Before the First Draft

The problem

An expert who cannot start does not usually say so. They agree in the meeting, the deadline arrives, and nothing exists. That reads as a motivation problem, but in the work I ran it was almost always a specification problem.

Ask a pathophysiology instructor to "make something on acid-base balance" and you have handed them a project with no edges. They carry a full teaching load, they are not trained in instructional design, and the request gives them no way to know when they are finished. Postponing it is the rational response.

What follows is the method I used to direct content production across six disciplines I was not qualified to teach, with faculty producing on top of full loads. It is organized around a single commitment: no one is asked to produce anything until a one-page brief exists that they have agreed to.

Stage one · Needs assessment

Evidence, not assumption

Three signals, and a rule about them

Topic selection is where most content libraries go wrong, because the default input is what instructors assume students need. That produces material nobody looks for. Three independent signals were available and each one is weak alone.

Search query frequency is unprompted demand. It captures what students went looking for at the moment they were stuck, which is the closest thing to a revealed preference available in an academic setting. Its weakness is that it only sees questions students knew how to ask.

Faculty report captures expert observation of where understanding breaks down. Its weakness is recency and salience bias: instructors remember the student who frustrated them most, not the most common failure.

Course-level patterns, meaning where marks are consistently lost and which questions recur, capture failure at scale. Their weakness is that they show where students fail, not why.

A topic earned production when two of the three signals agreed.

Each signal is weak on its own, which is why the rule required agreement between two of them. In practice it removed topics a single vocal instructor wanted and promoted topics nobody had raised but students were searching for. It also gave me a defensible answer for an expert asking why their suggestion was not in the queue, which is a conversation that otherwise damages the relationship.

Stage two · The scoping conversation

Where the project actually starts or dies

One call, up to an hour and frequently less, ending in a written brief. The slot is a ceiling, not a target. When the questions are answered the call ends, and finishing twenty minutes early is a good outcome. Experts notice, and it is part of why they take the next meeting.

The purpose is not to learn the subject. It is to extract the shape of the problem from someone who has automated their own expertise and stopped noticing what they know.

The questions below ask about incidents instead of content. An expert asked "what should students know about this" produces a syllabus. The same expert asked "tell me about the last time you had to correct this" produces a teachable moment with a misconception attached.

• What do students get wrong about this, specifically? Opens on the failure rather than the topic. The answer is the module's actual subject.
• Tell me about the last time you had to correct that misunderstanding. Incident recall. Produces concrete detail and usually a usable example.
• What do you find yourself explaining more than once? Surfaces repeated effort, which is the clearest signal that an asset should exist.
• What should a student be able to do afterward that they cannot do now? Converts the conversation into an outcome. If the answer is "understand," keep asking.
• If they remember one thing in a year, what is it? Forces prioritization from someone who considers all of it important.
• What are we deliberately not covering? Scope control, in the first conversation rather than the third draft. This is the one people skip.

The division of labor

Stating this explicitly is what makes an overloaded expert agree. Most refusals are about an imagined workload, because the expert assumes they are being asked to build a course.

I could not have specified what a microbiology explainer should contain. I could specify when it was due, what shape it should take, how its topic was chosen, and what question the reviewer was answering. Separating the subject from the structure is what let one coordinator direct production across six disciplines.

Stage three · Schedule

Contact scaled to work in progress

Asynchronous review is the backbone, because the alternative is asking people with full teaching loads to hold a recurring slot for work that is not their primary job. Meetings are for unblocking, not for status.

Frequency is set by how much is in flight and reassessed at each milestone instead of fixed at kickoff. This is the project management layer of the work. The failure mode on one side is a month of silent drift. On the other it is an expert who starts declining the invitation.

Stage four · Staged review

A draft sent with "any thoughts?" attached returns either silence or a rewrite, and both cost more than they should. Each round asks one question and the expert is told which one.

1. Accuracy only Is anything here wrong? Not unclear, not incomplete. Wrong. This is the round only the expert can do, so it is the only round that gets their full attention.
2. Emphasis Is anything over-weighted or under-weighted for this learner? Now that the content is correct, is it correctly proportioned?
3. Sign-off Are you willing to have your name on this? A yes or a specific objection, not another pass.

What this does not solve