The AI bubble is here and it's filled with failed experiments
What we're getting versus what we need.
A few months ago, I went to visit my primary care doctor. At the start of my appointment, she asked if she could record our conversation. I said, “Of course.” She took out her phone and launched an app.
During the visit, she was more attentive. Because she could focus on talking to me instead of typing notes on the computer. AI was running in the background, doing her documentation. I can’t imagine the number of hours doctors and other medical staff spend on notes, but I’m sure it’s a lot. AI lets them decrease their workloads and focus more on patient care.
That’s the type of AI technology we all need. Something that operates in the background and quietly makes work better. But that’s also not the type of AI that many companies are promising with their “life-changing” AI workflows that will replace humans entirely.
An AI company called Friend spent $1 million to advertise around New York, including in the subway tunnels. Friend is a wearble that users can talk to throughout the day.
New Yorkers did not take kindly to the ads for Friend, defacing the ads with graffiti lines like “Surveillance capitalism” and “Get real friends.”
Meanwhile, the world collectively assumes that we are on the brink of an AI bubble. McKinsey estimates that companies will need to invest $6.7 trillion by 2030 to keep up with the demand for AI compute — a staggering investment by any measure. But for all the money, it seems like companies are largely building and pushing the wrong solutions.
AI notetaking for meetings and appointments = helpful
AI “friend” that you can wear around your neck = not helpful
The AI we’re getting versus the AI we need
In the past year, CEOs have loved to drop announcements like “We’re freezing hiring due to AI” or “We’re restructuring the organization because of AI efficiency.”
Microsoft has cut thousands of jobs in 2025 alone. In a blog post, chairman Satya Nadella mentions that “AI transformation” is a business priority and that “getting both the product and platform right for the AI wave is our North Star.”
Collectively, companies have invested tens of billions of dollars in so-called AI transformation. Yet, according to research from MIT, 95% of companies have gotten zero return on investment in their AI pilot programs.
On one hand, leaders are acting like AI is about to make the entire company more efficient, so much so that they need to restructure and lay off people. On the other hand, they can’t get basic pilot programs off the ground.
I spent the first ten years of my professional career implementing enterprise software. I know exactly why pilot programs fail. Leadership teams buy products based on a promise, but implementation requires a lot of thinking and planning. Internally, employees who actually do the work have to be on board and understand the promise of an outcome (if implemented correctly).
The biggest failures happened when no one internally was driving the bus. In turn, the employees expected to use the product never received the proper training. Or there wasn’t an internal “champion” excited about change enough to get other people excited to learn. Instead, employees saw the software as an unwelcome disruption and distraction from their work.
With AI, tech companies are sprinting toward “transformative” experience. Workers are asking for practicality — and need the training and proper implementation to make it happen. Instead, they work under the threat that AI could replace them.
And also: “boring and genuinely helpful” doesn’t attract venture capital investments like “flashy and transformative,” so here we are, on the brink of AI bubble collapse.
The contradiction of money, capabilities, and human involvement
The fear of AI’s impact on jobs isn’t universal. According to a survey in Germany, only 5% of employees are worried that AI could threaten their jobs within the next five years. In the same survey, under 40% of employees had received any training related to AI.
No wonder AI experiments at companies fail. I know firsthand from many years of experience training thousands of people to use enterprise software that people often can’t just “figure it out” on their own. They need to be given specific guidelines. And that was for a product that was far more straightforward than AI use cases.
Researchers at MIT’s Sloan School of Management have tried to steer the conversation back to humans. They introduced the EPOCH framework: empathy, presence, opinion, creativity, and hope. These are the human capabilities that AI can’t replicate.
If companies are trying to introduce pilot programs to replicate (or replace) the human qualities introduced in the EPOCH framework, of course they will fail. Think: the companies that have tried to replace customer service agents with bots. That only succeeds in creating a doom loop when customers simply want to talk to a human.
I go back to the example of AI “running in the background” at my doctor’s appointment. I needed empathy, presence, and an opinion from my doctor. And while documentation is important, that’s not a part of her job that requires specific expertise.
If companies centered around EPOCH, the AI conversation wouldn’t be “replacement.” It would be augmentation. Humans are doing meaningful and/or creative work. AI is doing some of the necessary (albeit boring and repetitive) work underneath.
Some of the most transformative technologies in history didn’t replace humans. They replaced tedium. The calculator didn’t eliminate math. It eliminated the tedium of doing math by hand.
But right now, companies are still chasing spectacle over substance. It’s not that AI isn’t powerful. It’s that leaders want the benefits of AI without doing the hard, boring, systems-level work required to make AI actually useful. And when they get to those use cases — like an AI notetaker during medical appointments — does that justify the amount of investment in AI companies?
So yes, I’m excited about the AI that quietly makes life easier — the kind that helps my doctor be present, reduces tedious work, and helps teams spend less time on repetitive tasks. But I’m not sure that’s what CEOs are looking for when they talk about “transformational” change.
If you want to support my work as a writer, you can buy me a coffee or subscribe to receive additional issues I publish.
Have a work story you’d like to share? Please reach out using this form. I can retell your story while protecting your identity, share a guest post, or conduct an interview.