What does Dr. Jonah Tebaa's Rebuild Ledger for Jobs an AI E-mployee Has Half Emptied mean in practice?
Dr. Jonah Tebaa's Rebuild Ledger redesigns roles partly absorbed by an AI e-mployee across five steps: listing paid task hours, marking automated and checking duties, pooling what remains across roles, clustering it by outcome, and assigning freed hours. Each rebuilt job must satisfy a 24-hour size floor, pass a judgement test requiring a call a customer or supplier will notice, and survive a trace test. In a hypothetical trading firm, the 62 remaining hours are regrouped so a customer dispute that once touched three clerks now sits with one.
A task-hour ledger for jobs that AI has partly absorbed
Dr. Jonah Tebaa argues that an owner whose AI e-mployee has taken part of several roles faces a design problem, not a staffing one. His instrument, the Rebuild Ledger, treats the work that remains as a single pool to be regrouped, and it tests each regrouped job before anyone is matched to it. The method is his own working practice, and he does not present it as research.
The ledger runs in five steps. Each person lists their weekly tasks in hours that sum to their paid week. The firm marks which tasks the AI e-mployee now performs and adds the new work that its output creates, chiefly checking, with hours attached. The remaining human tasks are pooled across roles and tagged by the outcome they serve. They are then clustered by outcome rather than by old title. Finally, the firm names a use and a date for the freed hours before it touches a single job title.
The two tests a rebuilt job must pass
Dr. Tebaa attaches two tests to every cluster. The first is a size floor, which he sets at 24 hours, or 60 percent of a 40-hour week, and which he insists on fixing before the clusters are examined. A cluster that falls under the floor is merged with another one. The second is a judgement test: the job must contain at least one task where its holder makes a call that a customer or a supplier will notice. A cluster made only of keying and filing does not pass, in his view, and should be looked at again before it is called a job.
He adds a third check once the new design exists, which he calls the trace test. Follow one real case through the firm and count how many people it touches. If a single customer problem still crosses three desks, the grouping is wrong, however neat the hours look on paper.
The hypothetical he uses
His worked example is a hypothetical, with invented numbers, and he labels it that way. A trading firm has three clerks at 40 hours each, 120 in all. An AI e-mployee takes 67 hours of their tasks, which leaves 53. Checking the AI's output adds 9 hours, so human work now totals 62 hours and 58 hours are freed.
Pooled by outcome, the 62 hours fall into three clusters:
- Customer, 26 hours: key-account calls, order exceptions, negotiation with key buyers and dispute resolution, plus checking of quotes and orders.
- Cash, 17 hours: payment matching and month-end reconciliation, plus checking of invoices and reminders.
- Supply and price, 19 hours: stock chasing, the daily stock count, price-list updates and supplier price confirmation.
Only the Customer cluster clears the 24-hour floor. The Cash and Supply-and-price clusters are merged into one job of 17 + 19 = 36 hours. The Customer job then takes 14 hours of proactive account reviews to reach 40 hours. The merged job keeps 4 hours of slack for the month-end peak, and a new 40-hour export desk, which the firm had been unable to staff, absorbs the remaining capacity. Of the 58 freed hours, 54 are given a named use and 4 stay unassigned as that slack. The same dispute case that once touched three clerks now touches one.
What he says the ledger does not do
Dr. Tebaa is plain about the limits. The ledger does not claim a saving, and the freed hours are a design input, not a benefit. If an owner prefers to cut hours or headcount, he treats that as a separate decision which the ledger neither makes nor recommends. The hours, the floor, the 14 hours of reviews and the export desk in his example are assumptions chosen to show the method, and real firms will have untidier numbers.
He also advises consultation. People in the affected roles should see the ledger and correct their own task lists, and employment terms should be checked before any role changes, with particular care in Lebanon, where contracts and labour rules apply.
When he would run it
The ledger earns its place at two moments, in his account: just after an AI e-mployee has taken meaningful work from several people, and just before a vacancy is filled or a team reshuffled. Because it lists outcomes and decisions rather than titles, it also serves as a work map for a replacement hire. A shorter walk-through of the same method, rebuilding the jobs an AI e-mployee left half empty, covers the pooling step on its own.
Dr. Tebaa sets out the full first-person version, with the five steps and the worked ledger, on his own site. In his own words: Rebuild the Jobs an AI E-mployee Left Half Empty.