AI gave your team hours back. The default move is spending all of them on more output, and it is quietly costing you.

Key Takeaways

For CEOs:

  • 43% of workers name losing human-to-human interaction as their top concern about AI, ahead of job loss (Workday, May 2026)
  • Workplace loneliness costs US employers $154 billion a year, driven mostly by absenteeism. Lonely employees are 5x more likely to miss work and 3x more likely to underperform
  • AI is delivering real gains, burnout down and confidence up, while a connection deficit widens underneath them. Only one of those shows up on your dashboard
  • The default is reinvesting 100% of AI's recovered capacity into more output. Made by default rather than by decision, that is what widens the gap

For CMOs & CROs:

  • Oxford studied 46,336 workers across 233 organizations. Individual wellbeing offerings, mindfulness sessions, resilience training, stress management, wellbeing apps, produced no benefit. On several measures participants came out slightly worse
  • What did move wellbeing was organizational change: scheduling, management practice, workload, job design
  • 1 in 5 Gen Z workers took time off this year because they felt lonely or isolated
  • Your wellbeing budget is not a strategy. Handing someone an app while their calendar stays broken is paperwork, not support

The Reality Check:

  • A 12-month study across four countries found that lonely people turn to chatbots, and the more they did, the lonelier they became. The coping mechanism your workforce reaches for on its own runs in the wrong direction
  • AI did not create loneliness at work. It removed the friction that used to prevent it: the hallway question, the walk to the car, the three minutes before the meeting started
  • 63% of workers expect AI to make their workplace feel less human this year
  • Efficiency does not build trust. Friction did

The gap nobody priced in

Ask knowledge workers what worries them most about AI and you would expect one answer. My job.

That is not what they say.

In Workday's global research released in May 2026, the top concern was losing human-to-human interaction. Forty-three percent named it, ahead of job loss.

People are more afraid of being alone at work than of being replaced. If you run a company, that reorders your priorities.

The gains are real. So is the bill.

The same research found AI is helping. Burnout down. Confidence up. This is not a piece about how AI is bad for people, and I have no interest in the version of this argument that pretends the productivity gains are imaginary. They are not. I use these tools every day.

But underneath the gains, Workday found a connection deficit opening. One in five Gen Z workers reported taking time off this year because they felt lonely or isolated. A separate 2026 survey found 63% of workers expect AI to make their workplace feel less human.

Your people are getting faster and more capable. They are also getting further apart. Both are happening in your organization right now, and only one of them shows up on a dashboard.

The relief is making it worse

The strange part is where people go looking for relief.

A 12-month longitudinal study across four English-speaking countries, published this year in Psychological Science, tracked what happens when lonely people turn to AI for company. Loneliness drove them toward chatbots. The more they used them, the lonelier they became.

The thing that felt like a fix was deepening the problem. Which means the coping mechanism your workforce is reaching for on its own, with no prompting from you, is running in the wrong direction.

Why your wellbeing budget will not touch this

The reflex is a program. An app, a stipend, a resilience workshop, a mindfulness subscription.

There is now good evidence that reflex does not work. Oxford researchers examined 46,336 workers across 233 organizations, comparing people who used individual wellbeing offerings, mindfulness sessions, resilience training, stress management courses, wellbeing apps, against people who used none of them.

No benefit. On several measures, participants came out slightly worse.

The researchers' conclusion is the part most leaders skip. What did move wellbeing was organizational change. Scheduling. Management practice. Workload. Job design.

Handing someone a meditation subscription while their calendar stays broken is not support. It is paperwork. It lets the organization say it did something without changing anything.

"The

Cigna put the cost of workplace loneliness to US employers at $154 billion a year, driven mostly by absenteeism. Lonely employees are seven times more likely to disengage, five times more likely to be absent, and three times more likely to underperform.

Connection is not a soft benefit competing against hard priorities. It is a retention and productivity line item that stays invisible right up until people leave.

What AI changed

AI did not create loneliness at work. It removed the friction that used to prevent it.

The status meeting became a dashboard. The question you would have walked over to ask became a prompt. The handoff conversation became a ticket with a summary attached. Every one of those was the right call in isolation. Together they deleted nearly every unscheduled moment two colleagues used to have.

Connection at work was never built in the meetings. It was built in the three minutes before the meeting started, when somebody asked how your weekend was and you told them the truth.

Efficiency does not build trust. Friction did.

The reinvestment decision

AI hands you back capacity. That part is settled. The open question, and the one that will separate teams over the next two years, is what you do with it.

The default is to reinvest one hundred percent of the gain into more output. More shipped, more pipeline, more coverage. It is the obvious move, and it is why the connection deficit is widening even at companies where the AI rollout is going well.

The teams that hold together will be the ones that route some of the recovered capacity back into each other, deliberately, as a matter of operating design rather than good intentions.

I call the mechanism a happiness tripwire. A small good moment, placed in someone's path on purpose, before they ask, that anticipates a need. I built the idea to get through cancer treatment. It works the same way on a Tuesday full of meetings.

 

Design it into the system, not the individual

If the Oxford finding is right, and I think it is, then the answer cannot be one more thing you ask individuals to do on their own time. It has to change how the work runs.

Three changes that operate at the organizational level:

    1. Put unstructured time back on the calendar on purpose. Not an offsite. Not a social committee. Two minutes at the top of standing meetings that belong to the people rather than the agenda, protected by whoever runs the meeting.
    2. Make noticing part of the operating cadence. Build a named slot into your weekly review where specific work by specific people gets said out loud. When appreciation depends on someone remembering, it does not happen. When it is on the agenda, it does.
    3. Point your AI at connection, not just throughput. The same tools that track velocity can tell you who has gone quiet, whose one-on-one keeps getting moved, and which new hire has not been in a room with anyone senior in a month. Let the tool find the gap. Then have a human close it, in their own words.

    That third one carries a rule. AI plans and prompts. People bring the attention and the words. The moment the tool writes the note, it stops counting, and everyone can tell.

    The unasked and necessary question

    Your team is more productive than it was 18 months ago. Most teams are.

    Ask a harder one at your next staff meeting. Are they closer than they were 18 months ago, or just faster?

    You already know which of those your dashboard measures. The other one determines whether the people delivering those numbers are still here next year.

    Sources

    Most companies are spending everything AI gave back on more output. The smartest ones are reinvesting part of it in the connection that decides whether their people are still there to deliver it.