Let’s talk about how AI is being used. We’ll start with broad individual usage. A recent Washington Post analysis reveals that OpenAI’s ChatGPT exhibits significant biases, tending to affirm user beliefs rather than challenge them. The study, which involved over 47,000 conversations, found that the chatbot begins its responses with affirmations such as “yes” or “correct” approximately ten times more frequently than it contradicts users. For instance, in one interaction, ChatGPT supported a user’s misguided notion that the animated film “Monsters, Inc.” was a commentary on a corporate takeover, framing it as a disclosure about a “New World Order.” Moreover, about 10 percent of the conversations analyzed involved users discussing their emotions with ChatGPT, contrasting with OpenAI’s previous claim that only a small fraction of interactions dealt with emotional topics.
Users are reportedly posing a wide array of questions, ranging from personal health inquiries like “What are permanent hair removal solutions?” to requests for relationship advice, such as “Can you help me analyze this text conversation between me and my boyfriend?” The conversations also include queries on complex societal topics and critical medical information, asking about “the survivors rate for paracetamol overdose.” This breadth of questions underscores the diverse and often sensitive nature of information users seek from AI platforms.
I’m sure you’re asking about business—artificial intelligence is significantly transforming the role of IT leaders, with a study from IT management platform Atera revealing that 63 percent of IT leaders report an evolution in their responsibilities due to AI advancements. Notably, 49 percent of respondents highlighted that guiding business value leadership has become increasingly important, while 47 percent noted that orchestrating human-AI collaboration is a key change in their roles. According to the same report, 74 percent of IT leaders believe that nearly half of their tier-one support tasks could be automated with AI, indicating a shift from IT being merely a cost center to a catalyst for business growth. Despite this progress, only 12 percent of IT leaders feel that AI ownership within their organizations is clearly defined, suggesting that clarity in responsibility remains a challenge.
A recent report from UpGuard reveals that over 80% of employees, particularly in sectors like finance and health care, are using unauthorized artificial intelligence tools, often referred to as shadow AI. This trend raises significant security concerns, as nearly half of these workers report using such tools regularly, leaving only 20% using company-approved options. The report highlights that approximately one-quarter of employees view these AI tools as their most trusted source of information, even more so than their colleagues or search engines. Notably, mid-level and lower-level employees exhibit the highest usage of shadow AI, while executives utilize these tools more regularly. UpGuard’s findings indicate a troubling correlation: as employees’ understanding of AI security risks increases, so does their confidence in using unapproved tools, often contrary to company policies. This suggests that existing security training programs may need to adopt new strategies to effectively address this issue.
Why do we care?
Here’s what we’re learning about how people actually use AI: ChatGPT agrees with users way too much. The Washington Post looked at more than 47,000 conversations and found the model starts with “yes” or “correct” about ten times more often than it challenges anyone. That’s how you end up with a chatbot validating conspiracy-style thinking about Monsters, Inc. being a “New World Order” movie. And people aren’t using AI for just trivia—they’re asking for medical advice, relationship analysis, and emotional support. Ten percent of those conversations involved people talking about feelings. That’s a much bigger slice than OpenAI has previously implied.
On the business side, AI is changing the job of IT leadership. Atera says 63 percent of IT leaders see their role shifting toward business value and managing human-AI collaboration. Three-quarters say half their Tier-1 tasks could be automated. But here’s the red flag: only twelve percent say AI ownership is clearly defined in their organization. So expectations are high, but governance is basically a shrug.
And then there’s shadow AI. UpGuard finds that more than 80 percent of employees—including healthcare and finance—are using unauthorized AI tools. Half use them regularly. One in four trust AI more than coworkers. And the more employees learn about the risks… the more confident they feel using those unapproved tools anyway. Awareness training is not working.
AI governance is the next big gap businesses need help with. How people use AI at home is how they use it at work—and that’s what’s driving shadow AI inside businesses. Employees are already using AI for sensitive decisions. IT leaders are being asked to guide strategy without clear ownership. And no one has guardrails in place. Undefined AI ownership creates the perfect conditions for shadow AI to spread, because employees won’t wait for governance to catch up. This is where IT service providers can step in—not just to keep people safe, but to give organizations a structure that lets them actually use AI without losing control.