Gartner, "Gartner Predicts AI Inference Costs Per Agentic Workflow Will Increase More Than Fivefold Through 2028," press release, August 17, 2026

There is a sentence that closes a conversation, and I have sat through it enough times now to feel it coming before it lands. Somebody proposes a way of working. A person with more authority than the proposer says the company tried that once and it did not work here. The room agrees, because the room has no evidence to the contrary and no appetite to go looking. The subject dies. What follows is a search for something else to try, which in three or four years will earn the same sentence from somebody else.

I stopped taking that sentence at face value after a graduate module on counterfeit goods, which is not where I expected to find a leadership idea. The obvious harm from a counterfeit is the revenue it takes from the company that made the real thing. The more interesting harm is what it does to the original's reputation. Somebody buys the fake, believing it is genuine, and it comes apart in a month. They do not conclude that they bought a fake. They conclude the brand is junk, and they tell people. The counterfeiter took one sale. The damage he did to the name outlasts him.

Most organizations that have sworn off agile never did agile. They did the visible parts, which is all a counterfeit ever is. The daily stand-up became a status meeting where people report upward to a manager who is keeping score. The sprint became a two-week deadline carrying the same fixed scope somebody committed to a quarter earlier. The backlog became a wish list nobody would order, because putting things in order means telling a senior person their item is fifth. The retrospective became a meeting where people learned which complaints were safe to voice.

Every one of those has the outline of the genuine article and none of its function. Run that arrangement for eighteen months and you get exactly the results the arrangement deserves. Then somebody senior looks at the results and draws a conclusion about the method, and the conclusion travels much further than any of the details would.

The blame lands on the category, which is what makes the failure so comfortable for everyone still employed. Nobody has to answer for having installed an imitation, because the method took the fall on their behalf. A leader who spent two years and a consulting budget on ceremonies gets to describe himself as a pragmatist who tested something and found it wanting. He is not lying, in his own view. He genuinely tried the thing he thought it was.

None of this is really about agile, which is a twenty-year-old argument I would rather not have again. Look at what has happened to OKRs, which in most companies are last year's annual goals typed into a different template. Coaching culture arrives as a mandatory quarterly conversation with a form attached to it. Psychological safety, in more places than I can count, is a survey question and not one observable change in what happens to whoever speaks first. Digital transformation has been a procurement exercise for so long that the phrase now means buying software. Each has a cheap version that installs the artifact and leaves the machinery alone, and each cheap version is busy spending down the credibility of the real one.

So the useful question is how you tell which one you own, and I think there is a single test that works across all of them. Ask what changed about who decides.

Not what meetings appeared on the calendar, and not what vocabulary the organization picked up. Every one of these methods, underneath the branding, is a proposal about moving a decision closer to the people holding the information. Agile moves the question of what to build next toward the team and the customer. OKRs move the question of what counts as success down from the executive who set it. Coaching moves the question of what a person needs from the annual form to the person. If you installed any of them and nobody's authority changed hands, you did not implement it. You bought the costume and wore it to work.

That test also tells you something uncomfortable about why the counterfeit is so popular, because the counterfeit is the version that survives contact with an organization unwilling to move authority. The ceremonies are adoptable precisely because they cost nothing that anyone in charge values. The decision rights are the expensive part, so they are the part that gets left out, and what remains is a set of practices that look like a method and function like a schedule.

Everything I published this week was a version of this. A safety form standing in for the safety to speak up. A frozen job market standing in for loyalty earned. A return review aimed at a delivery team when the value question was settled a year earlier and two floors up. A person named as human oversight who never sees what the machine actually did. Long hours standing in for the short decision loop that made long hours worth working somewhere else. In every case the visible object was purchased and the mechanism underneath it was assumed to have come along.

The counterfeit is the same failure with a longer tail. When a form fails, you can replace the form. When an organization runs a fake version of something and concludes the genuine article does not work, it is inoculated against the real thing for years, and the people who would benefit most from it are the ones who will never get to try.

If you are about to tell somebody your company already tried a thing, it is worth asking yourself what you would have to produce to prove it. Name the authority that moved. Name the person who could decide something in the second year that they could not decide in the first. If nothing comes to mind, then whatever you tried, you did not try that.