Who Can Still Do the Job When an AI E-mployee Is Unavailable?
When an AI e-mployee is unavailable, only human staff who maintain regular manual practice can do the job, according to Dr. Jonah Tebaa. To prevent skills from fading within months, he proposes The Unplugged Hour, a three-step framework of Name, Shrink, and Run, where a named cover person and a deputy spend sixty weekly minutes performing live tasks by hand. In a composite illustration, this practice tracks the cover ratio metric to maintain manual capability.
Most companies that put an AI worker on a repeated task measure the time they get back. Dr. Jonah Tebaa, an AI strategist and business transformation consultant based in Lebanon, asks a different question: on the morning that worker is unavailable, who in the company can still do the job?
He calls the AI worker an e-mployee, meaning an AI given a defined role in a team. His argument is that such a role is only as dependable as the human who can still perform it, and that this human skill fades within months once the task is handed over, unless the company schedules a small, regular practice to keep it.
A Tuesday morning in a composite company
In his work, Dr. Tebaa illustrates the problem with a composite scene. A Beirut distribution company has its quoting e-mployee paused for a vendor change. It is ten past nine, customers are waiting, and the two people who once wrote quotes by hand are asking each other where the price-exception sheet is. He is explicit that the company and its numbers are illustrative, built from patterns he has seen, and that no real client is described.
The scene is meant to be ordinary. Power cuts, internet outages, vendor changes, model policy changes and paused accounts all produce the same morning. What differs from company to company is how long it takes to recover.
Why the saving has two halves
Dr. Tebaa argues that the usual savings figure counts only the first half of the benefit, the hours returned to the team. The second half is the cost of keeping someone capable of taking over. A company that records only the first, he says, is quietly borrowing against the second, because the hand skill and the unwritten exception rules move out of people's heads and into the tool's setup.
To make that visible he proposes one measure, the cover ratio. It is the number of items people can finish by hand in a day, divided by the number the AI worker finishes normally. In his illustration, a practised pair writing quotes by hand covers 70% of a 40-quote day. After six months without practice, the same pair covers about 33%. A three-day pause then leaves 81 quotes waiting instead of 36. He asks readers to redo the arithmetic with their own figures rather than adopt his.
The framework: The Unplugged Hour
His remedy is a weekly practice with three steps.
- Name. Each e-mployee role gets one named cover person and one deputy, written on the role card.
- Shrink. The owner writes a one-page minimum version of the task, with a limited customer list, a standard price list and no exception handling, so that a person can finish it at reduced volume.
- Run. Each cover person does the real task by hand for 60 minutes a week on live work, with a timer on. Gaps are noted, the one-page sheet is updated, and the hour is logged.
In his example the practice costs two hours a week across two people. It keeps cover near 60% and cuts the three-day backlog to about 48 quotes, which is 33 fewer waiting than in the unpractised case.
What he cautions against
Dr. Tebaa is careful about what the hour is not. It is not a fire drill, because unannounced shutdowns teach fear rather than skill. It is not a punishment, because a slow first week is information about the gap. And it is not a second team, since the aim is a smaller version of the job, not duplicated output.
He also advises rotating the deputy so a single resignation does not remove the cover, setting the cover target by the role (he suggests 50% or more where the work touches revenue, and about 25% for back-office work), and dropping the practice altogether where a task can safely wait. The measure of success, in his view, is not the number of hours logged but whether the cover ratio meets the target the owner set.
Where it fits in his wider work
The idea sits alongside his other writing on the e-mployee operating model. His two-week ramp plan deals with bringing an AI worker into a team at the start. The Unplugged Hour begins after that, once the worker is running and the temptation is to stop training anyone on the task. His practical advice to owners is to choose the one role whose pause would hurt most, name two people, book the first hour in their calendars this week, and calculate the first cover ratio.
Related evidence: NIST's contingency-planning guide sets out how organisations evaluate an information system and its operations to determine what it takes to keep the service available when a dependency fails, and ties that planning to organisational resiliency and the system development life cycle. (NIST's contingency planning guide for information systems)
Google's SRE book describes how a system with insufficient capacity for its steady-state load saturates its queues, so latency and error rates rise at the point of measurement while the actual cause is contention somewhere upstream. (Google's SRE book on saturated queues and cascading failure)