Picture two rooms in the same company on the same morning. In the boardroom, the message is simple: we're all in on AI. Two floors down, in a daily standup, somebody asks the only question that actually matters. "Okay, so what does that mean for us by Friday?"
Nobody has a good answer. And that gap, right there, is the whole problem.
I'm Lance Dacy with Big Agile, and here's why that question matters right now. The pressure to "get AI" has moved to the very top of the organization. When that pressure arrives without a translation, it doesn't disappear. It rolls downhill into your teams. I do a lot of work with teams right now, and I get to float across the organization. What I see is that teams labor and toil in this stuff far more than most executives realize.
The Pressure Doesn't Disappear. It Rolls Downhill.
Let me show you what this looks like on the ground. A VP walks out of the leadership offsite fired up. The board wants AI everywhere, and fast, because we have to beat our competitors. So she sends the note. Every team has now gotten some version of it: "We need to be using AI more. Let's move."
The team reads it three times, and then they start guessing. One engineer quietly starts piping customer data into a tool nobody approved, figuring it'll give better insight into what's going on. A product owner pauses a release she had scheduled, because she honestly isn't sure whether AI should change that plan now and she doesn't have good enough data to decide. And two people keep their heads down and wait for it all to blow over. You know who you are.
Same email, three different reactions, zero shared rules. The team isn't being lazy and they aren't resisting the technology. They're missing the one thing the urgency forgot to include: a translation. The board handed down a goal. Nobody handed down how to act on it safely.
So most teams that look like they're dragging their feet on AI aren't scared of the technology. They're scared of getting blamed for guessing wrong on something the boss just called urgent, because then that same boss calls and asks, "What were you thinking?" That, to me, is the leadership gap. And the data says it's about to get a lot more common.
What's Actually Happening at the Top
This is my read, based on my experience, but the numbers back it up. The one I keep coming back to is from Boston Consulting Group and their AI Radar study, published this past January. They surveyed more than 2,000 executives, including 640 CEOs. Here's what they found.
72 percent of CEOs now say they are the main decision maker on AI at their company. That's double the share from a year earlier. The person who used to delegate this to the CIO is now holding the pen. But the number that should really stop you is the next one: half of those CEOs believe their own job stability depends on getting AI right this year.
Well, you'd want to control it too, wouldn't you? This isn't a side project anymore. It's personal. My name's on it, and by gosh we're going to succeed at it. Around 90 percent of them expect AI agents to show real, measurable returns this year. So you have ownership at the top, personal stakes, and a fast clock. That's a lot of energy pointed downhill.
What makes this honest instead of just scary is that the same executives told on themselves. In a BCG follow-up, a majority of CEOs said their own boards are pushing faster than the organization is actually ready for. That's not new. It happens all the time. But it means the people who own AI are the ones telling us the pressure is outpacing the readiness, and they're in a position to do something about it.
Ownership Went Up. Control Did Not Follow.
A second study shows you exactly where that gap lands. Grant Thornton surveyed nearly a thousand senior leaders here in the US. This one is self-reported and US-only, so treat it as signal, not gospel. They found that 78 percent of leaders are not confident they could pass a basic AI governance check within 90 days. They call it the proof gap: plenty of AI activity, very little ability to explain it, measure it, or defend it. I did a video on that a few months ago, and it's worth a watch.
Put the two studies side by side and the story gets clear. The top of the house has taken ownership and tied its own job to the results. The middle of the house can't yet prove what's working. And that squeeze is exactly what your teams on the bottom floor are feeling: high urgency from above, no clear evidence from below.
Owning a problem and controlling it are not the same skill. You can be fully accountable for AI and still have zero operating system for running it. That's what most companies are fighting with right now. So the question isn't whether the pressure is real. It absolutely is. The question is what you do with it before it reaches your team as raw anxiety, because they already have enough to worry about.
Convert the Pressure. Don't Pass It Down.
Here's the job, whatever seat you're in, whether you're the CEO who owns this or the chief AI officer I expect a lot of companies to hire next. In your next leadership meeting, don't let the pressure roll downhill the way it normally does. You take a board-level goal and turn it into clear decision rights and safe-to-learn rules before it ever lands on the team. We do this constantly in product development with a definition of done and acceptance criteria. Leaders need to do the same thing with AI. Let me show you what that translation looks like.
Turn the goal into one observable outcome
Say the goal from the top is the classic one: "Use AI to move faster." That sentence is useless to a team. They can't act on it and they can't tell you whether they succeeded, so it just produces anxiety. Translate it instead. Turn it into one outcome the team can actually see, like reducing the time from a customer request to a first working response. Name the number you're watching. Now it's observable, and a team can aim at it and win.
Draw the safe-to-try line
Next, draw a line most leaders skip. Name what's safe to experiment with freely and what needs a second set of eyes first. Drafting an internal document with AI? Go for it, no permission needed. Feeding real customer data into a tool you don't fully understand? Stop. That one gets reviewed first. And by whom? I bet you can't name that person yet. That single distinction is what kills the quiet, scary guessing from the story earlier.
Name who gets to say yes
Then name who actually gets to greenlight things. Who can approve a small experiment without booking a meeting? Who has to be consulted before anything touches a customer? Can you name those people? When decision rights are clear, people stop freezing and start moving toward the goal, failing, and learning. That's exactly what we want them doing.
Make it safe to surface the misses
The piece that protects everything else: decide out loud how a mess gets surfaced. If reporting a failed AI experiment feels like confessing a crime, people hide the misses. That's a psychological safety problem, and hidden misses are exactly how your proof gap gets worse. So say it plainly. If something doesn't work, you bring it to the next review, we look at what we learned, and we move on. No punishment for an honest, well-run experiment. This is where leaders need a bigger say.
That's the whole translation. A vague directive becomes an observable outcome, a clear safe-to-try line, named decision rights, and a safe way to report the misses. It's easy to say and hard to do. But notice what just happened: you didn't slow the company down. You made it possible to actually move, because the team finally knows the rules of the game instead of sitting paralyzed and flying under the radar.
A One-Sentence Exercise for Your Next Meeting
Grab whatever AI directive in your world your team is currently guessing about. The email, the offsite slide, the message from the top. Write it at the top of a blank document exactly as it came down. Then, underneath it, answer one question: what is the single observable outcome this is really asking my team to produce?
Not a vibe. Not "be more AI-forward." A real outcome you can point at and say yes, that happened, or no, it didn't. Bring that one sentence to your team and ask if you got it right. Then watch what happens. The fog starts to lift. We like to say CEOs are very good at telling everyone to drive fast in the fog. Start lifting the fog instead, and the goal becomes something the team can see and aim at.
One directive, one observable outcome, one short team conversation. You're done when your team can repeat that goal back to you in plain language. If they can, you've already done more translating than most of the organization above you.
Owning AI Isn't Controlling It
The pressure from the top is real. I have empathy for it, and it isn't going away. But your job as a leader is to be the place where that pressure gets converted into something your team can safely act on instead of hide from. The leaders who do this well this year won't be the ones with the boldest AI vision. They'll be the ones who turn that vision into rules their teams can actually follow.
If you want to go deeper on translating board-level AI goals into something your teams can run with, that's a lot of what we work on inside our classes and coaching at Big Agile. Come learn how to build the operating system, not just the mandate. And if this was useful, forward it to the next executive who owns your AI roadmap. They're probably feeling the squeeze too.
None of us is smarter than all of us together. Let's figure this thing out.