The scarce resource is attention

Output is getting cheaper. Attention is getting more expensive.

Emad Mostaque’s The Last Economy is useful here because it points at the operating constraint that remains after computation gets cheap.

Most teams still measure AI progress the old way: more drafts shipped, more tickets closed, more replies sent, more research summarized, more tasks touched per person. Those gains are real, but they are easy to misunderstand.

Once computation can produce endless text, options, reminders, and recommendations, raw output stops being the scarce input to a business. The scarce input becomes human attention: who reviews the exception, who decides which edge case matters, who notices the customer risk, who changes the rule, and who turns one good answer into durable company memory.

That shift should change how an owner designs the system.

Cheap output can still waste the owner

A messy business can now generate faster mess. An agent drafts the estimate, rewrites the follow-up, summarizes the inbox, classifies the support queue, suggests the next action, and prepares a daily report. On paper, the team looks more productive. The owner may now be reading more low-value summaries, clearing more soft escalations, and revisiting more work that was produced cheaply but not governed clearly.

The constraint moved. The owner did not.

If each new automation path creates another review surface without reducing ambiguity, the company has not created leverage. It has created attention debt.

Spend human focus where it changes the company

An owner system should ask a harder question than “Can AI do this step?” It should ask where human attention still improves judgment or durable state. In most businesses, that means exceptions that affect customer trust, money, compliance, or scope; priority calls that change the business; changes to the playbook, scorecard, or pricing rule; coaching an apprentice through reasoning they will need again later; and review of whether a new pattern deserves to become shared memory.

Everything else should be pushed toward clearer defaults, better evidence packaging, or full automation.

Many AI rollouts fail here. They automate the visible task but preserve the hidden uncertainty underneath it. The human still has to reconstruct context, guess which output matters, and rescue the workflow when the rule boundary is unclear.

The result is not an owner-making system. It is a faster claim on the same person.

Hadto’s thesis depends on attention discipline

Hadto is not trying to help companies produce more undifferentiated output. It is trying to help more people become capable owners, which requires a different standard for AI.

The point of automation is not to keep a technician busy with more throughput. It is to free attention for the work only an owner-operator can do well: define standards, govern handoffs, inspect exceptions, improve the customer promise, and decide what the business should remember.

A future owner should inherit a system that already knows the normal path. They should not inherit a dashboard that produces infinite drafts and asks them to babysit all of them.

When AI is used well, the operator spends less time re-reading, re-explaining, and re-deciding. They spend more time judging the few decisions that matter and improving the rules behind them.

One practical rule

Every new AI workflow should have an attention test. Before keeping it, the owner should be able to say three things clearly:

• what human judgment this workflow is protecting
• what durable record or rule it improves when that judgment is applied
• what attention it no longer needs from the owner, manager, or senior operator

If those answers are weak, the workflow is probably creating churn instead of leverage.

Cheap output is not the goal. Better use of scarce human attention is.

The businesses that benefit most from AI will not be the ones that generate the most. They will be the ones that reserve human focus for meaning, connection, judgment, and durable state change, then build the rest of the work so the system can carry itself.

Used that way, AI creates more capable owners instead of just more output.

---

Source context for this note: Emad Mostaque, The Last Economy, especially Chapter 21 on attention as the remaining scarce resource after cheap computation, reviewed alongside the completed Hadto study notes on 2026-04-23 and checked against the current Hadto blog archive on 2026-05-08 to avoid repeating earlier posts on visible judgment, playbook promotion, and owner-operator system design.