AI in HR

What these leaders agree on

Across eight conversations, the most consistent finding is that AI programmes fail for organizational reasons rather than technical ones. Alexis Fink, who has led people analytics at Meta, Microsoft and Intel, puts the failure points at integration, process redesign, change management and absent executive sponsorship. Jennifer Love makes the same argument from the governance side: organizations execute before establishing enterprise strategy, governance and data readiness. Ian O'Keefe adds the data foundation, warning that new technology layered on a broken foundation makes things worse, because AI will be confidently and eloquently wrong. Jessica Smith reduces it to a sequence: strategy first, tools second.

The second point of agreement is that managers, not executives or vendors, decide whether adoption happens. Symphone'e Lindsey calls them the most under-leveraged asset in any AI change programme because they sit where organizational strategy meets lived experience. Jennifer Love reaches the same conclusion from a different angle, arguing managers set the psychological safety that makes a learning curve survivable.

The third is that fear is the hidden tax. Fink draws on loss-aversion research to argue leaders consistently underestimate employee anxiety, because humans overestimate loss when work is at risk. Steve Cadigan describes the consequence: introduce fear and everyone starts playing career defence rather than offence. Nadia Vatalidis puts psychological safety ahead of reskilling entirely, on the grounds that people who do not feel safe admitting uncertainty will comply with adoption rather than genuinely embrace it.

Where they disagree

The sharpest split is over what AI actually does to a role. Jackson Lynch separates the two things most organizations conflate: tasks get automated, roles get redesigned, and treating one as the other demoralizes people. He frames AI as a mirror that surfaces redundancy and approval layers, and treats that as a design invitation rather than a headcount justification. Fink is more willing to concede ground, arguing the useful question is which roles are being augmented versus replaced, and that repetitive, well-described, lower-risk work genuinely is replaceable. Cadigan rejects the subtraction framing altogether, calling the use of the most capable technology we have ever had to remove people a strategic error rather than an efficiency.

They also differ on how to treat resistance. Lindsey argues employee scepticism about AI is an asset rather than a problem, because sceptical employees are the ones paying attention, and that treating their concerns as resistance shuts down the signal you most need. That sits awkwardly beside the more common framing of adoption as something to be driven through an organization.