A recent study supported by Apple indicates that a new artificial intelligence model can flag health conditions with an accuracy of up to 92%. This model, known as the Wearable Behavior Model, was trained on over 2.5 billion hours of data collected from wearable devices, outperforming traditional models that rely on basic biometric measurements. Researchers found that behavioral data, such as movement and sleep patterns, often provides stronger health signals than conventional metrics like heart rate. The model was evaluated against 57 different health-related tasks and showed superior performance, particularly in dynamic health prediction tasks like detecting pregnancy and assessing sleep quality.
Nonprofits are also leveraging artificial intelligence to address global inequities, particularly in education. Education Above All, a nonprofit organization based in Qatar, has partnered with institutions like the Massachusetts Institute of Technology and Harvard University to create a free and open-source program called Digi-Wise, aimed at improving AI literacy in developing countries. The initiative allows children to navigate and combat misinformation while utilizing AI tools in the classroom. Education Above All’s generative AI chatbot, named Ferby, is currently being used by over five million children in India to access project-based learning resources.
Great news, right? Well, a recent study from the AI benchmarking nonprofit, METR (Model Evaluation & Threat Research) reveals that current artificial intelligence coding tools may actually hinder productivity rather than enhance it. In a randomized control trial with 16 experienced developers working on 246 real issues in large open-source projects, METR found that using these tools increased task completion time by 19%, contrary to developers’ initial expectations of a 24% time savings. Participants attributed the slowdown to unrealistic expectations of AI usefulness, unfamiliarity with repositories, and the complexity of codebases, as well as the additional effort required to review AI-generated suggestions. These findings suggest that, while developers anticipated increased efficiency, the reality was a reduction in productivity—though METR notes results may vary for less experienced developers or smaller projects.
A recent report from CloudZero highlights that while 91% of companies believe their investments in artificial intelligence are paying off, only half can substantiate this with concrete data. The report indicates that AI budgets are set to increase by 36% year-over-year, driven largely by spending on public cloud services, yet many organizations lack visibility into their AI costs and return on investment. The survey, which included 500 software engineers and senior managers across the United States, found that 51% of organizations do not feel confident in their ability to evaluate the return on their AI investments. The report emphasizes the critical need for effective cost attribution and tracking to ensure sustainable growth in this sector.
Why do we care?
This is a classic case of AI overpromising and underdelivering at scale. On the surface, these stories suggest AI is achieving remarkable breakthroughs: 92% accuracy in health prediction, nonprofits using generative AI chatbots to scale literacy to millions of children, and 91% of companies believing their AI investments are paying off.
But two counterpoints disrupt the narrative:
- METR’s study shows seasoned developers became 19% slower using AI coding tools—not faster as expected.
- CloudZero reports only 50% of companies can prove ROI from their AI initiatives despite soaring budgets and optimism.
This contradiction is the signal IT providers must tune into.
AI tools often underdeliver on productivity promises—especially in complex, real-world environments.
SMB clients could easily overspend on AI without clear ROI tracking, leading to wasted budgets and disillusionment.
The skills gap and rising salaries for AI roles are particularly acute for small and mid-market firms, making reliance on external expertise inevitable.
If MSPs rush to adopt AI themselves or resell AI tools without understanding these pitfalls, they risk becoming part of the problem rather than the solution.
For IT service providers, it’s a cautionary and strategic moment:
Caution: Don’t sell AI as a magic bullet. Even experienced developers are struggling with AI-assisted coding, and half of companies don’t know if their investments are worth it. That same lack of clarity will cascade into the SMB space.
Focus your role on AI financial governance—helping clients evaluate, monitor, and optimize their AI spend and outcomes. This isn’t about building models; it’s about being the AI CFO for clients, ensuring their enthusiasm doesn’t lead to wasted budgets.
MSPs and vCIOs who can quantify AI ROI and align AI investments to specific, measurable business outcomes will stand out in a market full of vendors chasing the next AI headline.