A recent survey by Gartner, Inc. indicates that by 2030, no IT work will be done without artificial intelligence, with 75% of tasks performed by humans augmented with AI and 25% by AI alone. The survey, which included over 700 Chief Information Officers, emphasizes the need for organizations to balance both AI readiness and human readiness to derive value from these technologies. During the Gartner IT Symposium/Xpo, analysts underscored the importance of transforming the workforce to adapt to AI integration. Gartner predicts that AI will create more jobs than it displaces by 2027, asserting that the focus should shift from job loss to workforce transformation. The survey also revealed that 74% of CIOs reported their organizations are either breaking even or losing money on AI investments, highlighting the necessity for careful planning and cost assessments in AI implementation.
Artificial intelligence is significantly transforming the information technology industry, with IT jobs projected to grow at a rate of 9% from 2024 to 2034, compared to just 3% for all other occupations, according to the Bureau of Labor Statistics. This shift is largely driven by new security concerns related to AI implementation, as noted by Peter Tsai, head of technology insights at Spiceworks, who stated that “AI opened a whole new can of worms for security.” A recent survey conducted by Spiceworks found that 63% of IT professionals believe skills related to AI prompting are increasingly important, reflecting a 53% rise from the previous year, although fewer than half feel confident in their AI capabilities. Companies are investing heavily in AI, with AI software accounting for a median of 2.7% of IT infrastructure spending, and the emphasis on training remains crucial to leveraging these technologies effectively.
Artificial intelligence is beginning to yield measurable benefits across various industries, with a notable increase in adoption among S&P 500 companies. According to new data from Morgan Stanley, the percentage of these companies reporting quantifiable advantages from AI has risen to 15% in the third quarter of 2025, up from 11% the previous year. Among firms classified as “AI adopters,” nearly 25% report tangible performance improvements. The technology sector leads this trend, with 39% of tech firms highlighting significant AI gains, a substantial increase from 26% last year. Other sectors, such as Communication Services and Financials, also show positive results, as Energy firms see their AI adoption rise from 0% to 10%.
But AI is bad at math. Recent research reveals that leading artificial intelligence models struggle significantly with mathematical calculations, achieving scores of 63 percent or less on a new benchmark called ORCA. This assessment, conducted by scientists from Omni Calculator and various European universities, tested five prominent language models, including ChatGPT-5 and Claude Sonnet 4.5. Despite high scores on other benchmarks, such as GSM8K, the researchers argue that these models often rely on memorization rather than genuine computational reasoning, leading to errors in logic and arithmetic. For instance, Claude Sonnet 4.5 recorded only 45.2 percent accuracy overall, falling short on basic mathematical tasks, highlighting the gap between natural language processing advancements and reliable computational performance.
Why do we care?
Everyone is predicting that AI will be woven into every IT task by 2030, but the data shows companies still aren’t getting value out of the AI they’ve already bought. CIOs say three-quarters of them are breaking even or losing money on their AI projects. That’s not “AI revolution”—that’s organizations paying for tools they don’t know how to use effectively.
And this is happening while workers say they don’t feel confident in their AI skills. So you’ve got massive pressure to deploy AI everywhere, paired with a workforce that isn’t ready and models that still screw up basic math. That’s a dangerous mix.
Even at the enterprise level, where budgets and talent are strongest, only a fraction of companies are reporting measurable benefits. So imagine what that means for SMBs: they’re even more vulnerable to overspend, misconfiguration, and unrealistic expectations.
And then we have the math failures. If models can’t reliably handle arithmetic or logic, we cannot treat them as authoritative. That’s a huge problem for coding, security operations, financial workflows — all the places vendors are promising “AI will take over.”
So why do we care? Because the gap between AI hype and AI reality is widening. Your job as an IT provider is not to tell clients that AI will replace everything—it’s to guide them through where AI actually works, where it doesn’t, and how to adopt it safely. The MSPs who focus on governance, skills, and realistic outcomes will win. The ones who trust the hype cycle are going to get stuck holding the bag when these models fail.