This Week's Term: AI Theater - the performance of AI adoption without the substance behind it.
A grand keynote, a flashy demo, a pilot that never reaches production, or an internal usage leaderboard. These are activities that signal progress on AI while producing little or no real value.
The term has a respectable lineage. Bruce Schneier coined "security theater" in 2003 for measures that make us feel safer without actually making us safer. Steve Blank borrowed the structure with "innovation theater," the hackathons and design-thinking workshops that build a culture buzz but, as he put it, "rarely deliver shippable product." In 2026, AI has earned its own version of the word, and you now see "the end of AI theater" written across the business and technology press as the year's defining theme.
The wasted budget is the smaller problem. The bigger one is what AI theater does to trust. It erodes confidence in the genuinely useful work happening alongside the spectacle. Once people have sat through enough demos that went nowhere, they start discounting the next one too, including the demos that would actually have paid off.
For leaders, the practical question is how to spot your own theater. If the activity is measured by inputs rather than outcomes, you should be suspicious. Tokens consumed, demos delivered, pilots launched, or slides presented are your warning signals. Real value shows up as something a customer or a colleague can feel: a resolved ticket, a faster decision, a recruiter who got an hour back to create new value. If you cannot draw a straight line from the AI activity to an outcome someone would care about, you are probably watching a performance.
The cure is the same move I describe in this issue's piece on speaking the language of your audience. Show someone a visible before-and-after on a task they actually care about, and there is nothing to perform. They feel the value within the minute, and that is the whole point.
Your action step
Take your current flagship AI initiative and write down how it is being measured. If every metric is an input, tokens, pilots, demos, sessions, replace at least one with an outcome a customer or colleague would actually notice. Then check back in a month and ask whether the line from activity to outcome got shorter or longer.
For a sharp, skeptical take on how much of the current AI spectacle is recycled hype dressed as breakthrough, AI analyst David Linthicum's video is a good reminder, and it pairs well with this term.