Workday's new Sana Concierge is intended to resolve employee questions about things like leave and payroll before they become service tickets. My first question would be: what data is it using to answer them?
Most organisations have an HRIS. That doesn't mean they have the structured, current data needed to answer an employee's actual question.
Take a pay or leave query. The answer might depend on the employee's job, location, employment conditions, roster history, an enterprise agreement, a payroll rule or a policy. Some of that will be structured. Some will be buried in documents or maintained outside the HRIS. We see this regularly in HRIS and payroll work.
Before letting an AI service answer employees directly, we need to know which source it uses for each type of information, which source wins when they disagree, and what it does when the information isn't good enough to answer.
Otherwise, the organisation hasn't really automated the answer. It has automated whatever data happens to be available.