A VP I worked for once yelled at me that I didn't have executive voice. Not said. Yelled. He knew, he told me, because he'd taken several classes in it — he'd studied the thing, and I wasn't demonstrating the markers he'd been trained to look for. Then he waited, as though the volume had settled the matter and I would now fall in line.
That was the whole play, in his mind: say it loud enough and I'd start following his direction. Never mind that he'd hired me to forge a path — to bring judgment he didn't have and chart a way he couldn't see. He wanted that judgment right up until it pointed somewhere other than where he was already pointing. Then he wanted compliance, and he reached for volume to get it.
What bothered me wasn't the verdict. It was that the yelling was the proof. A man who raises his voice to establish his authority in a room is showing you exactly how much of it he has. I never expected to be quoting Tywin Lannister as a model of executive sense-making. But here we are: any man who must say "I am the King" is no true king. Real authority doesn't need the announcement, and it never needs the volume. It's already doing the work.
Executive voice is a useful thing to be accused of lacking, because there's nothing to check it against. No artifact, no output, no decision you can point to and measure. It floats. That means it can be applied to anyone, for any reason, and defended indefinitely.
And what it rewards, most of the time, is the performance of authority rather than the exercise of it — the volume, the certainty, the credential, sounding like the person everyone already pictures when they think "leader." It has never been a neutral measurement. Whose voice already registers as executive, and whose has to be coached into it, lines up a little too closely with who was already holding the power.
I've thought about that conversation a lot since, because the same reflex runs through how organizations decide what work counts. We reward the performance of judgment over the exercise of it. Often we do worse than that: we ask for the judgment out loud — hire for it, say we want people who can steer — and then reward the ones who fall in line, and punish the judgment the moment it points somewhere inconvenient. An organization that does this shouldn't be surprised when the judgment it hired for goes quiet.
And the exercise of judgment usually doesn't perform. It looks like the person who catches the cross-team dependency before it slips, who handles the stakeholder that would otherwise detonate the timeline, who keeps three teams telling the same story about what they're building. That is executive judgment — seeing around corners, holding the shape of the whole — done without announcement. Which is precisely why it goes unseen.
It has names. Tanya Reilly calls it glue work. Researchers call it non-promotable work: the tasks that matter to the organization and do nothing for the person who does them. It is real, skilled labor. And it is systematically invisible, for the same reason that VP couldn't recognize leadership that wasn't performing itself at him. We've trained ourselves to hear the sound of authority instead of watch for the substance.
Which brings me to what's about to happen to that layer of work. Agentic AI is aimed straight at it — the coordination, the summarizing, the chasing, the reconciling. That's the pitch, and it isn't wrong about the motion. Agents are good at the motion. But the motion was never the point. The point was the judgment threaded through it: which slipping dependency actually matters this week, which exception is the tell, which quiet status update is a person about to quit. An organization that never learned to see that judgment — because judgment doesn't raise its hand — looks at the coordination layer and sees only overhead. It automates the motion, strips out the discernment, and calls the result efficiency.
We have made a version of this argument before: when execution gets cheap, judgment becomes the scarce thing, and judgment is a team capacity, not an individual trait. Glue work is where a lot of that capacity actually lives. It isn't only tasks getting done. It's how an organization stays aware of itself. Automate it without seeing it, and you don't just lose the tasks. You go deaf in the channel that used to warn you, and you find out at the worst possible time.
None of this is an argument against automating coordination work. Plenty of it is genuine toil, and some of the people carrying it would gladly hand off the grind. The argument is against automating it the way most companies are about to — as a line item, fast, without ever naming what it was doing. The deliberate version starts with one question the fast version skips: what judgment is threaded through this layer, and which parts of it should stay human? That's where the accountability and the early warning live. Answer it before you automate, not after. And credit the work while a person is still doing it — because if the first time you acknowledge it is the moment you remove it, that isn't recognition. It's a confession. The time you free up is capacity to reinvest, not headcount to book: AI mostly relocates labor rather than removing it, and this layer is where a lot of it lands next.
There's a simpler way to say the whole thing. Stop rewarding the announcement of authority over the fact of it — in who you promote, in what you automate, in what you're willing to call leadership. The people actually holding your organization together are rarely the ones telling you that they are. That is not a missing executive voice. That is what judgment looks like when it's too busy holding the thing up to perform. Automate it, or promote past it, without ever learning to see it, and you'll keep mistaking the man who says "I am the king" for the one who is actually running the kingdom — the person you hired to forge a path, and then, the moment it led somewhere you didn't expect, told to fall in line.