS "Blocking AIA conversation" . If found, platform rate limiting is actively blocking executions — investigate trigger volume immediately. 5 To prevent future occurrences, carefully review trigger conditions before enabling them in production. 🔍 Use Case 6: Vanishing OOB Workflows — Collection Plugins Not Updated After Upgrade Plugin Config Problem OOB Agentic Workflows (e.g., "Plan Generation for Issue Resolution") are present in the development instance but missing in QA and Production instances after upgrade.

Resolution 1 OOB workflows are delivered via AI Agent Collection plugins (e.g., ITSM AI Agent Collection, CSM AI Agent Collection). Verify these collection plugins are installed and on the latest version in the affected instances. 2 Check the App Manager for pending updates to the AI Agent Collection plugins and apply them. 3 Repair the relevant AI Agent Collection plugin if the workflows are still missing after upgrade. 4 Verify that the sn_aia_usecase records exist for the missing workflows. 🔄 Use Case 7: Post-Upgrade Agent Amnesia — NAVA Stops Selecting Known Agents Discovery Plugin Problem After upgrading to a newer patch version, AI Agents that were previously discovered and selectable in NAVA are no longer being selected or invoked. The agent works in AI Agent Studio but NAVA does not pick it up.

Resolution 1 Verify all Now Assist store applications were upgraded alongside the platform patch (see KB-0001, Use Case 3). 2 Check the assistant configuration — ensure the agent is still promoted and enabled for the assistant after the upgrade. 3 Reindex the Skills indexed source. 4 Verify the sys_cs_context_profile is active and the sys_now_assist_deployment is in sync. ✏️ Use Case 8: Name Change Orphan — Renamed Agent Disconnects from Skill Discovery Config Problem When renaming an OOB AI Agent through the AI Agent Studio, the name change does not propagate to the corresponding sys_gen_ai_skill_config record, causing inconsistency between the agent name shown in the Studio and the skill name used in discovery.

Resolution 1 After renaming the agent in AI Agent Studio, manually update the corresponding sys_gen_ai_skill_config record to match the new name. 2 Also check and update the sys_gen_ai_skill record if the display name is used in NAP discovery. 3 Reindex the Skills indexed source after making name changes. ☁️ Use Case 9: Provider Ghost — Azure AI One Extend Definition Missing from Instance Config LLM Problem Users are unable to change the Now Assist provider, or the Azure AI One Extend definition is not available on the instance, preventing configuration of third-party LLM providers.

Resolution 1 Verify the Generative AI Controller plugin ( sn_generative_ai ) is on the latest version — Azure AI support may require a specific minimum version. 2 Check the sys_one_extend_definition table for the Azure AI One Extend record. If missing, repair the Generative AI Controller plugin. 3 Verify that the One Extend capability definitions are properly loaded by checking sys_one_extend_capability_definition . 4 If the instance was recently upgraded or cloned, re-repair the Gen AI Controller and related plugins. 🔗 Use Case 10: Knowledge Graph Dead End — "Knowledge Graph Not Found" During Execution Tool Execution Problem The Knowledge Graph tool within AI Agents is not working and keeps returning a "knowledge graph not found" error. The agent cannot retrieve data through the Knowledge Graph integration.

Resolution 1 Verify the Knowledge Graph plugin ( sn_kg ) is installed and on a compatible version. 2 Check that the Knowledge Graph schema is properly configured and the graph is indexed. 3 Verify the AI Agent has the correct permissions to access the Knowledge Graph tables and APIs. 4 Check the sys_generative_ai_log for specific error messages related to the KG tool invocation. 5 Ensure the sn_ais_assist.enable_knowledge_graph_nlq property is set to true . 6 Verify Knowledge Graph Core plugin ( sn_kg ) is on a compatible version (key versions: 4.1.5, 5.1.0). 7 If using TextToResult/KG NLQ: increase timeout via sn_ais_assist.kg.nlq.request_timeout = 20000 (default: 10000ms) — workaround for PRB1950955. ⚙️ sn_ais_assist.enable_knowledge_graph_nlq 🐌 Use Case 11: CMDB Search Timeout — Large Table Queries Exceed Agent Limits Tool Performance Problem The "CMDB Search" AI Agent is getting timed out and producing slow search results, causing agent execution to fail or take excessively long.

Resolution 1 Check the CMDB data volume — the search tool may be querying tables with very large record counts without adequate filtering. 2 Review and optimize the search query within the agent's tool configuration. Add filters to narrow the result set. 3 Check the sn_cmdb_gen_ai.import_set_query_limit property (default: 100) and adjust if needed. 4 Review the sn_cmdb_gen_ai.error_summary_max_length and sn_cmdb_gen_ai.error_summary_min_length properties for tuning accuracy vs. performance. 5 If using CI Summarization, check sn_cmdb_gen_ai.ci_summarization_aggregation property to control how related list data is handled. 6 Also check sn_cmdb_gen_ai.error_summary_max_length and sn_cmdb_gen_ai.error_summary_min_length for tuning accuracy vs. performance trade-offs. 7 For very large tables: index the queried fields, switch from Record Operation to FDIH tools, and set timeout > 60 seconds in sn_aia_agent_tool_m2m . 🔔 Use Case 12: Phantom Triggers — Successful Fire but Zero Execution Plans Created Execution Config Problem The trigger for an Agentic Workflow (e.g., "Resolve HR cases") shows successful execution in the trigger log, but no corresponding AI Agent execution plan or execution is created. The workflow never actually runs.

Resolution 1 Verify the trigger condition is actually met for the specific record — check the trigger's filter conditions against the record data. 2 Check the sn_aia_execution_plan table for any failed or cancelled plans that may have been silently discarded. 3 Verify that the Agentic Workflow is active and properly configured in the AI Agent Studio. 4 Check that the user context under which the trigger runs has the required roles ( sn_aia.read and domain-specific roles). 5 Review system logs ( syslog_app_scope ) for errors from the sn_aia scope around the trigger execution time. 🎯 Use Case 13: Probabilistic Drift — Predictive Agent Returns Different Results Each Run Execution LLM Problem The Predictive Intelligence AI Agent (e.g., "Predict service and transfer HR Cases") produces inconsistent results — sometimes predicting correctly, other times failing or producing incorrect predictions for similar input data.

Resolution 1 This is inherent to LLM-based predictions — results are probabilistic, not deterministic. The same input may produce different outputs across runs. 2 Review the agent's instructions for ambiguity that could lead to inconsistent LLM reasoning. 3 Increase the training data volume for predictive models if the agent relies on Predictive Intelligence models. 4 Test with the auto-chat evaluation feature to measure consistency across multiple runs and identify patterns in failures. 5 Consider using script tools for deterministic logic instead of relying on the LLM for classification tasks. 📊 Use Case 14: False "Excellent" — Assessment Agent Rates Empty Records as High Quality Execution LLM Problem The "Assessment of Change Quality" AI Agent returns "Excellent" ratings even for change records with practically no data populated. The quality assessment does not accurately reflect the actual data completeness.

Resolution 1 Review the agent's instructions and the assessment criteria prompts. The LLM may be defaulting to positive ratings when insufficient data is available. 2 Modify the agent instructions to explicitly handle empty or sparse data: e.g., "If required fields are empty or contain minimal data, rate the quality as Poor or Insufficient." 3 Add validation logic via a script tool that checks field completeness before passing to the LLM for assessment. 4 Test with records at various completeness levels to calibrate the expected outputs. 🗑️ Use Case 15: Sticky Skill — Cannot Remove a Skill from Now Assist Panel NAP Config Problem A skill (e.g., "Suggest configuration items for a change request") cannot be removed from the Now Assist Panel. The skill continues to appear even after attempts to deactivate or remove it from the assistant configuration.

Resolution 1 Check the sys_cs_context_profile_skill table for the skill-to-assistant mapping. Delete the unwanted mapping record. 2 Verify the sys_gen_ai_skill and sn_aia_skill_metadata records — deactivate the skill record if it should not appear. 3 Deactivate the skill from the Now Assist Admin Console under the appropriate assistant configuration. 4 Reindex the Skills indexed source after removal. 5 Clear cache ( cache.do ) and re-test. 🐢 Use Case 16: 17-Second Discovery Lag — BGE Reranker Bottleneck in Agent Invocation Performance Discovery Problem BGE reranker response is slow during agent discovery on Now Assist VA, with async calls taking 17+ seconds. This delays the agent invocation significantly and degrades the user experience.

Resolution 1 This is a known performance issue being tracked internally. Check for available patches or hotfixes that address BGE reranker latency. 2 Reduce the number of skills and agents to minimize the reranker workload during discovery. 3 Check the glide.cache.size.SkillDetailsCache property and increase it to reduce cache misses. 4 Monitor the Gen AI logs for "AIA-Agent Invoker" timing to measure the specific latency contribution from the reranker step. 🔌 Use Case 17: A2A Async Invocation & Grant Type Configuration Issues A2A Config Problem A2A (Agent-to-Agent) async invocation fails, or users have questions about the correct grant type to configure for A2A communication between instances.

Resolution 1 For async A2A invocations, ensure push notifications are enabled on the instance. Without push notifications, async mode causes the subflow to wait indefinitely and timeout. 2 Use sync mode for A2A testing in the AIA playground — async is NOT supported there. 3 Verify sys_cs_channel → "AI Agent A2A Channel" → synchronous field must be true (sys_id: 5b6acb7eff426e50d635ffffffffffd8 ). 4 For grant type configuration, use OAuth 2.0 with the appropriate grant type for server-to-server communication (typically client_credentials grant type). 5 Verify the sn_mcp_client.wait_for_server_connection_close property is set to true if the server sends multiple datagrams. 🌐 Use Case 18: Unwanted Translation — VA Auto-Translates Responses to Non-English NAVA I18N Problem When a user submits a non-English utterance to Now Assist Virtual Agent, the VA automatically translates the response into the user's language — even when the customer does not want translation and prefers all responses in English regardless of the input language.

Resolution 1 Check the sn_generative_ai.append_lang_in_prompt property. When set to true , the system appends the user's language preference to the LLM prompt, causing it to respond in that language. 2 If all responses should be in English regardless of user language, set sn_generative_ai.append_lang_in_prompt to false . 3 Alternatively, modify the agent/skill prompt to explicitly instruct: "Always respond in English regardless of the input language." 4 Verify the glide.ais.global_searchable_filter.kb_knowledge property if non-English users cannot find English KB articles through VA. 💥 Use Case 19: Post-Clone Database Corruption — Unique Key Violations Kill All Agents Config Execution Problem After cloning or maintenance activity, ALL AI agents hang and fail to execute with "duplicate key value violates unique constraint" errors in the database. Execution plans are created but fail at the orchestrator step. GenAI logs show execution starts but never completes.

Resolution 1 Repair plugins in dependency order: com.glide.ai_record_activity → com.glide.identity.agent.security → Now Assist AI Agents ( sn_aia ) → Platform AI Agents and Skills. 2 If repair doesn't fix: escalate to platform engineering support for database constraint repair. 3 Post-clone checklist: Verify no unique constraint violations in sn_aia tables. Test agent execution with a simple agent before deploying complex workflows. 📄 Use Case 20: Stale System Topics — All Agentic Workflows Stop Simultaneously Plugin Execution Problem ALL agentic workflows fail simultaneously with "Script evaluation error at [topic_AIA-Agent Invoker]" or "TypeError: Method toString called on incompatible object." System topics were modified in prior support interactions — the Design topic version ≠ Runtime version.

Diagnostic Steps 1 Navigate: sys_cs_topic_list.do → filter: name LIKE "AIA-agent" OR name LIKE "AIA". 2 Check "System Update Version" date on each topic. 3 Compare with latest OOB version (Source = 'Store Application: Now Assist AI Agents'). Example: topics from 2025-08-23 vs OOB 2026-02-18 = stale. Resolution 1 Mark ALL AIA system topics replace_on_upgrade = true in sys_update_xml . When bulk updating, change application scope per topic. 2 Repair plugins: sn_aia AND sn_nowassist_va . 3 Verify AIA-Agent Invoker topic is on latest OOB version after repair. 🚫 Recurring Issue: This exact root cause appeared 3 times at a single customer account. This is a recurring pattern. NEVER directly modify AIA system topics. Always toggle state via the Assistant Setup page UI.

👤 Use Case 21: Agent Worknotes Display "Null" User Instead of AI Identity Config Execution Problem Work notes or comments posted by AI Agent Workflow display the user as "null" instead of "AI" or the configured worker user name. Activity stream shows "[Now Assist Generated]" but user is "Null". Shows as "AI" in DEV but "null" in UAT.

Root Cause The AI Agent's "Run As" user configuration is missing, inactive, or not properly synced between environments. The ITSM Worker AI Agent user record may not exist or be active in the target instance.

Resolution 1 Verify "ITSM Worker AI Agent" user record exists and is active in the affected instance. 2 If missing: import via update set or create manually. Also check for "Platform AI Worker" user. 3 Repair sn_aia plugin to restore default user records. 4 Check the agentic workflow "Run As" user configuration. 🧭 Decision Tree — Which Fix Do I Need? Is it a plugin installation or licensing issue? "This app can only be installed as a dependency" → Use Case 1 "Security Constraints" or "license restrictions" in AI Agent Studio → Use Case 2 Agent crashes after plugin update → Use Case 3 Is it an agent state or lifecycle issue? Agent terminated but tool shows ongoing → Use Case 4 Need to stop unintentionally triggered agents → Use Case 5 OOB workflows missing after upgrade → Use Case 6 Agent not selected in NAVA after upgrade → Use Case 7 Is it a configuration or naming issue? Renamed agent not reflected in skill config → Use Case 8 Provider / Azure AI not available → Use Case 9 Cannot remove skill from NAP → Use Case 15 Is it a tool or performance issue? Knowledge Graph tool errors → Use Case 10 CMDB search timeout / slow → Use Case 11 BGE reranker slow discovery (17+ sec) → Use Case 16 Is it an execution quality or trigger issue? Triggers fire but no execution created → Use Case 12 Predictive agent inconsistent results → Use Case 13 Agent assessment returns wrong scores → Use Case 14 VA translates response unexpectedly → Use Case 18 A2A async / grant type issues → Use Case 17 All agents died after clone (unique key) → Use Case 19 ALL workflows stopped (script error in AIA-Agent Invoker) → Use Case 20 Agent worknotes show "null" user → Use Case 21 📚 Consolidated Reference Links ServiceNow KB Articles 📘 KB2666962 — Pre-Validated Suite Versions <a class="ref ref-kb" style="display: inline-flex; align-items: center; gap: 5px; font-size: 12px; font-weight: 500; padding: 4px 12px; border-radius: 6px; background: #eff6ff; color: #1e40af; border: 1px solid #bfdbfe; text-decoration: none; cursor: pointer; transition: background 0.2s;" href="https://support.servicenow.com/kb?id=kb_article_view&sysparm_article=KB2704710" target="_blank" rel="noopener noreferrer">📘 KB2704710 — ServiceNow KB 📘 KB2709990 — AI-Assisted Support Details 📘 KB2507579 — Agentic AI Troubleshooting Landing Page Known Problems & Defects 🔴 PRB1998233 — Script include typo in ITSM AI Agent Collection 🔴 PRB1994390 — Status 503 Service Unavailable in logs 🔴 PRB1999517 — Conversational Catalog does not check Autonomous setting Key Properties (Advanced Scenarios) ⚙️ sn_nowassist_admin.ignoreFulfillerSubscriptionCheck ⚙️ sn_generative_ai.append_lang_in_prompt ⚙️ sn_ais_assist.enable_knowledge_graph_nlq ⚙️ sn_cmdb_gen_ai.import_set_query_limit ⚙️ sn_cmdb_gen_ai.ci_summarization_aggregation ⚙️ sn_cmdb_gen_ai.error_summary_max_length ⚙️ glide.ais.global_searchable_filter.kb_knowledge Key Script Includes ⚙️ sn_itsm_aia.InvestigatingAndResolvingIncidentsUtilSNC — check for typo on line 53 ⚙️ sn_aia.AiAgentRuntimeUtil — license validation check