Microsoft’s Copilot+ initiative aimed to introduce AI-powered laptops but struggled to gain traction, accounting for less than 10 percent of systems shipped in the third quarter of 2024, according to Mercury Research. Despite initial enthusiasm, features such as the Recall function raised privacy concerns and failed to attract consumer interest, leading to disappointing sales figures. Microsoft has shifted its focus, now proposing to make every Windows 11 computer an AI PC with cloud-powered features, rather than relying solely on high-end hardware. While the company claims that the Copilot+ systems had some success during the last holiday season, ongoing research from Omdia projects AI PCs will dominate the market, potentially reaching 75 percent of all systems shipped by 2029.
Recent concerns have emerged regarding the quality of artificial intelligence research, highlighted by a single individual, Kevin Zhu, claiming authorship of 113 academic papers this year, with 89 set to be presented at a leading AI conference. According to Hany Farid, a professor at the University of California, Berkeley, Zhu’s work is described as a “disaster,” raising alarms about the proliferation of low-quality research in AI, which is currently facing a surge in submissions — with the NeurIPS conference receiving over 21,500 papers this year, a significant increase from fewer than 10,000 in 2020. This has resulted in a pressure-cooker environment where academics are incentivized to publish quantity over quality, leading to concerns about the integrity of the field and the ability to produce meaningful contributions.
The demand for artificial intelligence in the workplace is reportedly declining, with only 11 percent of Americans at large companies utilizing AI to produce goods and services as of October 2025. This figure marks a decrease from 12 percent just two weeks prior, according to a recent survey by the U.S. Census Bureau. The trend is particularly alarming among businesses with 100 to 249 employees, where the percentage of those not using AI has risen to 81.4 percent. Furthermore, a Stanford economist noted a significant drop in generative AI usage from 46 percent in June to 37 percent by September. The tech industry anticipates spending $5 trillion on AI infrastructure by 2030.
A recent report from OpenAI reveals that while artificial intelligence is increasingly adopted in workplaces, it only saves employees an average of 40 to 60 minutes per day. The “State of Enterprise AI” report, which analyzed data from over one million business customers and surveyed nearly 9,000 workers, indicates that 75% of respondents believe AI has enhanced their work quality or speed. However, the productivity benefits are modest, with many employees feeling that the time saved does not significantly alter their workday dynamics. Notably, heavy users of AI report savings of over 10 hours a week, suggesting a widening gap in benefits based on usage intensity.
A recent study by the Future of Life Institute assessed the safety measures of eight leading artificial intelligence developers, revealing concerning findings. Notably, the top performers—Anthropic, Google DeepMind, and OpenAI—only achieved grades of C+ and C, indicating that even the best companies lack sufficient safeguards to manage potential risks associated with advanced AI systems. The report emphasized that all companies need to implement concrete safety protocols, as the current lack of regulation in the industry raises alarms about the existential risks posed by AI technologies.
Why do we care?
This is the side of the AI story vendors don’t love to talk about. Not everything is going well. Microsoft’s AI PC push fizzled, workplace adoption is actually dropping, and the productivity gains—on average—are modest. Meanwhile, the research ecosystem is flooding itself with low-quality papers, and even the safest AI labs are getting C-level grades. That’s not exactly the foundation of a revolution.
What this really tells us is that AI is hitting the reality wall. Users aren’t buying hardware just because it has “AI” stamped on it. Workers aren’t sticking with tools that don’t fit the way they operate. And customers aren’t seeing transformative returns unless they’re very intentional, very heavy users.
The mismatch here is striking: trillions in infrastructure spending on one side, shrinking day-to-day usage on the other. That’s a warning sign for MSPs. The hype cycle is cooling off, and the expectations are getting sharper. If you oversell AI, you’re going to own the disappointment. And remember: when vendors overshoot on AI investment and the returns lag, they push that pressure downstream — in pricing, product shifts, and expectations MSPs will be the ones to deliver the value they promised.
This is the moment to own the narrative with clients. AI is useful—it saves time, improves quality, and opens new workflows—but it’s not magic. It’s incremental. It demands governance. And it only pays off when it’s integrated into real processes, not stapled onto them.
For MSPs, the opportunity is to be the calm voice in a noisy market: focus on value, not vibes. Help customers adopt AI where it works, ignore it where it doesn’t, and build the foundations that make the real gains possible. The hype may be loud, but the work is still the work.