The One-Week New-Hire Training Build: From Nothing to a Real First Month

The offer's accepted, the laptop's ordered, and someone finally asks the question everyone was avoiding: "What's the training plan?" And the answer — at most small companies, for most roles — is a slightly embarrassed version of "they'll shadow Sarah." Sarah, who has a full-time job. Sarah, who has never once been asked to explain how she does what she does, only to do it faster.

Shadowing feels like training because it looks like proximity to knowledge. But watching a competent person work is mostly watching the parts that are visible — and competence is mostly the invisible parts: what she checks before she acts, what she'd never do and why, which of the four ways to handle the situation she picked and what told her to. The new hire sees the keystrokes and misses the judgment. Six weeks later you're explaining the same things for the fourth time and wondering if you hired wrong. You didn't. You just never built the thing between "hired" and "competent."

Here's the good news: for one role, that thing is a one-week build. Not a curriculum project, not an LMS purchase — five focused days, a few hours each, most of it work you can delegate to AI once the thinking is done. Day by day:

Day 1 — Define the ten tasks that are the job

Ignore the job description; it was written to attract, not to train. Instead, answer one question: what does this person need to be able to do, unassisted, at the end of month one? List actual tasks — "process a refund end to end," "qualify an inbound lead," "close the books on a client month" — and cap the list at ten. If you write twenty, you're describing month six. Then rank them by frequency times pain: the tasks done daily, and the tasks that are catastrophic when botched, float to the top. This list is the entire skeleton of the training. Everything you build this week exists to produce those ten competencies — and anything that doesn't serve one of them is decoration, cut without mercy.

Day 2 — Capture the expert while she's still explaining

Now Sarah earns her mention — not as scenery to shadow, but as a source to mine. Book her for ninety minutes, recorder on, and walk the top tasks with three questions per task: Walk me through the last time you did this. What do you check before you start? What goes wrong, and how do you know early? Those questions surface the judgment shadowing never shows — the pre-checks, the smells, the unwritten rules. You're not asking her to design training; experts are famously wrong about what beginners need taught. You're collecting raw material. Ninety minutes of transcript beats nine weeks of osmosis.

Day 3 — Draft the modules with AI, against real artifacts

This is the day that used to take a month and now doesn't. For each of the ten tasks, you want the same four-part module: what this task is and when it happens, how to do it step by step, the judgment layer — what to check, what goes wrong, when to stop and ask — and one worked example using a real artifact: an actual (sanitized) ticket, invoice, or lead, not an invented tidy one. Feed the transcript and your task list to AI and let it produce the first draft of all ten; that's exactly the drafting labor machines are for. Then do the pass that keeps you honest — you review every module against one test: would Sarah wince at this? The wince-points are where the AI smoothed over a nuance the transcript only hinted at. Fix those; ship the rest. Drafting is delegated. The wince test never is.

Day 4 — Build the checks that prove it stuck

A module without a check is a hope. For each task, write one scenario check — not "list the steps of the refund process" but "here's the refund request in the queue; handle it, and tell me the one thing about it that should have made you pause." Recall quizzes measure memory; scenarios measure whether the judgment layer landed, and the judgment layer is the whole reason you built this instead of pointing at Sarah. Set the bar at demonstration, not attendance: a task is trained when the new hire has done it, unassisted, on a real or realistic case, and someone confirmed it. That confirmation — a checklist with initials, nothing fancier — becomes your competence record, and it's the difference between "training happened" and "training worked."

Day 5 — Assemble the month, not the firehose

The final discipline is scheduling. The instinct is to front-load — two intense days of everything, then "let them get to work." Cognitively that's a bonfire: day-one material is gone by Friday without reinforcement. Instead, spread the ten modules across four weeks in order of the Day-1 ranking — week one gets the daily-frequency tasks, week four gets the rare-but-critical ones — and pair every module with same-week practice on the real thing. Each Friday, fifteen minutes: what was trained, what was demonstrated, what's next week. Four weeks, ten demonstrated competencies, zero "go shadow someone."

Where the tool fits

Everything above is doable with a recorder, a doc, and a disciplined week — and the first time, I'd genuinely have you do Day 1 and Day 2 by hand, because defining the job and mining the expert are where your judgment lives. Days 3 through 5 are structure and drafting, which is LearningByDesign's home turf: it turns a topic and a transcript into structured modules with measurable objectives, generates the scenario checks alongside the content instead of as an afterthought, and assembles the spaced first-month plan so the sequencing discipline survives contact with a busy calendar. You review against the wince test; it does the typing. The next role's build takes an afternoon instead of a week.

The bottom line

"Shadow Sarah" persists because it costs nothing up front — and then quietly costs you the same explanations four times over, a slower ramp, and occasionally a good hire who concluded the chaos was the culture. One week buys the alternative: ten defined tasks, the expert's judgment captured while it's explainable, modules drafted against real artifacts, scenario checks that prove transfer, and a first month with a spine. Build it once for the role, not the person — the second hire into that seat walks into training that already exists, and Sarah gets her job back.

— Tom