AI & Prediction Pack

sn-ai-001

Predictive Triage for Bugs

Predictive Intelligence Priority: Critical Effort: Medium

To automatically predict and populate the Assignment Group and Priority for new bugs, reducing manual triage time and improving consistency. This addresses the fragmented intake process seen across multiple start states like 'New', 'Open', and 'Triage'.

Why this pack: This is a core Predictive Intelligence classification use case, using historical data to predict categorical field values, which is the primary focus of this pack.

Trigger

A bug record is created.

Conditions

·short_description IS NOT EMPTY

Actions

·Predict 'assignment_group' based on 'short_description', 'description', and 'project'.

·Predict 'priority' based on 'short_description', 'description', and 'project'.

·Set prediction confidence thresholds to only apply high-confidence predictions automatically, and show lower-confidence predictions as recommendations to the agent.

Evidence used

·High volume of tasks starting in 'New' (770), 'Open' (815), and 'Triage' (308), indicating a need for consistent initial routing.

·The 'Project' field shows strong concentrations (e.g., 'Keycloak' 14.6%, 'JBoss Enterprise Platform' 9%) which can serve as a powerful feature for a classification model.

·Inconsistent priority values are used ('Major', 'Critical', 'Blocker', 'High'), which a model could help standardize based on bug descriptions.

ServiceNow implementation

Target table

·rm_defect

Configuration area

·Predictive Intelligence

Configuration trigger

·Record Creation

Configuration conditions

·active = true

Implementation steps

·Navigate to 'Predictive Intelligence > Classification > Create New Solution'.

·Create a solution for 'Assignment Group' prediction. Name: 'Bug Assignment Group Prediction'. Table: rm_defect. Input Fields: short_description, description, project. Output Field: assignment_group.

·Train the solution using at least 30,000 historical bug records with populated assignment groups.

·Create a second solution for 'Priority' prediction. Name: 'Bug Priority Prediction'. Table: rm_defect. Input Fields: short_description, description, project. Output Field: priority.

·Train the priority solution using historical bug records.

·Activate the solutions and configure a Business Rule or Flow Designer action to call the prediction API on bug creation, populating the fields if the confidence score exceeds a defined threshold (e.g., 80%).

Fields used

·short_description

·description

·project

·assignment_group

·priority

Test cases

·Create a new bug with a description matching common 'Keycloak' issues and verify the assignment group is predicted correctly.

·Create a new bug with keywords like 'crash' or 'outage' and verify the 'Priority' is predicted as 'Critical' or 'Blocker'.

Rollback notes

·Deactivate the Business Rule or Flow that triggers the prediction.

·Deactivate the PI classification solutions from the solution definition records.

Requirements / prerequisites

·ServiceNow Predictive Intelligence plugin (com.glide.platform_ml).

·A minimum of 30,000 historical records on the target table with accurately populated output fields ('assignment_group', 'priority') to ensure model accuracy.

sn-ai-002

Similarity-Based Duplicate Bug Detection

Predictive Intelligence Priority: High Effort: Low

To identify and flag potential duplicate bugs at the time of creation. This helps reduce wasted effort on issues that are already being worked on or have been resolved.

Why this pack: This leverages Predictive Intelligence's similarity framework, an unsupervised learning technique to compare incoming text data against an existing corpus, which fits the AI & Prediction pack's scope.

Trigger

A new bug record is being created or is viewed by an agent.

Conditions

·short_description IS NOT EMPTY

Actions

·Display a list of similar open or recently closed bugs in the agent workspace form or via a UI action.

·Allow the agent to link the new bug to an existing one as a duplicate.

Evidence used

·A significant number of bugs are closed directly from early states: 'Triage -> Closed' (122 tasks) and 'New -> Closed' (96 tasks). This pattern often points to the closure of duplicate or invalid submissions.

·The process has a high number of variants (257 for 845 tasks), suggesting that similar issues may be handled in different ways, and identifying duplicates early can help standardize the process.

ServiceNow implementation

Target table

·rm_defect

Configuration area

·Predictive Intelligence

Configuration trigger

·Form Load / UI Action

Configuration conditions

·state IS ONE OF New, Triage, Open

Implementation steps

·Navigate to 'Predictive Intelligence > Similarity > Create New Solution'.

·Name the solution 'Duplicate Bug Detection'. Table: rm_defect. Fields: short_description, description. Filter: Active is true.

·Train the similarity solution.

·Activate the solution.

·Configure the 'Similar Open Incidents' (or clone for bugs) Agent Assist recommendation in Workspace to use this similarity solution.

·Alternatively, create a UI Action 'Find Similar Bugs' that calls the Similarity API and displays results.

Fields used

·short_description

·description

Test cases

·Create a new bug with a short description nearly identical to a recently created bug. Verify the existing bug appears in the Agent Assist panel.

·Create a new bug describing a known issue using different words. Verify the model still identifies the original bug as similar.

Rollback notes

·Deactivate the Agent Assist card configuration.

·Deactivate the PI similarity solution.

Requirements / prerequisites

·ServiceNow Predictive Intelligence plugin (com.glide.platform_ml).

·Agent Workspace configured for the bug management process.

·Sufficient historical data (10,000+ records) with good quality text in short description/description.

sn-ai-003

Bug Reopen Risk Prediction

Predictive Intelligence Priority: High Effort: Medium

To predict the likelihood of a bug being reopened after it is moved to a 'Resolved' state. This allows teams to take preventative action, such as performing a more detailed peer review or enhanced QA, on high-risk bugs to reduce rework.

Why this pack: This is a predictive classification model focused on mitigating business risk (rework). It uses AI to identify patterns in resolved tasks that correlate with future failure, a key use case for this pack.

Trigger

Bug 'state' changes to 'Resolved'.

Conditions

·state = Resolved

Actions

·Calculate and store a 'Reopen Risk' score (e.g., High/Medium/Low) on the bug form.

·If risk is 'High', display a prominent notification on the form.

·If risk is 'High', send a notification to the assignment group manager or QA lead.

Evidence used

·The 'Closed -> Reopened' transition occurred for 65 tasks, representing a significant source of rework.

·The overall process has a high rework percentage, consistently above 20% in historical metrics.

·The long duration from 'Resolved' to 'Closed' (avg 1778.32 hrs) could indicate delays or uncertainty in the resolution quality, which this prediction could help mitigate.

ServiceNow implementation

Target table

·rm_defect

Configuration area

·Predictive Intelligence

Configuration trigger

·State changes to Resolved

Configuration conditions

·state = Resolved

Implementation steps

·Create a true/false field on the bug table named 'u_reopened' and populate it historically using a script (true if reopen_count > 0).

·Navigate to 'Predictive Intelligence > Classification > Create New Solution'.

·Name the solution 'Bug Reopen Risk'. Table: rm_defect. Output Field: u_reopened. Input Fields: priority, project, reassignment_count, time_worked, short_description.

·Filter the training data to only include bugs that reached a 'Closed' state.

·Train and activate the solution.

·Create a Business Rule that triggers when state changes to 'Resolved'. The rule calls the prediction API.

·The Business Rule stores the prediction result in a new 'u_reopen_risk' field and can trigger notifications based on the result.

Fields used

·state

·reopen_count

·priority

·project

·reassignment_count

·time_worked

·short_description

Test cases

·Resolve a bug associated with a project known for high reopen rates. Verify the risk score is 'High'.

·Resolve a simple, low-priority bug. Verify the risk score is 'Low'.

Rollback notes

·Deactivate the Business Rule that triggers the prediction.

·Deactivate the 'Bug Reopen Risk' PI solution.

Requirements / prerequisites

·Predictive Intelligence plugin.

·Sufficient historical data of both reopened and non-reopened bugs to create a balanced training set (ideally 1000+ reopened examples).

·Fields that can act as strong predictors (e.g., complexity, reassignment count, comments/work notes analysis).

sn-ai-004

AI-Powered Resolution Notes Summary

Now Assist Priority: Medium Effort: Low

To use generative AI to automatically summarize the resolution steps from work notes and comments when a bug is resolved. This improves the quality of closure information, accelerates knowledge creation, and ensures a clear record of the fix.

Why this pack: This rule uses a core AI capability (Generative AI for summarization) to improve data quality and agent productivity, fitting squarely within the Now Assist and AI-driven recommendations theme of the pack.

Trigger

User clicks a UI Action ('Generate Resolution Notes') on a resolved bug.

Conditions

·state = Resolved

Actions

·Generate a summary of all 'work_notes' and 'comments'.

·Populate the generated summary into the 'close_notes' field for the agent to review and save.

Evidence used

·The high rework rate (over 20%) and 'Closed -> Reopened' transitions suggest that resolution information may be unclear or insufficient.

·The process has many complex paths (e.g., 'New -> ... -> Ready for QA -> QA In Progress -> Verified'), which generate a long history of work notes that are difficult to parse quickly.

·The long cycle times indicate complex problems where a clear summary of the resolution would be highly valuable for future reference.

ServiceNow implementation

Target table

·rm_defect

Configuration area

·Now Assist

Configuration trigger

·UI Action

Configuration conditions

·current.state == 'Resolved' || current.state == 'Closed'

Implementation steps

·Ensure the Now Assist for ITSM (or platform) plugin is active.

·Navigate to 'Now Assist > Admin > Skill Configuration'.

·Activate and configure the 'Resolution Notes Generation' skill for the target table (rm_defect).

·Map the input to the 'work_notes' and 'comments' journal fields and the output to the 'close_notes' field.

·The 'Generate Resolution Notes' UI action should become available on the form automatically based on the skill configuration.

Fields used

·work_notes

·comments

·close_notes

·state

Test cases

·Open a bug record with several detailed work notes. Resolve the bug and click the 'Generate Resolution Notes' button. Verify a concise, accurate summary is populated in the 'Close Notes' field.

Rollback notes

·Deactivate the 'Resolution Notes Generation' skill in the Now Assist Admin console.

Requirements / prerequisites

·Now Assist SKU and corresponding plugins must be active.

·Historical work notes must be of sufficient quality and detail to provide meaningful summaries.

·The organization has a 'low automation appetite', making an assistive, agent-triggered tool like this more culturally acceptable than a fully automated process.