Please find the list of sources used for this post, at the end of the post.
Preface
When AI models started to become widespread and take their place in our world, many people were caught up in the hype without giving the actual products that surfaced a chance.
It seemed like a very fast-paced race where new limits were tested daily and people worked around the clock, fascinated, trying to find out what more could be achieved with this wonderful technology.
Many employers, both large and small and from a variety of sectors, started fantasizing about a world where AI would replace some or even all of their workforce.
People started claiming that many jobs would soon be replaced by AI and that humans would have to find new ways to earn a living.
For some reason, two aspects of AI use were totally ignored:
- The first was obvious: you need tokens to advance your work, and the more complex the work, the more tokens you need. Most people have already experienced this themselves.
- The other was less obvious, and people didn’t think to test it first. The quality of the work being done.
Without considering these two aspects, companies started firing employees, thinking they could replace them with AI models. However, they then found that the cost of tokens for certain missions could be more expensive than hiring a human for the job, resulting in a crisis due to the poor quality of work.
A few examples
Uber
Uber gave roughly 5,000 engineers access to Claude Code in December 2025. Usage nearly doubled by February 2026, and by March, 84% of engineers were classified as agentic coding users. By April, four months in, Uber’s CTO had burned through the company’s entire 2026 AI coding budget. Monthly API costs per engineer ran between $500 and $2,000.
Microsoft
Windows lead Pavan Davuluri publicly acknowledged the OS “went off track” with aggressive AI expansion while basic stability issues went unaddressed. Microsoft has since cancelled deeper Copilot integration in Windows 11 and is removing Copilot from apps like Notepad and Paint, shifting focus back to fixing the Start menu, Taskbar, and File Explorer instead of adding more AI.
On March 17, 2026, Microsoft restructured its Copilot and AI leadership, confirmed in an official blog post from Satya Nadella and AI chief Mustafa Suleyman. The commercial and consumer Copilot engineering teams, which “have yet to gain broad adoption,” were merged into one unified org. Suleyman was freed up to focus on building models instead of running the product, a clear signal about how the product side was performing.
Consumers lost access to Group Chats, AI-generated podcasts in Copilot, Copilot Labs experimental features, and Deep Research by August 18, 2026, described as Microsoft “killing off unsuccessful AI features.”
Microsoft cut 15,000 jobs (about 10% of its workforce) in April 2026, and a separate 4,800 jobs, while simultaneously spending heavily on AI infrastructure cutting workforce while pouring money into AI that hasn’t yet delivered proportional results.
Separately, Microsoft reportedly canceled most of its internal Claude Code licenses about six months after rolling them out, partly over cost, redirecting engineers to GitHub Copilot CLI instead.
Meta
According to a recent Reuters story, one of the boldest moves planned months ahead of its time and hidden from the public was a Meta (Facebook, Instagram, WhatsApp) project to replace 60% of its workforce with AI. Nicknamed Project OT, this plan was met with three levels of real resistance:
1. Employees became suspicious when they realized that all their work was now being used to train AI models. They started to guess where this was leading and rebelled.
2. AI agents failed to comprehend complex tasks, making more mistakes and requiring more “babysitting” from human engineers.
3. Investors noticed that Meta was spending large amounts of money on AI with no visible return (no efficiency or improvement was observed). According to Reuters, the project was then abandoned, and Meta moved away from it, at least for now.
In the freelance world
As a freelancer working in the tech industry, I speak with many different freelancers. Many people are still finding it hard to comprehend what can be achieved successfully with AI and what still requires a human touch.
A friend of mine who is a visual designer said that, during the early days of the AI hype, she noticed that she was getting less work and no new clients. People were sure that replacing her services with AI was not a big deal. After a while, however, she started to notice existing clients returning and new work coming in again. People realized that many tasks still required a ‘human touch’ that AI could not provide.
My partner also works in marketing as a freelancer.
The marketing world did not embrace AI with the certainty that other professional sectors did. It was never easy for people to distinguish between marketing and sales, and they often got the two mixed up.
Then the digital world came along and created one term to cover everything, without having this understanding from the outset. ‘Digital marketing’ has become a term that encompasses many aspects of a digital presence, and the distinction between marketing and sales has become even more blurred.
Amidst this confusion, AI was introduced. It did not matter that most people did not realize that marketing and sales were two distinct professional disciplines. Many machines were created using AI to solve various sales issues, but they never addressed the marketing side of things or identified what was missing from the process.
While many people created amazing solutions for their own and others’ digital presence, marketing was mostly overlooked while people stood fascinated by irrelevant numbers with no strategy. Both marketing and sales require high level strategy first which is then translated into tactics and tools that AI can serve.
Test before making plans
Many AI models help me in my everyday life and work. AI has certainly changed the way we do things, but, as with so many other inventions, we should take a closer look at the real value and actual ability of AI agents today, rather than making overly ambitious claims without considering the bigger picture of AI.
List of sources
Microsoft shakes up Copilot AI leadership team, freeing up Suleyman to build new models
Microsoft kills off unsuccessful AI features while merging its separate Copilot apps
You won: Microsoft is walking back Windows 11’s AI overload
Microsoft to cut 4,800 jobs, joining the wave of AI-driven tech layoffs
How Meta’s AI layoff plans went kaput
Bosses Are Blowing More Money on AI Agents Than It’d Cost Them to Just Pay Human Workers
AI can cost more than human workers now
KPMG Says Nearly Half Of Executives Pulled Back AI Agents Over Cost
Uber, Microsoft, and Others Burning Through AI Budgets. Now What?