Table of contents
Highlights
- Case deflection measures whether an employee avoided filing a ticket. Whether their request actually got resolved is a separate question, which is why deflection metrics can improve while HR workload stays flat.
- Agentic AI is designed to close that gap by reasoning through HR requests and completing the underlying task, like submitting PTO requests, updating benefits, or provisioning onboarding access, within defined guardrails.
- Unlike chatbots or generative AI, agentic AI can act with governance oversight across integrated systems like Workday and ServiceNow, executing multi-step workflows while applying existing policies and permissions along the way.
- High-impact starting points include onboarding, benefits enrollment, performance management, and compliance tracking — all high-volume, rules-based HR workflows well-suited to end-to-end automation.
- Moveworks' Reasoning Engine and Tool Studio are built to support full resolution, helping complete HR workflows while maintaining human oversight and control, so employees get the outcomes they need.
When an employee types, "How much PTO do I have left?", your AI assistant answers right away.
Your dashboard logs the interaction as a win because it’s a deflected case, with no ticket needing to be submitted and answered by your HR team.
But the employee still has to open another tool or portal, find the right form, and submit the request. Sure, their balance question was answered, but how much fanfare does that deserve when most of the process itself is still manual?
You’ve probably already added chatbots and knowledge base search capabilities (tools that match questions to stored articles) for your employees. Ticket numbers might have improved, but your HR team still has to process PTO requests, benefits changes, and onboarding steps by hand.
In fact, 51% of HR professionals report spending half their work week on admin, and 91% of them are using two or more systems to complete a single HR-related task.
A better benchmark is resolution. Did the task actually get done? That means employees get support in the moment, and your HR team gets time back for more people-focused work like engagement and retention.
Below, you’ll see why deflection is an incomplete measure, how agentic AI can complete HR tasks from request to finish, and how to tell whether your tools resolve work or just reroute it.
What is agentic AI in HR?
Agentic AI in HR is AI that can reason, plan, and act across HR systems to complete tasks within defined guardrails.
It works through AI agents: software components built for specific tasks. Think of them as digital coworkers. One might file a PTO request on an employee’s behalf, while another nudges a new hire to finish required training.
How it differs from chatbots and traditional automation
Traditional automation follows fixed rules. Basic chatbots answer from scripts or a knowledge base. Generative AI (GenAI, the kind that produces text, code, or summaries from a prompt) can write well but typically stops before taking action.
Agentic AI is able to interpret a goal and act across systems, applying your policies and staying within your defined guardrails.
A traditional AI assistant might point employees to a benefits form. An agentic one could locate it, complete it, and file it.
From deflection to resolution: Why the old benchmark falls short
Case deflection measures success by whether an employee avoided filing a ticket, even if their need went unmet.
Say an employee asks about parental leave. They get a link to the policy… case closed. Your dashboard logs a deflected case. But the reason they were probably asking (to submit a leave request) is still waiting.
That's why deflection can look like an improvement on paper while HR workload stays the same. The system answered a question without completing the task. Then, when the ticket is submitted (manually), 84% of senior leaders surveyed say the request will often need additional follow-up or rework.
Resolution asks whether the task was completed, with governance (policies, permissions, and approvals that guide each step) and an audit trail that records the actions taken.
Deflection asks: Did the employee avoid filing a ticket?
Resolution asks: Was the PTO request submitted, routed for approval, and logged?
How agentic AI delivers full resolution
Three capabilities work together here:
- Perception: The AI is designed to pull signals from connected systems, such as Workday (an HR information system, or HRIS, that stores employee records) and ServiceNow (a platform that manages workflows and cases). Integration capabilities, the connections that let software systems exchange data, make this possible.
- Reasoning: A reasoning engine is built to plan multi-step workflows, select the right tool for each step, and work through missing information where possible. It can also apply your policies along the way and escalate to someone on your HR team when a request calls for it.
- Memory and policies: Short-term memory is able to track the current conversation or workflow. Long-term memory can recognize patterns from past interactions. Both are designed to operate within your existing access controls and permissions.
AI resolution example, step by step
Back to the PTO request. Say an employee types something like, "I'd like to take next Friday off."
Deflected: A chatbot-only interaction would give them the org’s PTO policy. The employee still checks their balance, fills out the form, submits it, and updates their calendar. That’s a lot of manual steps.
Resolved: An agentic AI assistant, with the right integrations in place, could:
- Check the policy for the employee's role and location
- Confirm their PTO balance
- Submit the request and route it to the manager
- Add the time off to the employee's calendar
The employee skips the hunt for a form, and your HR team stays out of the queue for in-policy requests, all while the manager still makes the approval decision.
Where full resolution shows up: Core HR use cases
The highest-impact starting points are generally high-volume, rules-based workflows where steps follow clear policies. These are typically well suited for end-to-end automation because they can be defined by your rules and permissions.
Six use cases to look at first might be:
- Talent acquisition: Between internal calendars and external applications, recruiters lose hours to calendar juggling. Agentic AI is able to schedule interviews, remind panelists to submit feedback, and flag delays to hiring managers across tools like Greenhouse (an applicant tracking system) and Slack.
- Onboarding: New hires need access, orientation, and paperwork on their first day. Agentic AI can help grant tool access by role, schedule orientation, and share department-specific resources that the employee is permissioned to access.
- Performance management: Review deadlines can slip when managers chase updates. Agentic AI can send review reminders and track goals, which can help free managers to spend that time coaching.
- Learning and development: Training can stall when approvals and reminders pile up. Agentic AI is able to recommend courses tied to each role, route funding requests, and nudge employees before deadlines pass.
- Benefits enrollment: Open enrollment often brings a flood of questions and record changes. Agentic AI can answer personalized benefits questions, update HRIS (like Workday) and payroll records where integrations allow, and open a ticket when someone on your HR team needs to review a change.
- Compliance and policy management: Tracking policy acknowledgments by hand is slow. Agentic AI is able to log acknowledgments, send certification reminders, and record HR cases, so you have a clear record sooner.
Getting started: Implementing agentic AI in HR
Whichever platform you choose, automating HR processes with AI agents generally involves similar steps and priorities:
- Identify repetitive workflows. Start with high-volume, low-judgment tasks like onboarding checklists and PTO questions.
- Centralize knowledge and integrations. Bring policies, documents, approvals, and system connections into one governed layer, so the AI works from a single source of clean inputs.
- Build or customize HR agents. Use plain-language instructions to create agents for tasks like time-off submissions or policy lookups.
- Deploy where employees already work. Chat, intranet, email, and portals are all options. One conversational entry point can help keep things consistent.
- Measure and refine. Track your resolution rate alongside deflection, which tasks employees complete on their own, and where issues remain.
When you’re evaluating an HR AI tool, three questions can be useful to ask:
- Does it complete the task, or hand the employee a link?
- How often does an employee return for the same issue?
- What share of requests finish without manual rework?
Even anecdotal answers from other organizations can give you a sense of how well the platform may work for your org.
Choosing agentic AI built for resolution
Moveworks is designed around the premise that an HR request is resolved only when the underlying task is done.
Two parts of the platform support that goal:
- A Reasoning Engine that’s designed to interpret an employee's request, plan the steps, and coordinate actions across different systems.
- Tool Studio, a low-code builder, that lets you configure workflows with minimal programming. HR and IT teams can use this to create or customize AI agents for onboarding, benefits, and policy workflows.
Through the AI Assistant, employees can ask questions in everyday language, get answers with source citations, and complete tasks like PTO requests and benefit updates. Pre-built integrations with systems like Workday and ServiceNow are designed to connect those actions within your permissions.
Moveworks is now also part of ServiceNow. Its conversational AI helps power ServiceNow EmployeeWorks, which pairs it with ServiceNow's workflows.
Learn how Moveworks supports full resolution for HR, so your teams can move past the deflection trap.
Frequently Asked Questions
No. Many enterprise-grade agentic AI platforms come with pre-built connectors to common HR systems like Workday, ServiceNow, and Microsoft Teams, so HR teams can deploy without a lengthy custom development cycle. The amount of IT involvement typically depends on how many systems need to be connected and how much workflow customization is required beyond standard use cases.
Agentic AI is best suited to high-volume, rules-based tasks (like processing PTO or updating benefits), rather than judgment-heavy decisions like promotions or terminations. For sensitive HR matters, the technology is typically configured to support human decision-makers with information and workflow coordination, while keeping a human in the loop for final approval.
Agentic AI platforms built for the enterprise are designed to operate within existing access controls, permissions, and compliance frameworks, and to limit an agent’s actions and data access to what an employee's role already permits. Governance features like audit trails, configurable approval steps, and role-based access can help HR support compliance requirements while automating workflows.
Trust tends to build gradually, and it usually starts with agentic AI handling low-risk, easily verifiable tasks (like confirming a PTO balance or updating a mailing address) before expanding to more complex workflows. Transparency also matters here. When employees can see what action was taken and why, adoption and trust tend to increase faster than when the system operates as a black box.
For decisions that need human judgment, like complex leave exceptions or benefits disputes, the system can escalate to an HR representative, so employees know when a person is making the final call.
Ticket volume alone can be a misleading metric, since it doesn't distinguish between a resolved request and one that was simply deflected elsewhere. Stronger indicators include time-to-resolution (the percentage of requests completed without requiring manual rework or escalation) and downstream effects like new-hire time-to-productivity or manager time reclaimed from administrative follow-up.
