Playbooks

An expert talking over slides is a podcast with worse acoustics — “interesting” is the tell. One decision instead of a topic, one practice rep in the room, one artifact that lives where the work does, and the twenty-minute SME prep that makes it repeatable.

Thirty days out, the course is a headline and a vibe — on schedule, per the oldest curve in learning science. Why re-reading can't fix it, why retrieval can, and the scenario-a-week plan with spaced rotation that runs on a calendar and a chat channel.

OSHA, state harassment mandates, your license, and the training clauses buried in contracts you’ve already signed — nobody hands you the list, so here’s the four-layer assembly method, plus the one-page register that proves it all happened.

Don’t ask the expert to design the training — mine her for the judgment she can’t articulate on demand. Recent cases instead of generalities, the hard calls instead of the happy path, and a 90-minute session that replaces a month of review cycles.

Training fixes can't-do — it has never once fixed won't-do or can't-because-the-system. Four diagnostic questions that sort skill gaps from system gaps in five minutes, the prescription for each branch, and how to annotate instead of gate when the sponsor wants a course anyway.

"Shadow Sarah" isn't training — it's hoping proximity works. The five-day build that turns tribal knowledge into a structured first month: define the ten tasks, capture the SME while she's still explaining, draft with AI against real artifacts, build scenario checks, and schedule it spaced.

Manager coaching at an SMB is either nonexistent or a one-time doc nobody revisits — good intentions die in week one because nothing was built to carry them. Turn coaching into a living plan: one behavioral goal, a few concrete actions, a review cadence, and evidence of progress that survives the quarter.

SMB training doesn't fail for lack of knowledge — it fails at the blank page, where the person who knows the content is asked to also be an instructional designer, and freezes. Describe the training in a sentence, let AI draft the structure you don't have, and spend your judgment editing for truth, context, and outcome.

A spreadsheet of hand-typed checkmarks proves you typed a checkmark, not that a person completed a training. You don't need an enterprise LMS — you need the narrow thing underneath compliance: assign to named people with deadlines, track as a live picture, and hold evidence you can hand over.

"Completion rates and a smile-sheet average" isn't an answer to "what is the training budget buying." The minimum-viable measurement stack — the behavior sentence, a performance check, the 30-day pulse, and one honest metric — that measures the levels that actually predict, on a budget of roughly zero.

A completed course used to weakly imply learning — AI vaporized that signal. The two kinds of evidence that survived (alignment and unfakeable performance), why "you don't need unfakeable, you need better than a checkmark," and how a team of one measures real cognition.

Asked to "make a training," your instinct is to open slides and start writing — which is how you get forty slides nobody remembers. Build the outline backward from the outcome: the six-step method that produces training people actually use.

The six systems an L&D team of one actually needs — needs analysis, design, production, delivery, evaluation, reinforcement — built on Backward Design, ADDIE, and Kirkpatrick, written for the practitioner.