Don't Waste a Good Complaint
Turn friction into leverage.
The Sorcerer's Apprentice Problem
In Goethe's 1797 ballad, an apprentice enchants a broom to carry water, then discovers that he knows how to start the work but not how to stop it.
The spell works. That is the problem.
The broom carries water faster than the apprentice ever could. Then it keeps going. More buckets become more water, more motion, and eventually a flood because the apprentice delegated the labor without defining authority, limits, or an exception path.
That is a useful description of what many companies are doing with AI.
They accelerate proposals, analysis, client communications, software, and internal reporting. But the same consequential decisions still return to the CEO. AI does not remove the bottleneck. It feeds it faster.
At one federal services firm, roughly 85% of capture activity depended on one executive. Opportunity awareness, relationship coverage, competitor research, and qualification all converged on the same person. AI could produce more briefs and draft more outreach, but without clearer ownership it would only deliver a larger queue to the executive already carrying the work.
The company bought the broom. The CEO still owns every bucket.
Tasks were delegated. Judgment was not.
Founder dependency is often diagnosed as a time-management problem. It is usually an operating-design problem.
The work may have been delegated, but four things remain concentrated:
Context. The CEO knows why the last three similar decisions succeeded or failed.
Criteria. The team knows the CEO has a preference but cannot reliably apply it.
Authority. People are responsible for an outcome but still need permission to act.
Trust. Clients and employees have been trained to return to the founder when the situation becomes important.
The leader's attention had become infrastructure.
That works until growth creates more decisions than one calendar can absorb.
AI makes the hidden queue visible
AI can draft the proposal in twenty minutes. It cannot decide who may change the price.
It can summarize a client escalation. It cannot invent the firm's risk tolerance.
It can recommend a staffing move. It should not quietly acquire authority over careers, contracts, or commitments.
Without explicit decision rights, faster production creates a larger review queue. The organization celebrates local productivity while proposal cycle time, delivery speed, and management attention barely move.
That is the Sorcerer's Apprentice mistake: automating the motion without governing the consequence.
Your turn
Where does work sit longest in your firm: waiting for a decision, waiting for a review, or waiting for information?
Reply with the workflow and the role it waits on. I am collecting the patterns.
Break the spell
Start with one recurring class of decisions that repeatedly returns to the CEO.
Name the decision. “Client issue” is vague. “Approve an out-of-scope request below $10,000” can be governed.
Write the criteria. Capture the evidence, thresholds, precedent, and risks the CEO actually considers.
Assign the right. Name who may decide, who advises, and what must escalate.
Use AI to prepare, not impersonate, authority. Let it gather signals, compare precedent, identify missing evidence, draft the decision packet, and route exceptions.
Measure the queue. Track decision latency, queue age, CEO review hours, repeat escalations, and decisions resolved without CEO intervention.
The goal is not to remove the leader from consequential judgment. It is to stop consuming leadership judgment on decisions the organization should already know how to make.
AI creates management leverage when it helps the business remember, prepare, route, and learn. It creates management drag when every automated bucket still requires the CEO to say where the water goes.
Calculate your approval drag
If routine approvals keep returning to the same leader, six inputs can estimate the annual leadership hours and directional economic drag tied to approvals, exceptions, and reopened decisions.
If the calculator exposes one higher-stakes AI decision with no clear owner, the One Decision Clinic is available for a bounded fit review.
Directional estimate only. Keep confidential, regulated, personally identifiable, security-sensitive, and proprietary information out. Inputs stay in your browser unless you request a teardown.
Know a CEO whose attention has become infrastructure? Forward this issue to them. If someone forwarded it to you, subscribe here.
One Hard Thing
Run around a mountain.
The 2026 Ultra-Trail du Mont-Blanc starts Friday, August 28: 174 kilometers, 9,900 meters of climbing, three countries, and a 46-hour, 45-minute cutoff.
The hard thing is not going fast. It is continuing when the finish is still a country away.
One Song Worth Hearing
“The Day John Henry Died” by Drive-By Truckers, written and sung by Jason Isbell.
John Henry tried to beat the machine at its own game. That is the wrong AI contest. The human advantage is not producing faster. It is deciding what matters, where authority belongs, and when the machine should stop.
One Moment of Awe
On August 12, the Moon's shadow crossed Greenland, Iceland, and Spain, briefly revealing the Sun's corona.