Workday is betting that agentic AI, software that acts autonomously across multi-step processes rather than just answering questions, becomes the operating layer of HR itself. Michelle Dawkins, VP of Solution Consulting, cites a 25% reduction in HR transaction volume for customers using Workday’s self-service agent, and 900 hours of annual savings on audit work via its financial audit agent. The strategic claim underneath those numbers is that HR’s new job is architect of work<\/a>, not administrator of it, with AI handling roughly 90% of end-to-end process orchestration in finance and HR operations.<\/p>\n The 25% transaction reduction figure is the number your CFO will ask about, but the more consequential question it raises is structural. If agents absorb that volume, the HR function either shrinks its headcount or redirects it. CHROs who haven’t yet mapped which roles exist primarily to move information between systems are flying blind. Organizations with heavy point-solution fragmentation, multiple HR tools that don’t talk to each other, are most exposed: their staff are already acting as human middleware, and that’s precisely the work agents will absorb first.<\/p>\n The responsible AI framing Dawkins offers is more operationally grounded than the usual ethics boilerplate. Her three-question governance test (culture, human-interaction design, skills reinvestment) functions as a policy architecture, not a values statement. The sharp edge of it is the third question: do employees actually have the capacity to do the higher-order work that AI is supposed to free them for? Most organizations deploying automation skip this entirely, treating displaced time as automatically productive. It isn’t. The skill gap between “no longer processing forms” and “redesigning job architecture” is real, and Workday’s Skills Cloud, which infers competencies from CVs, training histories, and work activity rather than relying on HR-defined taxonomies alone, is the product designed to make that gap visible before it becomes a retention problem.<\/p>\n Dawkins’ sharpest line is the dismissal of “sticky plaster” transformation, applying new tools onto a broken architecture rather than consolidating it. This is a direct pitch for platform consolidation over point solutions, and Workday has an obvious commercial interest in that argument, which should make you weigh the specific thresholds she implies carefully. But the underlying logic holds regardless of vendor. If the same data produces different answers in three different systems, agents trained on that data will produce unreliable outputs. That’s not a vendor problem; it’s a data governance problem that surfaces as an AI reliability problem. CHROs should audit that fragmentation now, because the renewal cycle on most HR point solutions is the natural decision point, and it’s arriving faster than most roadmaps assume.<\/p>\n Agentic AI refers to systems that don’t just respond to a single prompt but execute sequences of actions autonomously, the way a junior analyst would gather data, draft a report, flag anomalies, and route the output, without being asked for each step. In HR, this means an agent handling the full lifecycle of an employee request, not just retrieving a policy document. The business stakes are high because agentic systems inherit whatever is broken in the underlying data and process architecture, and they act on it at scale.<\/p>\nWhat this means for your business<\/h2>\n
Concept deep-dive: Agentic AI<\/h2>\n