McKinsey consultants sat down this year with executives at 15 companies recognized as AI leaders. The opening line of the resulting report: nearly every company has AI tools, but few actually know how to use them. If you were expecting a success playbook, it's a deflating way to start.
What makes the report worth reading isn't the prescriptions—it's the common mistakes. Even companies that call themselves AI-native trip over the same things. Seven patterns emerged. And most of those seven have nothing to do with tool performance or budget. The gap between accumulating tools and absorbing them turned out to be deeper than expected. That gap is where most organizations are stuck right now.
The Speed Illusion
The first error the report consistently surfaces is the speed illusion. The expectation that AI-generated drafts will accelerate the entire workflow plays out differently in practice. Reviewing AI output, editing it, deciding which parts to keep, and communicating those decisions within a team all end up consuming more time than anticipated. There are real speed gains—they exist. But new slowdowns appear alongside them. When the gap between how fast a tool generates and how fast people can absorb doesn't close, overall productivity falls short of expectations.
The second pattern is knowledge hoarding. It's always one or two enthusiastic employees who first pick up a new AI tool. The problem is that their experience stays personal. Six months later, the same people are using the tool the same way. AI adoption gets locked into individual capability rather than team capability. The result: the productivity gap between those who use the tools and those who don't widens over time.
Third is validation avoidance. When no explicit system is built from the start for checking AI-generated output, people implicitly begin to trust it. When errors surface, the refrain is "the AI got it wrong"—but the prior failure was never designing a verification process. Among the 15 companies, the difference between those that built explicit validation structures and those that didn't showed up not in error rates but in how quickly errors were caught.
The remaining patterns follow a similar thread. Data siloed by department, requiring fresh access negotiations for every AI project. Unclear ownership over who has authority to act on AI-generated recommendations. Staff whose time is needed to operate AI systems being squeezed out because AI adoption is read internally as a signal of headcount reduction. And tool fatigue—as new tools keep piling up, the overhead of deciding what to use where becomes its own cognitive burden.
Looking at all seven, one thing stands out: there is no technology problem on the list. Nothing about GPT-4 being slow, API instability, or runaway costs. Every pattern originates in how people and organizations receive tools—not in the tools themselves. Now that AI performance has crossed a meaningful threshold, the bottleneck has shifted. It's no longer on the tool side. It's on the habits side.
The Skeptics Have a Point
Among growth-stage startup CPOs who encountered this report, the pushback was immediate: "These seven principles are already written in enterprise language." Applying frameworks built for organizations of hundreds—verification systems, data governance—to a three-person team or a solo operator means spending more time on design than execution. That criticism isn't entirely wrong.
There's a more fundamental objection too. Codifying operating principles as received wisdom may itself suppress experimentation and iteration. Real cases are being reported of organizations that move fast, fail quickly, and adjust outperforming more deliberately structured ones on AI adoption. The McKinsey report doesn't fully refute this. Some of the 15 companies said their AI-use culture had developed organically, with no formal guidelines—patterns emerged through repeated experimentation, not pre-established principles.
A separate critique is also worth taking seriously: the seven findings were never measured. An interview-based report depends on executive self-perception. There's no controlled data in this report quantifying how much knowledge hoarding or validation avoidance actually affects performance. The sample of 15 companies is limited. The criteria for what counts as "ahead in AI adoption" are never explicitly defined.
None of that makes the report worth discarding. The seven patterns are more useful as post-hoc diagnostics than as design blueprints. When used as a reference for "where is our team stuck right now," the limitations of being an enterprise document shrink considerably.
Is Your AI Routine Actually Working?
Yes, the McKinsey report is written in enterprise language. But inside it are points that translate directly to solo operators and small teams.
It's worth asking whether the AI tools you're currently using are actually shaping your decisions. If you're stopping at draft generation, you're less using the tool and more just having it around. Mapping out which judgments you delegate to AI and which you insist on making yourself reveals your actual level of use. A good starting point: track how much you're editing AI-generated proposals, and what criteria you apply when you do.
Your verification routine also deserves a look. If you haven't made explicit what criteria you use to review AI-generated content, plans, or summaries, the review process will quietly disappear at some point. For a solo operator, this doesn't need to be a complex internal system. Writing down three things you personally need to verify in any given draft is enough to constitute a verification routine. That simple checklist makes a perceptible difference in the quality of a blog post or a proposal.
The number of tools you're holding is worth auditing too. If you add every new AI tool as it launches, you'll eventually find yourself spending more energy deciding where to use what than actually using anything. Tool fatigue—flagged in the McKinsey report—isn't just an enterprise problem. Reviewing your active tool list once a quarter and cutting what isn't genuinely contributing to decisions helps sustain real AI adoption over time.
Across discussions of what capabilities remain distinctly human in the AI era, one observation keeps appearing: as tool functionality expands rapidly, the gap grows between that functionality and the ability to judge where and how far to deploy those tools in any given context. Simply logging more hours with a tool doesn't close that gap. The capacity to read context and know when to withhold judgment is what creates meaningful differences in capability—and that's a different layer of skill than familiarity with any particular product.
That's what McKinsey observed across 15 companies too. Before the question of which tools to use comes the question of what decisions to make when a tool is in front of you. The teams called AI-native weren't the ones who deployed every new feature first. They were the ones who decided in advance what each tool was for, and had routines in place to correct course quickly when those decisions turned out to be wrong.
