Beyond the hype: what a real human–AI operating model looks like

Bolting AI onto yesterday's workflow isn't transformation. It's wallpaper. Richard Doherty of Publicis Sapient and Richard Preece of DA Resilience join PIMFA’s Philip Allen to argue the real question isn't adopting AI, it's redesigning who does what when humans and machines share the outcome. Their lens is simple. Repeatable and data-heavy? AI leads. High-risk and trust-critical? Human leads. Everything else is human-in-the-loop. But accountability still sits with the firm and named individuals under SMCR, so guardrails, explainability and redress are the price of entry. Governance has to span design-time, runtime, and post-decision. And ROI needs to move past revenue, cost and risk to what actually matters: decision quality, cycle time, and capacity unlock, tracked in 30/60/90-day increments and delivered through "steel thread" implementations, one slice at a time. Race with the machines. Not against them.