5 AI Delusions That Could Derail Your Business

Smart leaders rarely lose the plot in one spectacular act of stupidity. They lose it one reasonable assumption at a time. That’s one idea that stuck with me from workplace researcher Kelly Monahan’s new book, Reclaim the Plot. She calls it “the delirium.” Reasonable assumptions pile up until leaders lose sight of what their decisions cost, who pays the bill, and whether they’re still solving the problem they started with. Here are five seductive beliefs about AI that contain just enough truth to be dangerous. 1. If AI makes people more productive, it will make the company more productive. The calculus feels simple. If hundreds of employees become more productive with AI, surely the organization does, too. Except organizations are messier than simple arithmetic. In our latest Work AI Index from the Glean Work AI Institute, 75 percent of digital workers say AI has made them more productive. Yet only 13 percent say AI has significantly improved their organization’s performance and outcomes. One reason is what Chip Heath and Nancy Staudenmayer called “coordination neglect.” People focus on dividing up the work and underestimate the messy, expensive work of putting it all back together. AI can amplify that impulse. Individuals crank out more work faster and then dump a heaping pile of review and cleanup work on others downstream. One employee saves an hour, but three poor colleagues inherit 30 minutes of cleanup. That hidden human labor is called botsitting. Workers report spending 6.4 hours a week on it — feeding AI context, checking its work, fixing mistakes, and cleaning up confident-but-wrong outputs. 2. AI will free people up for higher-level work. Leaders have fallen head over heels for a comforting AI fairytale: let machines handle the drudgery so humans can graduate to higher-value work, like strategy, creativity, relationships, and judgment. The problem is that saved time doesn’t stay saved for long. Monahan cites research showing that 96 percent of C-suite leaders expected AI to increase productivity. Meanwhile, 77 percent of full-time workers said AI had inflated their workload, and 65 percent said employers were asking more of them because leaders assumed AI had made them more efficient. Instead of reinvesting the AI time savings into all that glorious “higher-order” work, workers are often rewarded with loftier targets and heavier workloads. Worse, AI sometimes cannibalizes the work people value most. More than half of workers in our Work AI Index say AI has already automated meaningful work they’d rather keep human. Don’t just ask: How much time did AI save? Instead ask: What happened to the time? 3. If you can automate it, you should automate it. Friction has become a dirty word at work. And it’s tempting to aim AI at every annoying ounce of it. But not all friction is bad. Some friction is productive because it forces people to slow down and think. Researchers studying the IKEA effect have shown that people often value things more when they’ve had a hand in creating them. The principle extends far beyond furniture. Strip away the struggle of creating something, and you can strip away the ownership and pride people feel over the outcome, too. Struggle is also how humans develop expertise. Monahan warns that companies risk “optimizing apprenticeships out of existence.” From the top of the org chart, junior tasks can look like low-value grunt work begging to be automated away. But that work is often where people make mistakes, learn what “good” looks like, and develop the know-how that isn’t written down in the training manual. Experts can skip steps only because they once spent years stumbling through them. The challenge isn’t to eliminate friction. It’s to distinguish between the bureaucratic sludge that wastes people’s time and the productive friction that teaches them something. 4. AI will fix a badly designed organization. When organizations encounter a problem, their knee-jerk reflex is often to hurl more resources at it. Throw more money at the problem — more people, more technology, and more bureaucracy. Indeed, across a series of experiments, researchers found that when people were asked to improve something, they systematically overlooked solutions that involved removing things — even when subtraction was the better solution. Stanford professor Bob Sutton calls this tendency “addition sickness.” It’s easy to bolt an AI agent onto a slow approval process. Or add a chatbot to a terrible expense system. Or deploy an AI notetaker to a soul-sucking recurring meeting that should have been put out of its misery in 2024. However, a bad workflow doesn’t magically improve when you bolt AI onto it. Before adding AI, make sure you understand how the work flows. If you can’t map out the process, you’re not ready to apply AI to it. Sometimes the smartest solution is to avoid using AI altogether. 5. Becoming AI-first is a strategy. Executives love to thump their chests and trumpet that their company is on a bold journey to become AI-first. The grand declaration is usually followed by hackathons, mandatory use cases, fleets of half-baked agents, and dashboards glowing with vanity metrics like logins, prompts sent, licenses activated, and tokens consumed. Monahan calls this the refounding illusion, or the naive belief that an established company can somehow acquire the speed of a Silicon Valley startup simply by declaring itself reborn as AI-first. But companies don’t magically shed decades of complexity, gnarly internal politics, and workarounds just because someone put “AI-first” on a strategy slide. And once AI-first becomes the goal, organizations start shoehorning AI into places where it doesn’t belong. It’s another nasty case of addition sickness. When drift becomes delusion None of these five delusions starts with a glaringly bad idea. They start with good ideas stripped of their conditions and context. That’s how organizations lose the plot. One seemingly sensible decision at a time, until a pile of reasonable choices becomes full-blown delirium. EXPERT OPINION BY REBECCA HINDS, REBECCA HINDS, AUTHOR OF YOUR BEST MEETING EVER @REBHINDS