McKinsey: leadership drives genAI success

Healthcare genAI has moved past the “should we try this?” phase. The real fight now is much less glamorous — getting useful models into messy, regulated, high-stakes workflows without breaking trust. That is the point sitting underneath McKinsey’s latest healthcare research and its broader 2026 organization data. The headline is simple: leadership matters a lot, but not in the vague inspirational sense. It matters because someone has to force the hard organizational choices that turn demos into operating systems. (mckinsey.com) ### What actually changed? McKinsey’s recent healthcare readout shows the sector is still pushing deeper into generative AI, but the center of gravity has shifted from experimentation to implementation. Healthcare organizations are using genAI for things like administrative workflows, stakeholder engagement, and other targeted processes, while still wres(mckinsey.com)into workflows to get material enterprise value. (mckinsey.com) ### Why is healthcare the hard mode? Healthcare has the usual enterprise AI problems, then adds extra landmines — patient privacy, clinical risk, fragmented data, and a much lower tolerance for hallucinations or biased outputs. McKinsey’s healthcare work frames adoption around both ROI and risk, with organizations trying to capture productivity gains while managing security, governance, and trust. That is why “just launch a chatbot” is not a serious strategy here. (mckinsey.com) ### So where does leadership come in? McKinsey’s State of Organizations 2026 report makes this unusually concrete. When executives were asked what drives AI adoption, 42% pointed to ease of use, while 36% named leadership championing adoption and another 36% named a dedicated team to drive it. That tells you something important — adoption is not only a product problem. It is also an ownership problem. (mckinsey.com) ### Why isn’t a good tool enough? Because workflow change is political as much as technical. A model can work perfectly in a pilot and still die if nobody decides who owns the budget, who signs off on risk, which team changes its process, and how success gets measured. McKinsey’s broader AI work keeps returning to the same blockage: organizations experiment widely, but scaling stalls when the operating model does not change with the technology. (mckinsey.com) ### What does “embedded in workflow” really mean? Basically, genAI has to disappear into the job. In healthcare that could mean drafting prior-authorization materials inside an existing process, helping clinicians with documentation in the systems they already use, or supporting call-center and payer operations where latency, auditability, and handoff rules are clear. If workers have to leave th(mckinsey.com)That is the difference between a demo and infrastructure. (mckinsey.com) ### Why do pilots keep stalling? Because pilots are cheap, but integration is expensive. The hard part is not proving that a model can answer a question. The hard part is connecting that model to real data, permissions, review loops, compliance controls, and frontline incentives. McKinsey’s surveys suggest many organizations are still stuck in that middle zone — lots of activity, uneven scale, and benefits that stay local instead of becoming enterprise-wide. (mckinsey.com) ### What should executives take from this? If you are pitching an AI-enabled data platform, the argument is not “AI is important.” Everybody already knows that. The argument is that leadership has to sponsor staged deployment, name a single owner, fund the integration work, and set governance before rollout. Otherwise the organization buys intelligence but keeps the same plumbing — and then wonders why nothing compounds. (mckinsey.com) ### Bottom line? The McKinsey read is not that healthcare needs more genAI enthusiasm. It needs more executive commitment to workflow redesign. That is the lever that makes the technology real. (mckinsey.com)