AI dashboards can’t replace leaders who truly see

AI dashboards – A CEO’s AI-powered people analytics platform delivered “real-time” insights—yet within six months two top managers quit without warning. The exits came with staggering replacement costs, a cautionary case against treating measurable HR signals as a substitute

A CEO once told me his HR transformation was working.

He had rolled out an AI-powered people analytics platform a few months earlier. and he sounded proud when he described what it could do: real-time sentiment data. predictive turnover scores. and engagement dashboards. Then he delivered the outcome in a single line—his HR team had been cut by a third.

Six months later, two of his highest-performing senior managers quit in the same quarter.

There were no obvious warning signs—no flags, no warning scores, nothing on the dashboard. What happened instead was quieter and more human. Two people had felt for a long time that no one knew them. Then, as the decision set in, they stopped waiting for that to change.

The price landed in the company’s balance sheet. One manager was a team lead carrying $4 million in client relationships. The other had spent two years grooming junior talent. Between severance. recruiting. onboarding. and the business that walked out with them. the company spent close to $600. 000 replacing people the dashboard had said were fine.

The system wasn’t wrong about what it measured. It simply couldn’t measure what mattered most.

Allison Pugh has spent years studying the professions we tend to trust with our most human moments—physicians. teachers. chaplains. therapists. She argues that in an AI-saturated future. the only irreplaceable work humans will do is relational: empathy. attunement. and genuine presence. In her book The Last Human Job, she calls this work “connective labor.”.

The definition is disarmingly simple: connective labor is the work of truly seeing another person.

It shows up in the check-in that surfaces a struggling employee before they spiral into a quiet quit. It’s the honest conversation that defuses conflict before it splits a team. It’s the leader who notices—without a dashboard or a survey prompt—that something is off with someone she has known for three years. Connective labor is invisible, relational, and load-bearing.

Pugh’s research has focused on professions associated with that kind of care, not middle management. But the pressure she describes is felt intensely where organizations often least protect it: inside people leadership roles. where connective labor quietly holds teams together—and where it is now being squeezed.

What AI is doing to your people leaders is not subtle. The transactional scaffolding of managing is being automated—scheduling, reporting, coordination, performance tracking. What remains after the measurable work gets stripped away is connective labor: the hardest part. the most invisible part. and the part almost no organization has learned to name. measure. or invest in.

The term “people leader” is deliberate here, because “manager” undersells what the role requires. People leaders create the conditions that make work sustainable—safety, belonging, challenge, and connection to something that matters. These aren’t just intentions. When they are absent, performance erodes, turnover climbs, and conflict festers until it doesn’t.

This kind of work cannot be automated in the way dashboards can be. Seeing a person requires being seen in return—a two-way act. A sentiment algorithm can detect that something is wrong. It can’t sit across the table and help someone figure out what it is. what they need. and whether there’s still a reason to stay. AI can’t inspire growth when growth is hard. It can’t disagree kindly in a way that forces honesty. And AI doesn’t care about a person’s specific way of thriving—though it can be very good at agreeing with us.

Connective labor keeps getting devalued because it’s hard to see. It doesn’t produce a neat deliverable. When a people leader spends 40 minutes on an unplanned conversation with someone who’s struggling. nothing gets logged for the effort. When she notices someone has gone quiet in meetings and makes a point of checking in. the outcome—trust rebuilt. a problem caught early. a person who feels less alone—doesn’t arrive as a line item.

So organizations stop investing in it.

They cut the slack that made it possible, stack more reporting, cascade more tools, and shrink people leaders’ real attention while their official headcount expands.

In the United States, the timing has been visible. The U.S. Surgeon General’s 2022 advisory named loneliness a public health crisis and pointed directly to the workplace as one of the primary places adults seek connection. That report landed a few years ago, and many organizations responded with another survey. Then they wonder why the best people keep leaving.

The problem, in this telling, is rarely an engagement problem. It’s a connective labor deficit—an HR strategy that automates around what holds people together without giving connective labor the protection and recognition it needs.

The cost is not abstract.

A manufacturing client described itself as proud of a lean management structure: each people leader carried about eighteen direct reports; there was a robust HR system; weekly automated check-in prompts; and a quarterly pulse survey.

But time was missing. People leaders were spending roughly twelve minutes per week per person. The relationships were cordial—functional, shallow. When the team was interviewed, one word kept coming up in different forms: invisible, replaceable, fine.

“Fine” becomes dangerous because it’s a kind of blind spot. Fine means nothing is on fire. Fine also means nothing is being tended to.

When two competitors began recruiting. the company lost seven people in eight weeks—smart. experienced. high-performing employees who decided that feeling genuinely known somewhere else was worth a lateral move. Their people leaders weren’t failing. They were structurally prevented from doing the work that might have kept those people. Connective labor had been engineered out of the system. Nobody meant for it to happen. It happened anyway.

This isn’t an argument against AI. The aim is not nostalgia for inefficiency. The question is what automation frees up human capacity for.

If the hours saved by automation are absorbed by more reporting and more throughput, the job hasn’t truly changed. The transactional parts of managing get faster while the relational parts remain starved.

Organizations that are “fit-for-human” do three things differently. They name connective labor explicitly and treat it as a business-critical function. They protect the conditions for it—people leaders need unscheduled margin. the kind of space that lets a real conversation happen. And they make it visible in how they evaluate and develop people. If connective labor never shows up in how people are recognized for doing it well. the organization is selecting against it. And “you will get what you measure.”.

Pugh’s research, in this view, was never only about physicians and chaplains. It was about what it costs to both parties when someone is seen versus processed. AI will not save a culture. The people leader who knows her people will.

So the real test is simple: make sure that leader still has time to do the work that only she can do.

AI people analytics employee engagement predictive turnover HR technology connective labor people leaders leadership accountability workplace loneliness turnover costs talent retention