When ChatGPT Beats Your Colleagues as a Teammate

False positives on AI gains, the real drivers of adoption, and the risk of profound losses of human connection

Jul 16, 2025

I've been neck-deep in newly-released AI research lately, and while the differences between studies can be stark, many of the lessons are increasingly clear – and sometimes alarming.

For example: the employees getting the biggest productivity gains from AI are also 2X more likely to quit. It’s not their marketable talents drawing them away — the most productive AI users have 88% higher burnout rates and are reporting they'd rather chat with ChatGPT than their colleagues. The demands to crank up outputs are driving down human connections.

If that doesn't make you pause your next "AI will revolutionize everything" presentation, I don't know what will.

When Perception Meets Reality

Five major studies dropped recently – BCG, Gallup, METR, Slack, and Upwork – painting wildly different pictures of AI's workplace impact. It's like asking four witnesses to describe the same car accident and getting reports of a bicycle, a pedestrian, two cars, and a dog.

The productivity data looks similar across studies: 40% gains in Upwork's survey among top AI adopters, with BCG finding that 55% of participants in workflow-redesigned organizations save at least an hour daily.

But perception isn’t reality: a new academic study found AI made developers 19% slower, even though they self-reported working 20% faster. They spent less time coding but way more time prompting, reviewing, and … waiting. As Kelly Monahan put it, "We're not great at noticing how our work shifts. We feel productive because we're busy, but the shape of that busyness is changing fast."

The top AI performers in Upwork's survey already know this: they're 32% more likely to say they have no idea how their company will achieve its AI goals. People with their hands deep in AI tools daily see the biggest gaps between leadership perceptions (it's nothing; it's revolutionary) and reality.

Organizations treating AI like a magic productivity pill, expecting people to just "do more, faster" without changing how work works, won't get the results they expect.

The Investment Gap That's Killing Results

Slack's data shows daily AI users are 73% more likely to be "highly satisfied" at work (38% vs 21%), with better work-life balance, stress management, and team belonging. When you look at BCG's data on what drives higher usage, the picture gets clearer: they're working in organizations that actually support them. It’s not the tools, it’s the leadership.

The numbers are brutal: only 36% of workers feel they’ve gotten sufficient training. As a result, only 18% of workers with no training support use AI regularly. Provide five or more hours of proper training, and adoption jumps to 82%. That's the difference between failure and success.

Leadership support shows the same pattern. Active executive engagement drives 82% adoption rates. Hands-off leadership? It drops to 41%. People aren't learning AI because leaders aren't teaching it, and they're not adopting it because leaders aren't modeling it. Sadly, only 25% of employees have actively engaged leaders.

This isn't rocket science, but it requires time investment by leaders – who are often the most burned out people in organizations. Companies have de-layered management, creating 1.6X more direct reports on average for managers. No wonder 59% of mid-career employees report significant burnout, compared to 45% overall.

More load, fewer hours for learning and coaching – right when those leadership layers matter most. You can't expect AI to enhance management if people aren't equipped to adopt AI. Seems obvious, right?

The Shadow AI Problem

If you don't provide tools and guidelines, 54% of your employees will bring their own. That jumps to 62% for Millennials and Gen Z.

This isn't rebellion; it's adaptation. When organizations fail to provide adequate tools and training, employees find workarounds. The question is whether you want to shape those workarounds or discover them during your next security audit.

Think Gen Z is driving this? Think again. Slack found that 33% of Millennials use AI daily – more than Gen Z's 28%. The generation that survived dial-up internet and learned to troubleshoot computers is better at adopting workplace AI than the TikTok natives.

Millennials are in their career sweet spot: experienced enough to know good work, hungry enough to embrace edge-giving tools. They're juggling mortgages, kids, and career advancement. If AI can help them write better proposals while their toddler melts down, they're all in.

When Bots Beat Humans

The scariest finding? High AI adopters are developing better relationships with their AI teammates than their human ones. They're more polite to AI, AI is more polite to them—and they've anthropomorphized relationships with AI that work better than with teammates.

Heavy AI users are emotionally disconnecting from human colleagues. This ties back to productivity demands: when the focus is more output, time for connection and relationship development disappears.

As Kelly Monahan told me, "AI's benefits are real, but they're only sustainable if we redesign work systems to support human connection, purpose, and growth."

The difference isn't the AI – it's everything else at work.

The Capability Expansion Revolution

Slack extended the term "vibecoding" to include any way of using AI to perform work outside core expertise. I prefer "capability expansion." Marketing people suddenly do basic data analysis. HR professionals create decent graphics. Product managers build prototypes. I'm an old dog suddenly doing new tricks.

We're witnessing the birth of Swiss Army knife employees, happening organically among people who understand both expertise value and "good enough" power.

These new skills will be crucial as organizations redesign themselves over the coming years. While Gallup shows job displacement fears steady at 15%, BCG reports 41% of workers globally believe their jobs will "certainly or probably disappear entirely" in the next decade.

Teams can become more self-sufficient. Projects move faster without waiting for specialized resources. Individual contributors solve problems that previously required expensive consultants.

But this only works when people feel safe experimenting. Organizations treating AI adoption like a performance metric create anxiety. Organizations treating it like professional development create capability.

Thanks for reading Work Forward! This post is public so feel free to share it.

What Leaders Must Do Differently

Invest in real training, not awareness sessions. Five hours isn't nice-to-have: it's the minimum effective dose. Handing out tools and asking people to figure them out on magic time doesn't work.

Get your hands dirty. Successful AI implementations happen when executives participate, not just mandate. You can't lead what you don't understand, and you can't understand AI's impact from 30,000 feet.

Focus on capability expansion, not just efficiency. Instead of asking "How can AI make you faster?" ask "What could you accomplish with AI that you can't do today?" This shifts conversations from replacement anxiety to growth opportunity.

Create clear permissions and boundaries. People need to know what's acceptable. Ambiguity breeds anxiety and shadow usage. Only 22% of organizations have clear plans and guidelines—that's unacceptable.

Create room for connection. People are missing time for human connection. This isn't about office days (though time together matters) – it's about putting limits on the non-stop efficiency drive to build relationships and trust.

The Choice Ahead

The research tells us something critical: AI adoption without intentional leadership creates productivity gains at the cost of human connection, employee satisfaction, and retention. The technology isn't the problem – our implementation approach is.

Organizations that figure this out first will have massive advantages. Not just because they'll be more productive, but because they'll develop workforces capable of things that seemed impossible two years ago. In a world where AI handles routine work, having people who can expand beyond traditional capabilities might be the only sustainable competitive advantage left.

The question isn't whether AI will change how work gets done. It's whether you'll shape that change thoughtfully or let it reshape your organization by default. Based on the research, most companies are choosing the latter.

Don't be most companies.

Work Forward