Understanding Your AI Readiness Findings: A Practical Guide to Every Result

[Sharon_Barnes]

Understanding Your AI Readiness Findings:

A Practical Guide to Every Result

Table of Contents

• How to Read Each Finding
• Findings by Section
• AI Search
• Virtual Agent
• ITSM
• CSM
• HRSD

Articles Hub

https://sn.works/coe/nachub

 

Requirements

Family Release Australia P3
Zurich P10

Now Assist Center  4.0.2

Role sn_na_center.nac_admin

What You're Looking At

AI Readiness (AIR) is a technical scanner that inspects your ServiceNow instance before you implement Now Assist or Agentic AI. Think of it like a pre-flight checklist. It surfaces potential friction points, helps you understand what work might be needed, and gives you talking points to have with your technical team.

For more info on what AIR is and where to find it, see: Using AI Readiness in Now Assist Center

This guide explains what each AIR finding actually means for you, what outcome it leads to, and what your technical team should do about it.

One important thing upfront: AIR is tuned to be thorough. It would rather surface something that might be relevant than miss something real. Most enterprise instances will have some findings. The goal is not to get a perfect score. The goal is to understand which findings actually require work before go-live, and which are just informational.

How to Read Each Finding

Every finding has three possible outcomes. Learn to spot them quickly:

Green check (✅) means no action needed

The scan found no issues in this area. The AI feature should work as expected out of the box. Move on.

Yellow warning (⚠️) means review recommended

Something was found that could affect the AI feature, but might not apply to your setup. Your technical team needs to look at it and decide whether it matters in your context. Not a blocker by itself.

Red X (❌) means confirmed finding, action needed

The scan found a specific customization or gap that is known to create friction. Your technical team needs to assess and address it.

Findings by Section

Below are all AIR findings organized by product area. Each finding explains what it means, what outcome it leads to, and what action to take.

AI Search

Is AI Search live?

What it means: AI Search has not been activated on your instance. Without it, users rely on keyword-only search. It's like searching a dictionary by flipping through pages instead of using an index. You find things, but it's slow and you miss context. What to do: Enable AI Search. It's the foundational layer that everything else sits on top of.

Is AI Search configured for Next Experience?

What it means: Next Experience is where most users will interact with the platform going forward. If AI Search hasn't been configured for it specifically, users on Next Experience will fall back to basic search, missing the full benefit of AI-powered results. What to do: Configure AI Search for the Next Experience portal to ensure users are getting the good version of search, not the legacy one.

What are the experiences in which AI Search is configured?

What it means: AI Search deployment might have gaps across different portals or experiences. Users on unconfigured portals won't benefit from AI-powered search and will fall back to basic results. What to do: Map which portals are and aren't covered. Users on unconfigured portals get a noticeably worse search experience. Extending coverage to all relevant experiences is usually straightforward once the gaps are visible.

Is the latest version of the AI Search widget used?

What it means: An older version of the AI Search widget is deployed. Running an outdated version means you're missing bug fixes, performance improvements, and features that Now Assist depends on to work correctly. What to do: Update to the latest out-of-box widget version. This is one of the simpler fixes on the list but it's mandatory.

How many published/valid articles?

What it means: Published article volume is being measured. AI Search quality depends directly on the richness of your knowledge base. More well-written, published articles means better answers and more accurate AI-powered suggestions. What to do: If article coverage is thin, work with the Knowledge team to publish and validate more articles. This is a Knowledge team engagement opportunity as much as a technical one.

What are the Knowledge Bases and KB Categories to target?

What it means: AI Search knowledge targets may not be defined. Without targeted configuration, it may surface irrelevant results or miss the most valuable content. What to do: Define which KBs and categories should be prioritized for AI Search indexing. You want the AI surfacing your best content, not pulling from everything including outdated or low-quality sources.

Are there any group restrictions on knowledge bases?

What it means: Some knowledge articles have read restrictions limiting who can see them. That's normal and expected for sensitive content. What matters is confirming that AI Search respects those boundaries and doesn't accidentally surface articles to people who shouldn't see them. What to do: Your team should audit KB permissions to verify AI Search is enforcing access correctly.

What are the top searches?

What it means: The most frequent search queries are identified. Your top searches are a map of what your users struggle with most. What to do: Validate whether AI Search is returning high-quality, relevant results for these specific terms. If it doesn't, enriching the associated knowledge articles or tuning search profiles for these queries will have an outsized impact on overall user experience.

What are the top articles viewed?

What it means: High view counts on certain articles signal strong user demand in those topics. If those articles are outdated or poorly written, AI Search will surface them prominently, amplifying low-quality content. What to do: Review the quality of your most-viewed articles before go-live. Your most-viewed articles are going to get even more visibility once AI Search is surfacing them proactively. Making sure those specific articles are high quality is one of the best pre-go-live investments the Knowledge team can make.

Virtual Agent

Is classic Virtual Agent live?

What it means: The Virtual Agent capability is currently active on your instance. This is the foundation for conversational fulfillment and automated request handling. What to do: If not yet live, activation is a prerequisite. If it is live, you can build conversational capabilities and automation on top of it.

Is NLU live?

What it means: Natural Language Understanding (NLU) is enabled, allowing the VA to understand user intent from conversational language rather than just matching keywords. Without it, the VA is limited to basic keyword matching. What to do: Activate NLU if not already live. It's a significant capability enabler for conversational fulfillment.

If live, how many active topics?

What it means: The Virtual Agent has a set number of active topics configured. Topics are the conversation flows that handle specific user intents. The breadth of topics determines the range of requests the VA can handle. What to do: Review active topics against user requests to identify any high-demand topics that are missing or inactive.

How many active catalog items are available?

What it means: The Virtual Agent has visibility into how many active catalog items exist for potential fulfillment. This tells you the scope of what the VA can potentially assist with. What to do: Use this inventory to prioritize which items should be made available to Virtual Agent and which ones to configure as conversational.

How many are conversational?

What it means: Out of all available catalog items, a specific portion has been set up for conversational fulfillment. The higher this percentage, the better the user experience for multi-step requests. What to do: Look at expanding conversational coverage to more catalog items. Start with the high-demand ones and build from there.

Are there any common conversational stoppers?

What it means: Conversational patterns that cause the Virtual Agent to reach a dead end or fail are being identified. These stoppers interrupt the user experience and reduce agent effectiveness. What to do: Work with your VA team to identify and resolve these conversational dead ends through topic refinement and better fallback handling.

What are the top 10 catalog items?

What it means: The most frequently requested catalog items are identified. These represent what users ask for most and where automation and Virtual Agent assistance will have the biggest impact. What to do: These are your high-ROI candidates for Virtual Agent fulfillment and conversational setup. Focus your VA configuration efforts here first.

Are there conversational items in the top 10 catalog items?

What it means: The most-requested catalog items may or may not have conversational fulfillment set up. Conversational items in high-demand categories offer a much better user experience than traditional catalog interaction. What to do: Look at converting your top 10 requested catalog items to conversational flows. This is where users notice the difference immediately.

Is there an opportunity to convert any of the top 10 catalog items to conversational?

What it means: The system is evaluating whether your most-requested catalog items could benefit from conversational setup. Multi-step or complex requests are especially good candidates. What to do: Identify which of your top 10 items have multi-step fulfillment flows and prioritize those for conversational conversion.

Automation found?

What it means: The Virtual Agent has identified automation opportunities within existing interactions. These are chances to streamline what users currently do manually. What to do: The VA team has already done the work to identify what's automatable. Now it's about prioritizing which ones to turn on first. These are essentially free deflection waiting to be activated.

What are the Service Catalog names?

What it means: The system is inventorying the service catalogs available on your instance. Users can find and request items through Virtual Agent only if those catalogs are exposed to it. What to do: Review which catalogs are exposed to Virtual Agent. If some aren't, and they should be, configure exposure to expand what users can request through the VA.

ITSM

Do incident customizations exist?

What it means: Custom fields or logic have been added to the Incident table. Most enterprise instances have these. The question is whether any touch the specific fields that AI reads. What to do: Review which customizations exist and confirm they don't interfere with the core incident fields that AI uses.

Are there updates to the OOB state/field choice values?

What it means: Standard incident state values have been customized. AI relies on standard values to understand where a ticket is in its lifecycle. Modified values can cause the AI to misread record state. What to do: Audit which values were changed and confirm the AI can still accurately interpret incident state.

Any special field configuration to be aware of?

What it means: Non-standard field configurations have been found that may affect how AI skills read incident data. What to do: Review these configurations to ensure AI can access the information it needs to generate accurate summaries and recommendations.

Are Short Description, Description, Additional Comments and Work Notes standard or are custom fields used?

What it means: Agents might be entering information in custom field substitutes instead of the standard ones. AI is designed to read standard fields. What to do: Either redirect agents back to standard fields or configure the AI to read the custom ones your team actually uses.

What is the volume of past incidents with at least 50 words?

What it means: Incidents with sufficient narrative text give the AI rich material to work with for summarization and knowledge generation. Low volume means thin records. What to do: Encourage agents to write more complete incident notes going forward. Setting that expectation before go-live shapes behavior at exactly the right moment.

What is the percentage of incidents where Short Description, Description, State, Priority, Work Notes, Additional Comments and Resolution Notes are filled in?

What it means: Key incident fields may have low completion rates. AI uses these fields to generate summaries and suggest resolutions. Missing data directly degrades AI output quality. What to do: Set field completion standards and consider making critical fields required. Even modest improvements before go-live make a visible difference in what the AI can produce.

How many past incidents are there with no resolution details?

What it means: Many resolved incidents have no resolution notes captured. Resolution notes are the raw material for AI-generated knowledge and for surfacing relevant fixes in the future. What to do: Start requiring resolution notes before incident closure. This is one of the highest-ROI behavior changes you can introduce before go-live.

Is Create Knowledge UI Action custom in both UI16 and Workspace UIs?

What it means: The Knowledge creation UI action may be customized in one or both interfaces, potentially preventing the AI from using it to generate knowledge from incidents. What to do: Verify the action works in both interfaces or configure it to support AI knowledge generation.

Are there additional mandatory fields as part of Knowledge creation that need to be factored?

What it means: Your knowledge creation process may require additional fields beyond the standard ones. The AI needs to know about these to successfully generate knowledge articles. What to do: Identify any extra mandatory fields and configure the AI knowledge generation process to handle them.

Are agents assigned out of box roles or do custom roles need to be factored?

What it means: Custom roles may not include the permissions required by Now Assist AI features. What to do: Ensure custom roles include all necessary Now Assist permissions so agents can access AI features without blockers.

Are there improvement opportunities in customers Knowledge Base?

What it means: Knowledge base content quality or coverage could be improved. Improving article quality and coverage enhances AI Search results and AI's ability to recommend relevant resolutions. What to do: Engage the Knowledge team and turn tribal knowledge into well-structured articles the AI can actually use. This is a good moment for that engagement.

Are there any custom values or alterations on the Incident table that the AI agent depends on for category, subcategory, or configuration item assignment?

What it means: The AI agent uses standard category and subcategory values to route tickets. If those values have been customized, the agent may be working with a different vocabulary than it was designed for. What to do: Align custom values with what the agent expects, or configure the agent to understand your custom vocabulary.

Are there custom Business Rules that restrict or alter updates to category, subcategory, or CI fields?

What it means: Business Rules you've built might silently block or override the agent's attempts to update fields. The agent looks like it's running fine, but its updates are being blocked behind the scenes. What to do: Review each Business Rule for compatibility with agent write operations. This is tricky to catch after go-live, so it's worth the upfront check.

Are there any custom ACLs or access controls restricting the AI agent user from updating the category, subcategory, configuration item, or work notes fields?

What it means: Access restrictions may prevent the incident agent from writing to fields it needs. Without field-level access, the agent cannot perform its core functions. What to do: Review and grant the necessary permissions to the agent user. This is one of the most common agentic blockers and also one of the easiest to fix.

Are there any customizations done on the 'Not Allowed CI Actions' table that could affect the AI agent?

What it means: This table defines what the system won't do automatically with CIs. Customizations may inadvertently block the AI agent from performing legitimate CI-related actions during incident resolution. What to do: Review the table to ensure agent actions are not unnecessarily restricted. This is usually not intentional but easy to miss.

Are there any custom triggers already present that execute on Incident creation/update and conflict with the AI agent trigger?

What it means: Custom triggers may conflict with the agent's trigger logic, causing race conditions, duplicate actions, or agent failures. What to do: Identify and reconcile these triggers before going live. This is coordination work, not a tear-down.

Are there ACLs or customizations made to the standard Change Request table, particularly affecting Implementation Plan, Test Plan, and Backout Plan fields?

What it means: ACLs or customizations may prevent the AI agent from reading or writing the plan fields. These fields are critical for AI-assisted change management. What to do: Grant the agent user permission to write to these fields. This is typically a focused permissions fix rather than a major change.

Are there customized business rules or server-side scripts on the Change Request table that might alter, revert, or restrict updates to plan fields?

What it means: The agent writes values to Implementation, Test, and Backout plan fields. A script immediately overwriting them means the agentic action is silently lost. What to do: Audit these scripts and add logic to respect AI-generated values. Silent overwrites are the hardest issue to debug after go-live.

Have journal fields or activity logs for Change Requests been customized or disabled?

What it means: Activity logs have been customized or disabled. The agent relies on the full history to understand context. Degraded logging reduces the agent's ability to reason about past actions. What to do: Restore full activity logging or ensure the agent can still access enough historical context.

CSM

Do case customizations exist?

What it means: Custom fields or logic have been added to the Case table. The question is whether any touch the specific fields that AI reads. What to do: Review which customizations exist and confirm they don't interfere with core case fields.

Are there updates to the OOB state/field choice values?

What it means: Standard case state values have been customized. AI relies on standard values to understand case lifecycle. What to do: Audit which values were changed and confirm the AI can interpret case status accurately.

Any special field configuration to be aware of?

What it means: Non-standard field configurations may affect how AI skills read case data. What to do: Review these configurations to ensure the AI can access the information it needs.

Are Short Description, Description, Additional Comments and Work Notes standard or are custom fields used?

What it means: Agents might be using custom field substitutes. AI is designed to read standard fields. What to do: Redirect agents to standard fields or configure the AI to read your custom ones.

What is the volume of past cases with at least 50 words?

What it means: Cases with sufficient narrative text give the AI rich material for summarization. Low volume means thin records. What to do: Encourage agents to document cases more fully. Expectations set before go-live stick.

What is the percentage of cases where Short Description, Description, State, Priority, Work Notes, Additional Comments and Resolution Notes are filled in?

What it means: Key case fields may have low completion rates. AI uses these for summarization and recommendations. Missing data degrades AI quality. What to do: Establish field completion standards. Even modest improvements before go-live make a visible difference in AI output.

Is Create Knowledge UI Action custom in both UI16 and Workspace UIs?

What it means: The Knowledge creation action may be customized, preventing the AI from generating knowledge from cases. What to do: Verify the action works in both interfaces or configure it for AI knowledge generation.

How many past cases are there with no resolution details?

What it means: Many resolved cases have no resolution notes captured. Resolution notes are the raw material for AI-generated knowledge and for surfacing relevant solutions in the future. What to do: Start requiring resolution notes before case closure. This is one of the highest-ROI behavior changes you can introduce before go-live.

Are there additional mandatory fields as part of Knowledge creation that need to be factored?

What it means: Knowledge creation may require additional fields. The AI needs to know about these for successful knowledge generation. What to do: Identify extra mandatory fields and configure AI knowledge generation to handle them.

Are agents assigned out of box roles or do custom roles need to be factored?

What it means: Custom roles may not include permissions required by Now Assist. What to do: Ensure custom roles include necessary Now Assist permissions so agents can access AI features.

Are there improvement opportunities in customers Knowledge Base?

What it means: Knowledge base content quality could be improved. Better articles enhance AI Search and AI recommendations. What to do: Engage the Knowledge team to improve coverage and quality.

Are there any custom fields or alterations on the Interaction or Case tables that the AI agent depends on?

What it means: Custom fields may interfere with the AI agent's ability to identify customer context and issue type. What to do: Review customizations to ensure the agent is reading from the right sources for customer context.

Are there any custom Business Rules on the Interaction or Case tables that might impact the creation, updating, or linking?

What it means: Custom Business Rules may affect how interactions and cases are created or linked. The AI agent depends on clean record relationships. What to do: Reconcile rules so the agent and existing automations don't conflict.

Are there any custom ACLs (Access Control Lists) directly or do custom roles that need to be factored?

What it means: Custom ACLs may restrict the AI agent's access to fields it needs. What to do: Grant the agent user the specific field-level access required to triage cases and interact with customers.

Have customizations been made to the email handling logic that might impact verifying email sender details?

What it means: Email sender verification is how the agent confirms it's talking to the right person. Customized email logic might break this, introducing accuracy and security risks. What to do: Review email handling customizations to ensure sender verification still works.

Have you altered or customized the AI search profiles, specifically the Quick Action KB Search Profile?

What it means: The KB search profile is the agent's lens for finding relevant articles during triage. If modified, the agent may look in wrong places or with wrong filters. What to do: Review what changed and whether it affects triage relevance. Usually a targeted configuration fix.

HRSD

Do HR Case customizations exist?

What it means: HRSD is typically the most customized module. The question is whether customizations touch specific fields the AI reads. What to do: Review customizations for AI skill compatibility. Usually not a deal-breaker, just verification.

Are Short Description, Description, Additional Comments and Work Notes standard or are custom fields used?

What it means: Agents might use custom field substitutes. AI reads standard fields. What to do: Redirect agents to standard fields or configure the AI for custom ones.

Any special field configuration to be aware of?

What it means: Non-standard field configurations may affect how AI skills read HR case data. What to do: Review configurations to ensure the AI can access needed information.

What is the volume of past HR cases with at least 50 words?

What it means: HR cases may naturally have less narrative text since they're process-driven. The AI is designed with this in mind. What to do: Set realistic expectations about summarization depth with the customer upfront. This prevents disappointment after go-live.

Are there updates to the OOB state/field choice values?

What it means: Standard HR case states have been customized. AI relies on standard values to understand case lifecycle. What to do: Confirm the AI can map custom states to its understanding of case progression.

What is the percentage of HR cases where the Short Description, Description, State, Priority, Work Notes, Additional Comments and (for resolved HR cases) are filled in?

What it means: HR case field completion rates may be low. HR cases are process-driven, not note-driven, so expectations differ from ITSM. AI accounts for this, but missing data still limits output. What to do: Introduce consistent field completion standards. Even modest improvements help AI effectiveness.

Are there any ACLs or custom access control configurations on the standard HR case table that restrict the AI agent user, particularly affecting fields such as Opened For, Opened By, Assignment Group, Assigned To, Short Description, Description, Work Notes, Additional Comments, Priority, State, and HR Service?

What it means: Access restrictions may prevent the AI agent from reading or writing fields it needs. What to do: Grant the agent user field-level access to perform its assigned actions.

Are there any custom fields, values, or table-level modifications in HRSD that the out-of-box Agentic AI depends on (Opened For, Opened By, Assignment Group, Assigned To, Short Description, Description, Work Notes, Additional Comments, Priority, State, and HR Service)?

What it means: HR-specific customizations may not map cleanly to what the AI expects. HR processes are unique. What to do: Review custom fields to ensure the AI can read the context it needs or configure it for your custom setup.

Are key HR Case fields consistently populated?

What it means: Consistency in field population directly impacts AI reliability in understanding and processing HR cases. What to do: Review field population patterns and establish standards for critical fields to ensure the AI has consistent data to work with.

Are there custom Business Rules that restrict or alter updates to Opened For, Opened By, Assignment Group, Assigned To, Short Description, Description, Work Notes, Additional Comments, Priority, State, or HR Service fields?

What it means: Custom Business Rules might interfere with the AI agent's ability to update fields correctly. What to do: Review rules for compatibility with AI operations and reconcile conflicts.

Do any custom journal fields exist on HR case that could be mistaken for Work Notes or Additional Comments?

What it means: If custom journal fields exist and agents are using them instead of standard ones, the AI may miss critical context about case progression. What to do: Audit journal field usage and ensure agents are documenting in the standard Work Notes and Additional Comments fields that the AI monitors.

Are there custom triggers (Flow Designer, Workflow, or Scripted triggers) already present that execute on HR case creation/update and conflict with the OOB AI agent trigger?

What it means: Custom triggers may conflict with the agent's trigger, causing failures or duplicate actions. What to do: Identify and reconcile these triggers before go-live. This is coordination work.

Next Steps

Look through your AIR results and use this guide to understand what you're seeing. Prioritize your red X findings first. Those are confirmed work items that your technical team needs to address.

For yellow findings, involve your technical team to decide whether they actually apply to your instance. Many will turn out to be false positives once your team investigates them. Nine times out of ten, the answer to a yellow finding is "that doesn't apply to how we use the system."

Most findings are fixable and shouldn't block your go-live. You're just making sure you know about them upfront instead of discovering them during implementation.

Explore more in the Now Assist Center Hub