Start here
The 1 AM Test
Every training artifact answers one question: does this help someone who is stuck right now, alone, at 1 AM?
Users tell you how they want to learn. They email asking for materials. They ask for "anything you can send me." They want the answer at the moment they are stuck, not an appointment next week at a time convenient for a trainer.
The library is how you say yes to all of them at once. The economics are settled (the full arithmetic lives in the Rollout Playbook): self-paced formats retain more, cost less per learner at any scale, and carry near-zero marginal cost. What the economics don't tell you is how to make library content people actually use. That craft is this cookbook.
One field result frames everything that follows. In one deployment, two workflow videos, built in a day or two, began producing fully SOP-conformant requests from users who had never booked a training session. Asked how they knew the process, the answer was always the same: "someone shared the video." Content built right distributes itself.
A training calendar reaches the people who sign up. A library reaches the people they forward it to.
Part I · The Recipes
Three Formats, Three Disciplines
Every artifact is one of these. The discipline in each recipe is what separates a library from a folder of files.
The scope rule: one guide, one task. "Create a matter team" is a guide. "Working with Teams" is a folder of guides that hasn't been split yet.
- Title is the task, verb first. The title should match what a user would type into a search bar mid-problem.
- Prerequisites in one line. What access, what role, what must exist first.
- Numbered steps with screenshots. One action per step. Screenshot anything a user must click, with the click point marked.
- "If something goes wrong." The three most common failure points and what each one means. This section prevents more tickets than the steps do.
- Two pages, hard limit. If it doesn't fit, the task is actually two tasks. Split it. The limit is the discipline.
The scope rule: one video, one task, two minutes. If the cut runs past three, it's two videos.
- Skip the ceremony. No intro animation, no logo sequence, no "hi everyone." First sentence: "Here's how to [task]." Users grant you about ten seconds of trust; spend them on the task.
- Record the real interface. Actual product, realistic sample data, current version. Polish is optional. Accuracy is not.
- Script skeleton: say it, show it, recap it. One sentence on what's about to happen, the doing, then a ten-second recap of the steps just taken.
- Cursor discipline. Move deliberately, pause on each click, let the interface settle. Viewers can't follow a cursor that darts.
- End on the finished state. The last frame is what success looks like, so viewers can verify their own screen against it.
The scope rule: a complete process a user performs start to finish: a provisioning flow, a request lifecycle, a setup sequence. This is the format for the workflows that generate the most one-on-one requests.
- Chapter it. Timestamps for each phase, so returning users jump straight to the step they're stuck on. Nobody rewatches eleven minutes to find one click.
- Show the standards inline. Naming conventions, required fields, and approval criteria demonstrated in the flow, not appended as a policy slide.
- Narrate the why at decision points. "We name it this way so it sorts by matter number" turns compliance into understanding.
- Pair it with the two-page guide. The video teaches the flow; the guide serves the mid-task lookup. Same task, two doors.
Part II · The System
The Library That Gets Found
Content nobody finds is content nobody built. Findability is half the craft.
Naming that matches the moment
Users search the way they think when stuck: with the task. Name every artifact verb-first, task-first, in the user's vocabulary, not the platform's.
| Instead of | Name it |
|---|---|
| Teams Training Part 3 | Create a matter team |
| AI Module 7: Advanced Features | Summarize a document with AI |
| New Hire Deck v4 FINAL | Set up your accounts on day one |
| Q3 Platform Update Overview | What changed in [platform] this quarter |
One front door
Every artifact lives behind a single searchable entry point: one portal, one search bar, every tool and guide one click away. Scattered content in email threads and buried folders fails the 1 AM test no matter how good it is. The library needs an address people remember at midnight.
The feedback pulse
Every guide and video carries two questions: Did this solve it? What was missing? The answers are the roadmap. "What was missing" tells you the next artifact to build; falling "solved it" rates tell you which artifact went stale. The library grows in the direction users actually pull it.
Freshness has an owner
- Every artifact has a named owner, not a team.
- Platform release notes trigger a review of every artifact touching that platform. A stale guide is worse than no guide: it costs the user the task and their trust in the library.
- Retire loudly. When an artifact is superseded, redirect the old link to the new one. Dead links at 1 AM are how libraries lose their reputation.
Part III · Distribution
Content That Travels On Its Own
The best distribution channel is a colleague saying "oh, there's a video for that."
- Put links where questions happen. Ticket auto-replies, chatbot answers, intake confirmations, error messages where you control them. The guide should arrive in the same channel as the question, seconds after it's asked.
- Answer with links, every time. When anyone on the team answers a question live, the answer ends with the artifact link. Every answered question becomes a distribution event, and every "there's no artifact for that yet" becomes a build ticket.
- Feed the peer network. The field data says users share good content with each other faster than any announcement reaches them. Make sharing effortless: clean URLs, no login walls inside your own organization, titles that make sense out of context.
- Train the trainers to produce. A training team that only delivers sessions is a bottleneck with a calendar. A training team that produces artifacts is a publishing house. The Rollout Playbook's train-the-trainer principle applies doubly to production skills: recording, cutting, writing to the two-page limit.
The Production Line
Minimal tooling, batch rhythm, honest metrics.
- Tooling stays small. A screen recorder, a decent microphone, a quiet hour. Production value is the enemy of production volume; users choose the accurate video over the cinematic one every time.
- Batch the recording. One prepared afternoon produces four to six micro-videos. Context-switching is what makes video production feel expensive; batching removes it.
- Template everything. One guide template, one video intro line, one thumbnail format. Decisions made once, reused forever.
- Measure what matters: self-service resolution rate, sessions avoided, artifact views at off-hours (the 1 AM test, empirically), pulse scores, time from "what was missing" to shipped artifact.
Notes & Provenance
- Field results (workflow videos producing SOP-conformant requests; 2-minute module producing a shadow AI report; peer distribution effect): BlockBrain Labs deployment observations, anonymized.
- Delivery economics and retention research: see The AI Rollout Playbook, "The Math of One-on-One at Scale," with sources.
- Companion frameworks: The Vendor Verification Playbook and The Shadow AI Field Guide.