Share with your CHRO
Three CHROs from GSK Pharmaceuticals, BSE, and Sonata Software drew a hard line at the ETHRWorld Annual Conclave 2026: efficiency gains alone do not constitute proof that an AI investment in HR has delivered value. The emerging standard of proof covers decision quality, talent pipeline strength, risk reduction, and what happens to displaced workers. Their argument reframes the ROI conversation away from headcount eliminated toward whether the organisation actually works better after the technology went live.
What this means for your business
Most HR AI deployments right now are being evaluated on the wrong scoreboard. Time-to-hire, cost-per-screen, tickets-closed: these are activity metrics dressed up as outcomes. If your current AI business case rests on hours saved, you’re one board question away from a credibility problem, because “we saved 4,000 recruiting hours” invites the obvious follow-up of whether the hires performed better. CHROs who built their cases on efficiency alone are most exposed here; CHROs who tied deployment to a specific decision quality gap, like bias reduction in shortlisting or attrition prediction accuracy, are on firmer ground.
The BSE example from the panel carries a structural insight that applies well beyond financial services. When AI and a human reviewer catch different categories of error in document review, the combined system outperforms either one alone. That pattern, call it complementary error coverage, is the actual design principle worth institutionalising. It reframes the build decision from “how much human review can we remove” to “where do human and machine error profiles diverge most, and how do we stack them.” That’s a workflow redesign question, not a procurement question, and it belongs in the hands of the CHRO before the vendor gets the contract.
Balaji Kumar’s point about workforce redeployment is where this conversation gets expensive fast. A narrow ROI model that counts eliminated roles as savings will look attractive in year one and generate attrition, legal exposure, and capability gaps in years two and three. The CHROs arguing for a broader value stream aren’t being soft on AI; they’re pricing in a cost that narrow models leave off the spreadsheet. The falsification condition here is clear: if your organisation is tracking AI-driven role changes but not tracking where those people went and whether the organisation can still do what they used to do, your ROI number is incomplete by design.
Based on reporting from AI in HR is proven only when it improves outcomes, not just efficiency: HR Leaders, ETHRWorld, originally published 2026-09-23 03:00:00.
