General
Q: What is AI Agent Advisor?
AI Agent Advisor analyzes operational data in your instance to identify recurring patterns across records — each pattern representing a candidate process for automation. It prioritizes those opportunities by volume, matches them to existing OOTB agents and tools, and guides you through deploying or building the agents needed to act on them.
The outcome is a comprehensive blueprint for deployment that is grounded in data rather than assumptions.
It seeks to address questions such as:
- What should we automate first?
- What is the prospective ROI from implementing an AI Agent?
- What OOTB AI Agents and Tools are best for my business?
Q: How does it work?
AI Agent Advisor operates across three sequential phases:
Mine
Uses AI-based clustering to group similar incidents or cases, extract intents, and identify resolution steps that are consistent across the group.
Match
Prioritizes opportunities by volume of related record and maps each resolution step to existing OOTB and custom ServiceNow AI Agents, Workflows, and Tools, generating an actionable solution blueprint with pre-built solutions already identified.
Make
Provides native integration with AI Agent Studio to create and deploy AI Agents. Tools identified during the Match phase are automatically added. Curated datasets stemming form the records using during the Mine process are generated for use in testing.
Q: What do I need to gain access to AI Agent Advisor?
License and Version Requirements
- You have a license that includes access to Now Assist (Pro+ OR any New Commercial Model license)
- Zurich Patch 8+ or Australia Patch 1+
- Instance has at least version 3.0.5 of the Now Assist Center plugin (installed automatically with any Now Assist for [x] plugin)
Data Requirements
- Minimum 500 resolved/closed records in the incident or CSM case tables
- Records must have populated work notes and/or comments
Q: When should I use AI Agent Advisor?
Before purchase
Run against a cloned or demo instance with representative data to replace weeks of discovery workshops with an automated analysis. Review the outcomes to verify if the OOTB tools will serve your organization prior to purchase.
Implementation
Run against a clone of the production environment (e.g., Dev or Test) to deliver a prioritized automation backlog within hours and jumpstart agent creation using OOTB components.
Post-deployment
Review results regularly to identify new automation opportunities. Feed output into AI adoption plans, executive review decks, and developmental roadmaps. Re-run and compare coverage percentages over time to track improvement.
Configuration & Execution
Q: How do I configure AI Agent Advisor after installation?
From the AI Agent Advisor home page, click 
Q: What record types are currently supported?
Incident and CSM Case tables are supported. Additional record types (including HR Cases and custom tables) are on the roadmap.
Q: Is the 500-record minimum configurable?
No, it is a hard minimum and cannot be modified. However, it is only a floor — most enterprise Incident and CSM Case tables will have thousands or millions of records, so this will rarely be a constraint for production data.
Q: How do I start AI Agent Advisor?
It will automatically run once installed, and then again every 90 days. This is modifiable from within the AI Agent Advisor Settings page, in the section labelled Schedule.
If you wish to execute it manually, you can also find a button labelled Execute Now in the setting page. The process of the execution will be displayed.
Usage
Q: How should I interpret the dashboard results?
Each line in the table reflects an automation opportunity discovered by analyzing the records in your instance. We present the following details for each one:
- The estimated savings results are calculated using the values set in the Potential Value Calculator found in the settings page along with the number of associated records that relate to this opportunity. Do take the time to tailor these values to what aligns with your organization.
- The number of requests column holds the number of records associated to this particular opportunity – the more there are, the more frequent this issue has occurred. We recommend reviewing example records within a cluster to verify the generated intent accurately represents the record group before acting on it.
- The number of AI assets indicates how many OOTB tools can address the process. A lower number can mean higher implementation complexity as it requires custom tools to be built.
Automation opportunities are sorted by record volume. Focus on the those with the greatest volume of records for highest ROI impact.
Q: Can I edit the resolution steps generated by the analysis?
Yes. Resolution steps can be created, reordered, and edited within the opportunity view. If a new step is added manually, a matching action is available to discover AI assets for that step. Original values can be restored at any time by clicked the Revert All button.
Q: How do I create an AI Agent from AI Agent Advisor?
From your opportunity page, you can click the large Continue in AI Agent Studio button.
After you can confirmed your action, it will begin to generate an AI Agent using the details of your opportunity. Your generated AI Agent will have the following auto-populated:
- Name
- Description
- Role
- Steps (derived from your list of resolution steps)
We also perform the following:
- Any matched tools are added automatically to the AI Agent
- Do note that you must have the associated AI Agent installed. If not, we will prompt you to install them on the opportunities page. If you proceed without doing so, you will find that the tool is not added.
- A dataset is created that contains all the records that were identified as having an intent that matches the current opportunity. This dataset can be used in the automated evaluation process, so you can test that your AI Agent executes well against the records that originally highlighted the issue.
Additional Details
Q: How many Now Assist assists does it consume?
None.
Q: Which LLM providers does it use?
By default, a combination of NowLLM and GPT Large. If a global default LLM provider has been chosen for the instance, AI Agent Advisor will use that instead.
Q: Is it available in GCC or Domain Separated environments?
GCC testing is currently underway. Domain Separated environment support is on the roadmap.
Q: What is the difference between AI Agent Advisor and AI Ops LEAP?
AI Ops LEAP is recommended for use with playbook generation. AI Agent Advisor is the preferred tool for agentic opportunity discovery.
Q: What is the difference between AI Agent Advisor and Process Mining?
AI Agent Advisor is for discovering where to start with automation; Process Mining is for diagnosing why an existing process is underperforming.
Q: Is there a minimum word count requirement on individual records?
No specific word count per record is enforced. The system looks at the entirety of each record and determines which records to sample from. The important requirement is that records have populated work notes and/or comments — records without this content cannot yield meaningful intent extraction.
Q: Do OOTB AI agent plugins need to be activated for matching to happen?
No. The system scans all available OOTB agents and tools regardless of whether they are installed on the instance. If a resolution step maps to a tool from an agent that is not installed, the UI will prompt the user to install that agent before proceeding. The matching itself is not blocked by installation status.
Q: How does the matching logic work — does it just match on agent descriptions?
No. Agent descriptions can be quite broad, so the system primarily matches on tools — the specific capabilities each agent exposes. The matching also considers agent proficiency descriptions (what the agent says it can do), but tools are the primary signal because they define what the agent can actually execute. The system matches each resolution step to the most relevant tools first, then surfaces the agents those tools belong to.
Q: What happens when a resolution step has no matching OOTB tool?
The UI flags the step as unmatched and indicates that a custom tool will need to be created. This could mean building a subflow, creating a custom skill, or creating a custom AI agent. Unmatched steps are still valuable — they tell you exactly what needs to be built and where the gaps are before you start development.
Q: Can AI agents be matched across all ServiceNow product areas?
The matching covers all OOTB agents and tools across the platform. However, if a customer is not licensed for a particular tool, it will not be added to the generated agent. The UI performs license checks and surfaces which tools the customer has access to and which require an additional install or license.
Troubleshooting
Q: Where do I find error messages?
Navigate to the AI Agent Advisor settings page using this button 
Q: The analysis is failing or producing poor results. What should I check?
The most common cause is insufficient record volume for the clustering algorithm to run.
To remedy this, configure the record filter on the Settings page to increase the number of included records.
Do also ensure the records have populated work notes or comments — records without this data cannot yield meaningful intent extraction.
Q: How do I cancel an execution?
- In the Filter Navigator, search for Active Transactions.
- In the list, locate the record with Agent Miner or AI Agent Advisor in the URL column.
- Select the record, then choose Kill from the Actions on selected row dropdown.
-
Navigate to the sn_agent_miner_executions table.
-
Find the record with Status InProgress.
-
Click the status field, type Failed, then click the green checkmark to save.
Without step 6, new executions will not trigger as only one execution can be in progress at a time.
Q: What if some tools are missing from the generated agent?
Missing tools indicate the source agent is not installed on the instance. The platform will flag which agent needs to be installed. Once installed, you can go back and manually add the tool to the generated agent.
Q: Is there a way to evaluate the generated agent against real data before deploying?
Yes. At GA, AI Agent Advisor automatically generates an evaluation dataset from the cluster records. When you navigate to the auto-evaluation tool, this dataset is pre-selected — the customer just runs auto eval without needing to manually choose or prepare data. This lets you validate that the agent actually resolves the incidents that triggered the cluster before committing to a production deployment.
