A fictional exploration of how measures can inherit the legitimacy of the problems they were created to address, even after they stop establishing whether those problems are being solved.
You can see it as a continuation of When the Moat Becomes the Product
Observation
Marcus had rehearsed the pitch for a week.
The workshop was free. The results were documented. The marketer had generated millions of views with an AI clone doing most of the work.
It was elegant. Scalable. And it solved the production problem entirely.
He opened his laptop.
“I found something that could actually move the needle,” Marcus said. “An AI-clone workshop. You build a digital version of yourself – Brain, Image, Voice and Video – and it produces content under your name. Your presence multiplies. You barely have to touch any of it.”
Richard, who had run the company for eight years after twenty-three years in the British civil service, looked at the screen without moving.
“So let me understand this clearly,” he said slowly. “You want to build an AI version of me that can publish continuously.”
“It would multiply your presence.”
“And if it works, what exactly do we still need you for?”
Marcus paused.
“That isn’t the intention.”
“I’m sure it isn’t.”
Richard looked back at the presentation.
“Revenue is down four percent. How does this fix that?”
Marcus had anticipated resistance.
“There’s a lag between visibility and conversion. The metrics show…”
“Stop.”
Richard held up a hand.
“I’ve tried. Genuinely. I’ve tried to learn your language. KPIs. Indexation. Semantic search. AI-readable pages. I’ve read your explanations. I’ve sat through your presentations.”
He leaned forward.
“And there’s one question I still can’t get answered.”
Marcus waited.
“How does any of this connect to getting the right people to notice us and what we do?”
“The metrics show people are finding us.”
“Which people?”
Marcus hesitated.
“People searching for the terms we rank for.”
“Are they the people who need what we do?”
“We don’t track it at that level.”
Richard nodded.
“Then what do we know?”
Marcus glanced at the screen.
“That we’re visible.”
“Visible to whom?”
The question remained between them.
Richard sat back.
“You know what I notice? Revenue. Revenue means someone gave us money because they wanted what we’re selling. Revenue isn’t the whole truth. But it’s an outcome we were trying to influence, and none of your metrics has explained why it isn’t moving.”
Marcus looked down at his notes. The pitch no longer felt as solid as it had earlier that morning.
Structural Tension
Richard turned the laptop slightly toward Marcus.
“Walk me through what happened.”
Marcus began with the strategy.
They had improved their search rankings. Restructured their pages for AI search systems. Increased indexation. Produced more content. Tracked engagement, bounce rates and click-through rates.
Each stage had made sense when they introduced it.
Each had produced measurable improvement.
“And now,” Richard said, “you want to automate the production of more of it.”
“Not just more. Better distribution. More consistency. The clone can sound like you. It can publish at three in the morning. It can maintain a presence even when you’re occupied with other things.”
“Can it tell us who needs to hear what we’re saying?”
Marcus looked at him.
“It can help us reach them.”
“How do we know?”
Marcus opened his mouth, then looked at the dashboard.
“We’d track the response.”
“Which response?”
“Engagement. Clicks. Conversions.”
“Those are measures.”
“Yes.”
“Of what?”
Marcus felt irritation rise in him.
“We need measures, Richard.”
“Of course we do.” Richard’s voice remained level. “But a measure doesn’t become an answer just because we can put it on a dashboard.”
He looked at the green charts.
“There’s a story my mentor Clarence told me when I was young in the civil service. The cobra bounty.”
Marcus had heard the expression before, but never in this context.
“Delhi, 1870s. Too many cobras. A British administrator creates a bounty. Pay for each dead cobra. Simple. Measurable. An obvious connection between the problem and the solution.”
Richard paused.
“People start killing cobras. Bounties go up. The numbers look good. Then someone works out that breeding cobras is cheaper than catching wild ones. So they breed them, kill them and collect the bounty.”
“And the population of cobras?”
“Didn’t decrease. Could have increased. But the measure kept improving.”
Richard looked at the screen.
“Dead cobras up. Bounties paid up. System working beautifully.”
Marcus said nothing.
“The measure had inherited legitimacy from the reason it was created. Solve the cobra problem. But eventually nobody stopped to ask whether it was still solving the cobra problem.”
Richard turned back to him.
“That’s what I’m asking you.”
Contrast
Marcus looked at the presentation deck.
The workshop had seemed like an answer because it addressed something tangible.
- They had not been producing enough.
- They needed greater visibility.
- Their competitors were publishing more frequently.
- AI systems were changing how information was discovered.
- Everyone was talking about content scale.
The clone offered a way to respond.
But the original question had been different.
They had something to say. And they needed the right people to notice them, to hear it.
At some point, the question had shifted.
- Could people find them?
- Could they rank?
- Could AI systems understand and retrieve their pages?
- How much content could they produce?
- How much of that production could be automated?
- Each question had been reasonable.
None had established that the answer to the previous question had solved the original problem.
Ranking was not attention. AI readability was not attention. Content volume was not attention.
And an AI clone producing ten thousand pieces of content did not create ten thousand units of human attention.
It created ten thousand pieces of content competing for the attention that had been scarce in the first place.
The organization might solve the production problem while deepening the discovery problem.
And the dashboard might not show the difference.
Marcus looked up.
“So what are you saying? That we shouldn’t use the clone?”
“No.” Richard’s answer came quickly enough to surprise him. “I’m saying that we shouldn’t confuse the ability to produce something with hope that producing it will solve the problem.”
He pointed at the screen.
“Your rankings may matter. Your content may matter. The clone may even help. But you keep presenting the measures as though they have already established the connection.”
Marcus looked back at the green columns.
He had been treating each improvement as confirmation that the strategy was working.
But confirmation of what?
Structural Principle
“You’ve been optimizing rankings,” Richard said, “because rankings once correlated with attention. That’s why they were a useful proxy.”
Marcus nodded slowly.
“But you’ve optimized the proxy so intensely that you no longer know whether the original relationship holds.”
He let the sentence settle.
“The measure inherited the legitimacy of its original purpose long after it stopped establishing that purpose.”
Marcus looked at the dashboard.
“Then what should we measure?”
Richard gave a small, tired smile.
“You can build another proxy.”
“That isn’t very helpful.”
“No. It isn’t.”
Richard folded his hands on the desk.
“That’s the difficulty. We need measures. We can’t run a business without them. But every measure creates the possibility that we will start optimizing it instead of the thing we care about.”
He glanced at the laptop.
“The question isn’t whether we can measure attention perfectly. We probably can’t. The question is whether we keep asking whether our measures are still telling us something useful about it.”
Marcus thought about the AI clone. A digital version of Richard could speak in his voice, publish his opinions and maintain a constant presence. It could create the appearance of activity at a scale no human could sustain.
But it could not, by itself, establish that the right person had noticed, understood or cared.
The clone could multiply the signal.
It could not guarantee that the signal mattered.
“That’s a harder problem,” Marcus said.
“Yes.” Richard stood. “That’s probably why you kept solving the easier ones.”
Closing Observation
Marcus closed his laptop.
At the door, he paused.
“What did Clarence do when he realised the measure had become the target?”
Richard considered the question.
“He asked the original question again.” Richard looked at the presentation deck, with its green charts and rising metrics.
“He kept asking it after everyone thought it had already been answered.”
Marcus waited.
“Did it get easier?”
“No.”
Richard turned back to the spreadsheet.
“You get better at noticing when you’ve stopped asking.”
Marcus left.
Richard closed the door and sat down.
Revenue down four percent.
Every other column green.
He opened a blank document and typed a single line.
What are we actually solving?
He looked at it for a while.
Then he opened the revenue spreadsheet again.
(Not written by an AI clone)
