The Leaderboard That Watches

Salesforce, per 404 Media, has built an internal leaderboard that ranks teams by how much they use AI tools. Earn the right training badges and you ascend a three-tier hierarchy of corporate self-actualization: Champion, Innovator, Legend. Roughly 30 percent of the company has clawed its way to Champion this year. Around 15 percent made Innovator. Fewer than 10 percent are Legends. The leaderboard is sorted by executive, so each VP can admire the column of people beneath them, glowing or not.

And then there's the feature that gives the whole thing away. Next to the teams that haven't earned their badges, the dashboard offers a small, friendly prompt: click to see who 👀.

That eyeball emoji is the entire product. Everything else is decoration.

Here's the trick, and it's an old one wearing a new lanyard. You can change behavior two ways. You can argue people into it — make the case, win the mind, lose half the room. Or you can change the room. Move the furniture so the behavior you want becomes the path of least resistance and the behavior you don't becomes conspicuous. The second method works far better, which is exactly why it should make you nervous. A leaderboard doesn't tell you to use AI. It just arranges the social physics so that not using it costs you visibility, status, and a quiet line item in somebody's mental performance review.

Be precise about what's reported and what's projected, because the gap is where a piece like this usually cheats. What Salesforce built is a dashboard that shows who hasn't earned a badge. That's it — visibility, not yet a watchlist, not yet a firing. The menace isn't smuggled in by the feature; it's invited by the architecture. You don't have to call a leaderboard "surveillance" for it to behave like the first draft of one. The badges are real. The choice is theater. "Champion" is a compliance receipt with a gold star printed on it, and "click to see who" is the part where the receipts quietly itemize themselves.

Now the part that should actually worry a company, and doesn't. 404 Media reported earlier this month that Amazon killed its own internal AI leaderboard after employees, predictably, started gaming it — racking up tool-usage numbers that measured enthusiasm for the metric and nothing else. Goodhart's law, the oldest bug in the management codebase: the moment a measure becomes a target, it stops measuring anything but the desire to score. But the deeper failure isn't that the metric gets gamed. It's that by measuring AI use, the company optimizes for the appearance of adoption and destroys its own ability to learn the only thing worth knowing — whether any of this AI makes the work better. The leaderboard doesn't merely fail to answer that question. It blinds the institution to it, and files the blindness under progress.

And the question was never going to be asked, because the employees were never the audience. Look again at "sorted by executive." The number isn't built to be read by the people it ranks; it's built to be shown upward — to a VP, to a board, to a market that prices companies on how confidently they can narrate AI adoption. The leaderboard is a story executives are telling investors, and the workers are the props in it. The surveillance lands on them; the purpose sails over their heads. That's the real inversion: a measurement instrument aimed at the people inside, built to serve an audience that will never be in the room.

Which is why the employees' worry has a direction. 404 Media notes they're already nervous the rankings will bleed into performance reviews. Maybe they will, maybe they won't — that part is genuinely unwritten. But the architecture has a gravity to it. You don't build a dashboard, sort it by executive, and add a "see who hasn't earned the badge" button because you intend to never look. The structure leans somewhere, and everyone standing on it can feel the slope.

There's a version of behavior design that's beautiful — the quiet hush of a great library, where the room itself invites a certain kind of attention and nobody has to be told. Salesforce built the inverse. Same architecture, opposite purpose: instead of structure that serves the person inside it, structure tuned to the institution's story and pointed back at the person as a data point. The genius of environmental design is that it shapes you without a single command. The horror of it is exactly the same sentence.

So here's the slope, not the prophecy. The badges can become a column in the review software. The column can become a number. The number can become a reason, six months from now, that's already on the spreadsheet before anyone walks into the room. Nobody will have ordered you to use AI. You'll just have noticed, somewhere around the second quarter, that the people with the trophies kept getting to stay — and that no one ever checked whether the trophies meant the work got better.

The eyeball was watching the whole time. It just wasn't watching for you.