Brazil, Australia Patch 5, Australia Patch 3, Zurich Patch 12, Zurich Patch 10
Generative AI
The L1 IT Service Desk AI specialist will autonomously resolve high-volume, routine issues by triaging the incident, identifying the issue, researching the resolution, and resolving it end-to-end. It runs inside ServiceNow workflows you already operate, follows the policies and guardrails you set, and reassigns the moment something falls outside what it's cleared to handle.
Getting Started:
- Implement with a ServiceNow expert. Please reach out to your account representative to plan your implementation with a ServiceNow professional implementation expert. This will ensure proper configuration of triage, knowledge sources, confidence thresholds, and governance to make autonomous resolution safe and effective in your environment.
- Identify the right candidates first. Before deployment, use the L1 IT Service Desk AI Specialist Onboarding Tool to determine which clusters are primed for autonomous resolution. It assesses your environment and pinpoints where the AI Specialist will perform well based on your KB and catalog coverage, so you start with high-confidence use cases and expand from there.
- End-to-end resolution. A coordinated team of specialized AI agents works together to understand the problem, find the fix, and propose the resolution to the caller. Each agent has a defined role, so the incident moves all the way to resolution without a person having to drive it.
- Investigates like your best agents. It draws on knowledge base articles, past incidents, known errors, and catalog items to reach a resolution, and asks the requester follow-up questions when it needs more to go on.
- Request fulfillment. Some incidents come in as requests. In those cases, the system proposes or submits the appropriate request item, such as a software access request. When a person must confirm details first, it provides the exact link required instead of a dead end.
- Confidence-gated action. Before it acts, the system evaluates and scores its own proposed answer. When confidence is high, it proceeds; when it is not, it continues researching or escalates. A weak or partial-match answer never reaches the caller. Instead, the incident is reassigned to a service desk agent.
- Seamless reassignment to a service desk agent. When confidence falls below the defined threshold, the system reassigns the incident to the designated assignment group or person, with the full history of what it already attempted attached.
- Two execution modes. Run each specialist in autopilot, where it posts resolutions directly, or in copilot, where a person approves the resolution first. Switch between them per specialist at any time, so you can validate behavior before granting full autonomy.
- Monitor the AI Specialists' performance. Track its auto-resolution rate, how many incidents it manages through the full lifecycle, which knowledge base articles it uses to resolve them, and more.
- Complete traceability. Every search and step the system performs is recorded. If a resolution falls short, your team can review exactly what happened and why, and address the root cause rather than work from assumptions.
- Choose your controls. Decide how the AI Specialist handles an incident at every stage. Configure how it classifies and routes each incident, which knowledge sources it draws on, how deeply it researches, and when it steps back for a person to confirm. Sensitive or high-risk work can be held for review while routine work runs on its own
- Governed by default. The AI specialist and its team of AI agents is automatically registered and kept current in ServiceNow's AI Control Tower, alongside the rest of your governed systems.
New
- Requesters get follow-up questions for mandatory catalog variables the AI Specialist couldn't fill from the incident; once all are answered, it submits the request on their behalf. AI Admins cap the exchange via Replies before reassigning, after which the requester gets a link to a draft request with answers so far.
- Requesters who submit a drafted catalog request from the AI Specialist's link have their incident resolved automatically, with a resolution note appended to close notes.
- Service Desk Managers get a Catalog tab on the AI Specialist Performance dashboard covering catalog items used, requests drafted and submitted, AI vs. requester submissions, and outcome cards (Zero touch drafts, Abandoned in drafts, Cancelled after submit, Reassigned to SDA, Drafts with incomplete variables). Hidden when the catalog slot-filling API is unavailable.
- AI Admins can switch response templates between fixed and dynamic mode, where the AI Specialist writes activity notes, comments, work notes, and close notes from free-form admin instructions.
Changed
- When a requester escalates to a human agent, catalog request items the AI Specialist submitted are closed as incomplete with a cancellation comment instead of left open.
- Requesters reassigned after a follow-up reply get a reassignment note written by the AI Specialist instead of a generic scripted message.
- Service Desk Managers see Coverage rate calculated as closed incidents durably resolved by the AI Specialist over those assigned to eligible groups, and Durable auto-resolve rate against incidents attempted. Subtitle and tooltips reflect the new calculations.
- AI Admins with custom incident states get hold reasons, close codes, and cross-group routing state resets that follow the worker's state mapping.
- AI Admins see the Auto-submit catalog requests setting reworded around drafts: the AI Specialist submits only when it filled every mandatory variable; otherwise the employee completes the draft.
Fixed
- Performance dashboard:
- Incidents closed by eligible assignment groups counted every group and included canceled incidents.
- Incident outcomes total counted only resolved, reopened, and reassigned autopilot incidents, so it didn't match the routing funnel.
- Incident outcome counts could differ from the linked list when the viewer's time zone differed from the PA collection job's.
- KB articles tab summary metrics didn't display because data access was granted to the worker role instead of the AI worker manager role.
- KB articles tab took 23-28 seconds to load and didn't reliably refresh when selected.
- Incidents resolved after a positive requester reply left the Save activity notes step showing as ongoing for three days.
- A "no article found" knowledge feedback task could be created even when the AI Specialist cited a matching article whose link was removed from the customer note for access reasons.
- KB article attachments weren't used when composing responses, and one failed knowledge search request returned empty results for every query in that search.
- Incidents the AI Specialist reopened itself were counted as reopened by a person, marking its proposed solution as rejected.
- Outcomes for incidents handled before the AI Specialist moved to its own application records could be downgraded from resolved to reassigned by the post-execution metrics job.
- Coaching application files were missing after upgrading.
Removed
- Performance dashboard
- Copilot gauge removed from At a glance; the section is hidden in Copilot mode.
- Coverage autopilot numerator/denominator scorecards replaced in the detailed view by cards for incidents closed by eligible groups, attempted, and durably resolved.
Not applicable for this application version.
IT Service Management Prime
Otto for IT Service Management
Australia Patch 3+ or Zurich Patch 10+
Other Requirements:This version does not support GCC environments
This version does not support Domain Separation