Employees ask HR the same questions all year: leave balances, payslips, benefits, policies, how to update details. An AI HR service desk answers them instantly and hands the hard cases to people.
How it works
- The assistant is connected to your policies, knowledge base and HR systems.
- It answers routine questions and completes simple transactions (for example updating an address, requesting a letter).
- It escalates anything sensitive or complex to a human advisor, with the context attached.
Design choices that matter
- Tiering: decide what the AI handles alone, what it prepares for a human, and what it never touches (grievances, harassment, health, performance disputes, terminations).
- Source of truth: answers must come from your current policies, with a link, not from the model's general knowledge.
- Country rules: the same question has different answers by country or contract.
- Easy exit to a human: a visible "talk to a person" option at every step.
- Logging and review: sample conversations weekly; fix wrong answers in the knowledge base.
Metrics
Containment rate (questions answered without a human), accuracy of sampled answers, time to resolution, employee satisfaction, escalation quality, and the HR advisor hours moved to higher-value work.
A real example
IBM's AskHR is one of the best documented. IBM reports a 94% containment rate for common questions, 11.5 million employee interactions in 2024, a 75% reduction in support tickets since 2016 and a 40% reduction in the HR team's operational costs over four years. It runs a two-tier model: AI handles routine inquiries while human advisors manage more complex needs.