HR's AI Problem Is Actually an Automation Problem
Gartner surveyed HR leaders last fall. Eighty-eight percent said their organizations had not realized significant business value from AI tools. (1) The analysis pointed to governance gaps, misaligned expectations, deployment decisions made without HR involvement.
All true. But there’s a more specific problem underneath those explanations.
HR teams are standing up automations and calling them AI initiatives. I don’t believe those are the same thing, but only because people tend to conflate three things.
Automation, Chat AI, & Agentic AI.
Those are not the same. And calling them the same thing could be costing you.
Above the line, below the line.
Automation is a doer and has been around for a long time. It takes work you shouldn’t have to do and removes it. A rule fires, a task completes, a form routes itself. This is real value. It frees time, reduces error, and returns capacity.
It’s below the line support, clearing what bogs you down without changing what you’re capable of thinking or deciding.
Chat and agentic AI are something else. Used well, they can be a reasoning partner. They process context, surface patterns, and push back on assumptions. They don’t just execute what you tell them to do, but instead can help you question whether what you’re doing makes sense. That’s above the line support. They change what’s possible, not just what’s faster.
HR teams have a pattern that predates AI by decades. When a process is broken, the instinct is to build around it. A form that takes too long gets a shortcut. A reporting structure that creates confusion gets a workaround meeting. The underlying problem stays. The workaround becomes permanent.
Automation is the best tool ever built for scaling that instinct from what I’ve seen. It makes workarounds fast, reliable, and invisible. The broken process is still there. Nobody has to think about it anymore.
But that’s the real danger of calling automation either chat or agentic AI. Not a labeling error. The wrong intervention at the wrong level. You needed something that would ask whether the process should exist. You got something that executes it without complaint, forever, at scale.
Gartner found only one in fifty AI investments delivers transformational value. (2) That number makes more sense when you understand what most organizations are actually investing in.
This isn’t an argument against automation. Automation is amazing when it’s done correctly and maintained. The work it removes is real, and the time it returns is real. The problem is when it substitutes for the harder question.
The harder question is what AI as a reasoning partner forces you to sit with. Does this process deserve to be faster, or does it deserve to be rethought? Those are different questions. One is an efficiency question. The other is a design question.
The next time your team announces an AI initiative, ask what it’s actually doing. If it’s routing, triggering, completing, or removing, that’s automation. Probably good automation. But not chat or agentic AI, and not a substitute for the question you were supposed to ask first.