Seventy-two percent of chief executives now say they are their company’s main decision maker on AI, twice the share of a year earlier, in BCG’s latest global survey of roughly 2,400 executives (BCG, 2026). Read that number slowly, because it is the largest organizations on earth arriving, at considerable expense, at a conclusion my doctoral research reached from the inside more than a decade ago: awareness of a new technology moves through an organization from the top down, or it does not move at all. This piece gives that finding a working model. I call it the awareness cascade, three gates in a fixed sequence, where each gate opens only after the one before it, and where the height of the first gate sets the ceiling on everything downstream. The first gate is the leader’s personal engagement. The second is the team’s awareness. The third is organizational fluency. If your team’s AI capability is stuck, the model says the blockage is almost never where the frustration is pointing.
The finding underneath the model
The cascade is not a metaphor reverse-engineered from this year’s headlines. It is the pattern my 2014 dissertation kept surfacing across nineteen interviews with Fortune 500 leaders, their employees, and technology provider staff, a study built deliberately as a side by side comparison of five leaders who adopted a wave of new technology and five who resisted it. The full construction of that comparison is in adopters and resisters, and one of its conclusions was stark enough that I have repeated it ever since: where the leader was unaware of the technology, the subordinates were unaware of it too (Goodwin, 2014).
What made the finding land was its consistency in both directions. Awareness did not percolate upward from younger employees. It did not arrive through vendors, and it did not seep in from the trade press. In the resisting organizations, the leaders had not personally engaged with the platforms, and so their teams had not engaged either, whatever tools the company technically owned. In the adopting organizations, the leaders had engaged personally, could explain the technology in their own words with real examples, and their organizations moved behind them. The direction of travel was one way in every case the study examined, and that one way direction is the entire load bearing claim of the cascade model. Awareness behaves like water moving downhill through an organization chart, and the leader is the elevation.
Gate one: the leader’s own engagement
The first gate is the leader’s hands on the technology, and the standard is specific. It means personal use on real tasks, recently and repeatedly, rather than sponsorship from a distance. A leader who approved the budget, attended the demo, and assigned the rollout has done three useful things and has not touched the gate. The camp test in adopters and resisters probes this directly with its second question: have you personally used it on a real task in the last month, not a demo, a task. The symptoms of a closed first gate are just as specific. Delegation language, as in someone is on that for us. Borrowed opinions that trace back to a vendor deck rather than to use. A vocabulary of features with no examples attached. None of those symptoms feel like resistance from the inside, which is what makes the gate dangerous.
The current executive data has become unusually candid about this gate. In BCG’s research on companies lagging with AI, only 8 percent of C-suite leaders are deeply engaged with the technology, and the same research names the act of delegating AI down to middle management as a defining failure of the laggards (BCG, 2025). The 2026 survey measures the other side of the gate in hours: the large pragmatist middle of the CEO population it profiles now spends roughly seven hours a week using, studying, or thinking through the technology, and the report’s closing counsel to executives is to develop personal fluency through repeated practice (BCG, 2026). Half of the surveyed chief executives believe their own job is at stake if their AI efforts do not pay off, which goes some way toward explaining the hours. The corner office has concluded, later than the research but with real money behind it, that this gate cannot be delegated open.
Gate two: the team’s awareness
The second gate only opens from the far side of the first, because awareness is transmitted by observed behavior rather than by announcements. A leader who uses the technology in visible, ordinary ways licenses the team to take it seriously: the summary produced in minutes instead of an afternoon, the draft that arrives already structured, the question answered from a tool in the meeting rather than deferred to next week. A leader who announces an initiative while personally abstaining teaches the opposite lesson with perfect clarity, since people learn what matters in an organization by watching where their leader’s own hours go. My study saw both classrooms. Teams under engaged leaders were aware of the technology and what it was for. Teams under disengaged leaders were not aware of it in any operational sense, no matter what had been purchased on their behalf (Goodwin, 2014).
Language is the tracer for this gate, and you can hear its state in a hallway. Teams under engaged leaders talk about the technology the way they talk about any working tool, in terms of tasks: what it drafted, what it caught, what it still gets wrong. Teams under disengaged leaders either do not talk about it at all or talk about it in the abstract, as an initiative, a mandate, a thing the company is doing, because the only vocabulary that ever reached them was the announcement’s. An organization’s words for a technology are inherited from wherever its awareness came from, which under the cascade means they are inherited from the leader or from nobody.
You can read a closed second gate off the shelf, sometimes literally. Tools bought and unopened. A pilot that needs constant pushing because nothing in the team pulls. Training that evaporates within a month because nothing in the daily environment reinforced it. The modern numbers put a multiplier on the open version: McKinsey finds the organizations capturing real value from AI are three times more likely than their peers to have senior leaders who own and actively champion the work (McKinsey, 2025). Championing, in cascade terms, is what an open first gate looks like from below, and the case that this makes AI a leadership problem before it is ever a technical one is argued in full in AI is a leadership problem, not an IT problem.
Gate three: organizational fluency
The third gate separates knowing about the technology from working differently because of it. An aware team can name the tools and has tried them. A fluent organization has rebuilt pieces of its daily work around them, so that the technology sits inside the flow of the business rather than beside it. That rebuilding is where the financial return has been hiding all along. McKinsey’s research keeps arriving at the same answer about value: of everything organizations do with AI, redesigning workflows end to end is the change most associated with bottom line impact, and only about one in five organizations using the technology has done that work (McKinsey, 2025).
The distinction is easiest to see at small scale. An aware five person firm has a subscription, a couple of enthusiastic users, and its old workflows with the technology sprinkled on top, so every gain depends on someone remembering to reach for the tool. A fluent five person firm has changed what the workflow is: the intake that populates the calendar on its own, the draft that starts nearly finished, the summary that files itself into the record, and nobody has to remember anything because the flow itself carries the technology. Awareness produces users, while fluency produces a different operation, and only the second one shows up in the financial results the surveys keep failing to find.
The model also explains the strangest pair of numbers in the current data. Adoption at 88 percent of organizations alongside scaling at 7 percent looks like a contradiction until you place the population on the cascade. Nearly nine in ten organizations have cracked the second gate somewhere, one function, one team, one tool in real use. Fewer than one in ten have passed the third. The gap between those figures is not a technology shortfall, since both groups can buy the same products on the same day. It is tens of thousands of organizations standing between gate two and gate three, and in most of them, the model predicts the stall traces back upstream to a first gate that never opened wide enough to carry the weight.
Why the gates cannot be skipped
The practical force of the model comes from mapping spend onto gates. Buying an enterprise platform is gate three spending. Commissioning a training program is gate two spending. Both are legitimate, and both starve behind a closed first gate, which is why so many well funded initiatives produce shelfware and a quietly cynical team. The sequence only runs one direction, so capacity added downstream of a closed gate is capacity that cannot flow. Run the arithmetic of ceilings and the conclusion gets uncomfortable in a useful way: the cheapest capacity increase available to most organizations is at gate one, and gate one happens to be the only gate that money cannot open directly, because the purchase it requires is the leader’s own attention. The reason understanding must precede adoption at the personal level, for the leader as much as anyone, is the subject of you cannot adopt what you do not understand, and the full set of levers the leader controls is mapped in the 5 drivers of AI adoption.
The same mapping exposes the most popular skip attempt of all, which is hiring the gate open. Bringing in a talented specialist, or naming an AI lead, is a legitimate gate two and gate three investment, and it changes nothing about the first gate, because the cascade runs on the leader’s observed behavior rather than on the org chart’s coverage. A specialist reporting to a disengaged leader inherits a closed gate and spends their tenure pushing against it, which is why so many of those hires produce impressive demonstrations and no organizational change. Delegation is the failure the current research names outright, and a hire is delegation with a salary attached.
What the model does not claim
It does not claim the leader must become an engineer. The gate one standard is a plain language explanation of what the technology does in a business like yours, plus recent use on real tasks. That is a modest bar, and the adopting leaders in my study cleared it without technical backgrounds, in their own words rather than a vendor’s (Goodwin, 2014). A leader who can describe last week’s real use of the tools has an open first gate, whatever their job history says.
It does not claim bottom up enthusiasm is worthless. Individuals anywhere in an organization can and do build serious personal skill, and that skill is real. The claim is narrower and better supported: in the organizations my research examined, awareness at the organizational level only ever arrived through the leader, so enthusiasm below a closed gate stayed personal. It made individuals more capable and made the organization no different, which is its own kind of loss.
And it does not claim the survey correlations prove the direction of causation by themselves. A skeptic can read the three times champion finding in reverse, with success producing engaged leaders rather than engaged leaders producing success. The reason I read the direction with confidence is that qualitative work observes mechanism where surveys observe outcomes. The interviews did not merely find engaged leaders standing near capable teams. They watched awareness fail to move whenever the leader had not moved first, across every case, in the participants’ own accounts. Mechanism from the small study plus correlation from the large ones is how the two kinds of evidence earn each other’s trust, and on this question they agree.
Reopening the first gate
The discouraging half of the cascade is that nobody can open your first gate for you, not a consultant, not a hire, not a platform. The encouraging half is that gate one is the only gate that requires no budget, no committee, and no permission. So the move is small and starts this week. Pick one real task from your own desk, not a demo, and run it through the technology yourself. Put a recurring hour on your calendar for the practice, which is a fraction of what the executives in the current surveys are logging, and judge yourself a month from now by the camp test’s standard of recent real use. Your team is watching where your hours go whether or not you intend the lesson, so the hour teaches either way.
If you want an honest reading of your own first gate before you start, the Omnine AI Readiness Assessment measures the leader side levers directly. It takes about three minutes, and it will tell you whether the ceiling over your team is the one you suspected.
References
Goodwin, M. R. (2014). A qualitative descriptive multiple-case study: Fortune 500 leaders’ social business platform adoption (Doctoral dissertation, University of Phoenix). ProQuest Dissertations Publishing (UMI No. 3648813).
McKinsey & Company. (2025). The state of AI in 2025: Agents, innovation, and transformation. McKinsey Global Survey on the state of AI.
Boston Consulting Group. (2025). Are you generating value from AI? The widening gap.
Boston Consulting Group. (2026). BCG AI Radar 2026: As AI investments surge, CEOs take the lead.