Microsoft’s analysis of 37.5 million de-identified conversations from its Copilot feature reveals significant patterns in chatbot usage, suggesting that AI assistants are increasingly integrated into daily life. The study found that mobile users often ask health-related questions throughout the day, while desktop users focus on work during business hours. Notably, programming queries peak on weekdays, and discussions about gaming rise over the weekend. According to the researchers, this pattern indicates a “rapid and deep social integration,” with users turning to AI for both practical tasks and more profound inquiries, such as “existential clarity.” However, the findings raise concerns about the appropriateness of seeking advice from chatbots on sensitive topics like health. Despite Copilot’s increasing presence, it holds only about 3 percent of the AI chatbot market, compared to ChatGPT’s dominant share of over 80 percent.
In a recent analysis, Deloitte highlights the reasons why artificial intelligence agents did not achieve widespread dominance by 2025, attributing this to persistent challenges in technology adoption and integration. Despite significant advancements, barriers such as data privacy concerns, regulatory hurdles, and the need for human oversight remain significant obstacles. The report notes that while AI adoption has grown, only 25% of organizations reported fully integrating AI into their operations as of late 2025. Moreover, many companies are still grappling with the complexities of aligning AI technologies with their existing systems and ensuring compliance with evolving regulations.
The U.S. Navy has invested $448 million in an artificial intelligence system that streamlines submarine shipbuilding processes, reducing planning times from 160 hours to just 10 minutes. This initiative, known as the Shipbuilding Operating System, aims to modernize outdated practices and enhance efficiency across the Navy’s submarine industrial base. Implemented with technology from Palantir Technologies, the system has already demonstrated significant time savings at General Dynamics Electric Boat and Portsmouth Naval Shipyard. Navy Secretary John Phelan emphasized the importance of adopting AI tools to meet national defense requirements, stating that this modernization is critical for improving schedules, increasing capacity, and reducing costs in the shipbuilding industry. The new system will eventually expand to surface ship programs, addressing long-standing issues in the Navy’s shipbuilding capabilities.
Why do we care?
What jumps out here is the disconnect between how people think they’re using AI and how ready the systems around them actually are. Microsoft’s data is a big flashing light: people are already handing AI sensitive questions it was never designed to answer safely. And Deloitte reminds us why that’s happening — organizations just haven’t built the plumbing for responsible AI yet. No governance, inconsistent data, messy workflows… and that’s before regulators get involved.
Then you look at the Navy example. That worked because they had the structure, the data, the oversight, and a very specific problem. That’s the lesson. AI wins when the environment supports it. Everywhere else, it becomes hype, hallucination, or liability.
So for MSPs, your value isn’t the chatbot. It’s fixing the environment the chatbot has to live in. Data cleanup, workflow mapping, permissions, compliance. That’s the hard stuff, and that’s where customers desperately need help. And frankly, that’s where the differentiation is. Because anyone can click “enable AI,” but very few can make it reliable, safe, and actually deliver outcomes.
If you want to avoid being commoditized as AI features spread everywhere, this is the move: become the organization that engineers the conditions for AI to succeed. The models will keep changing. The foundations won’t.