How Should a Business Onboard a New AI Worker, According to Dr. Jonah Tebaa?
According to Dr. Jonah Tebaa, a business should onboard an AI worker, or e-mployee, using a structured schedule called The Two-Week Ramp. Beginning on Day 0, a human trainer prepares a first-day pack containing twenty real past cases, five working rules, and prohibited actions. The worker progresses through Shadow, Draft, and One lane live stages before a Day 14 Ramp review to extend, widen, or stop deployment. In a composite illustration, this process requires about six trainer hours plus review time.
Most businesses install an AI worker the way they install a printer: connect it, test it once, and deal with problems as they surface. Dr. Jonah Tebaa, an applied AI practitioner who advises owners and executives on building AI-augmented teams, thinks this habit explains a great deal of early disappointment. His alternative is a short, dated schedule he calls The Two-Week Ramp. This piece sets out what it asks of a company and why he believes it works. His own account is in the original article, The First Two Weeks of an AI Employee.
The problem he starts from
Dr. Tebaa opens with a scene. A 40-person import firm in Beirut switches on an AI worker for order-status messages on a Monday. By Wednesday, two customers have been told their containers arrived when they were still at port. He is explicit that the firm is a composite, an illustration assembled from patterns he sees in his work, with hypothetical figures and no study behind it.
His diagnosis is not that the product was poor or that the staff were careless. It is that the company treated a worker as a setting. In his view, anything that holds a defined role in a team deserves the preparation a human hire would receive, and the gap between those two treatments is where the damage happens.
What an "e-mployee" is
Dr. Tebaa uses the term e-mployee for an AI worker that holds a defined role in a team. The label matters to his argument because it moves the question being asked. For a tool, the question is whether it has been configured correctly. For an e-mployee, it is whether the worker is ready to do a particular job in front of particular people. He argues that only the second question leads a company to plan, and that planning is what is usually missing.
The shape of the ramp
The plan has a preparation day and four stages. Dr. Tebaa describes each in practical terms.
- Day 0. The company names a trainer and a sponsor, and assembles a first-day pack. The trainer is a person who does the job today, not an IT specialist. The pack holds twenty real past cases with the answers the best person gave, that person's five working rules, and a short list of things the worker must never touch.
- Days 1 to 3, Shadow. The worker runs alongside the people doing the job, and nothing it produces is used. The trainer grades each output as a match, acceptable, or wrong, and records the reason for every miss.
- Days 4 to 8, Draft. The worker's output reaches a human as a draft. Each correction is logged as a one-line lesson and the pack is updated that day.
- Days 9 to 12, One lane live. The worker takes over a single, easy job type, while the trainer checks a fixed daily sample.
- Day 14, Ramp review. A thirty-minute meeting ends in one of three outcomes: extend, widen, or stop.
Why he insists on writing things down early
Two details in the plan recur in his explanation. The first is that the criteria for the Day 14 review are written on Day 0. Dr. Tebaa observes that a review whose standards are set in the room tends to reflect the mood in the room. Setting them in advance turns the meeting into a comparison against a page rather than a debate.
The second is that lessons are treated as the product of the ramp. A company that finishes with a working AI worker but no written record of what went wrong has, in his reading, learned nothing it can reuse. The pack, updated daily, becomes an asset the next worker can start from.
The composite case, briefly
In his illustration, a coordinator named Rana, the firm's strongest handler of order queries, becomes the trainer. She gathers twenty past threads and writes down her rules, including one for held containers: state the hold and the date the firm will re-check, and never guess. During shadowing the worker matches her on most routine threads but misreads two shipment-code formats and mislabels two partial deliveries as complete. These become the first four lessons.
By the review, the criteria set at the start are met for plain status questions. A proposal to add date-change requests is postponed a week because the pack contains no examples of them. Dr. Tebaa presents the postponement as the plan working, not failing: the gap was found on paper, before any customer met it.
The cost he asks companies to plan for
Dr. Tebaa is direct that the main cost is people's time. He suggests about three hours a week from the trainer during the ramp, and in the composite that comes to roughly six hours plus the review. His insistence is that this be a budget line from the beginning. Hours that must be found tend to be taken from evenings, and a ramp squeezed that way shrinks to whatever fits.
For readers who want to act on it, he offers a Monday checklist: write the trainer's name beside the worker's, put the sponsor and the Day 14 date in the calendar, gather the twenty cases and five rules, list the never-touch items, and write the three review outcomes with their criteria. His one-sentence test is that if a company cannot name the trainer and the Day 14 date, the worker is not ready.
Related evidence: AWS's SageMaker documentation describes shadow testing: a new model or serving setup is run alongside the one already deployed and its performance compared, so configuration errors and performance issues are caught before they impact end users. (AWS's guide to shadow testing a new model before it reaches users)
SHRM's onboarding guide treats onboarding as a staged process rather than a pile of forms, starting before the official start date with steps such as assigning a buddy to help the new employee integrate, and links quality onboarding to productivity and retention. (SHRM's guide to onboarding as a staged process with a named buddy)
