The previous units gave you the diagnostic and design foundations — how to read what's actually behind low adoption, how to map the mess before automating it, and how to bring the right people into the redesign. This unit addresses what happens when someone at the top skips all of that and drops a mandate in your lap. The panel tackled this head-on with a scenario many in the room recognized: a CEO who returns from a board meeting ready to cut 50% of headcount using AI as the justification. What followed was the panel's clearest guidance on how HR leaders can push back without losing credibility — by reframing the conversation, insisting on transparency, and refusing to treat AI outputs as gospel.
When the CEO mandate scenario landed, one panelist's response was immediate: "can we play for some upside here?" The argument wasn't that cuts never happen — it was that starting with cuts poisons everything downstream, incentivizing "people locking stuff down, locking information down, not sharing, job protection, not collaborating openly." Another panelist shared a reframe that stuck: a workshop leader who opened by asking "what would you do if you had 30% more capacity or resources?" — because if a function can articulate what it would do with that capacity, you've made the case for expansion before anyone reaches for the headcount spreadsheet. Lead with opportunity, and you unlock the collaboration that makes AI adoption actually work.
Building on this, the panel was equally direct about transparency when AI enters candidate- or employee-facing interactions. One panelist described giving candidates "the option to do the recruiter screen with either a human recruiter or an agentic recruiter," noting that roughly 35% choose the agentic path. The principle isn't about predicting preference — it's that offering genuine choice is what builds the credibility to sustain transformation. When employees and candidates feel they have agency in how they interact with AI, they become participants in the transformation rather than subjects of it. As another panelist put it,
