Pilot or Purgatory? Hard Calls on HR AI Projects

Dr. Denise Turley

AI Advisor

Lilia Hayrapetyan

Managing Director, Global Head of People Analytics and Insights · Citi

Ketaki Sodhi

Head of AI Enablement · Moody's Corporation

Arjun Asokakumar

Pilots are everywhere, but the real risks are capacity drain, credibility erosion, and opportunity cost. Four leaders share how they gate ideas, instrument for impact beyond usage, involve Legal and Finance early, and design adoption that lasts. Leave with a practical playbook and time-boxed decision cadence to scale, pivot, or stop responsibly.

• Pilot triage playbook – capacity-first gate with owner, baseline, 90-day metric, privacy and bias checks, and a viable funding path.
• Impact instrumentation grid – measure time-to-insight, rework and error reduction, policy adherence, adoption depth, and diffusion across teams.
• Governance handshake – who does what in the first 30 days and the artefacts that are table-stakes.
• Adoption patterns that stick – agentic assistants, prompt libraries, role-based access, workflow hooks, and visible logging.
• 90- or 180-day decision path – monthly huddles to invest, pivot, or stop based on credibility, capacity, and opportunity cost.

• Apply a pragmatic capacity-first gate to green-light only pilots with clear owners, baselines, and 90-day success metrics.
• Evidence value beyond usage by tracking faster insight, fewer reworks and errors, stronger compliance, and deeper adoption.
• Secure early alignment with Finance, Legal, Privacy, and IT using concise, auditable artefacts.
• Design enablement that builds trust quickly through prompts, assistants, access controls, and workflow integration.
• Run a disciplined 90- or 180-day cadence to invest, pivot, or stop while protecting credibility and capacity.

Executives expect visible AI progress without compliance surprises. HR and analytics teams must prove value that Finance believes, meet privacy and bias controls, and avoid zombie pilots that consume scarce capacity. Clear selection gates, compact metrics, and auditable governance make AI progress safer, faster, and more credible in regulated environments.