A form you have never written before needs its conventions requested first

The difficulty is the convention, not the prose

Sooner or later you have to produce something in a form you do not work in: a grant application, a legal-adjacent letter, a technical specification, a press release, a eulogy, a board paper, a research protocol. You can write perfectly good sentences and still produce something that reads as wrong to anyone who handles that form regularly, because forms carry expectations that nobody writes down for outsiders.

Those expectations are specific and largely arbitrary. What goes first. What is stated rather than argued. Which sections a reader skips and which they read closely. How much hedging is normal. Whether the thing is written for one reader or for a file.

A generated draft of an unfamiliar form usually gets the surface right and the priorities wrong, which is the worst combination, because it looks convincing enough that you stop asking.

Ask what the form contains before you ask for the form

Split the request in two. First ask what this kind of document normally contains, in what order, and what each part is for. Ask what a reader of it is looking for, what makes one obviously weak, and what a newcomer characteristically gets wrong. You are gathering a specification rather than a draft.

This is a different request from describing the output shape you already have in mind, which is the ordinary case and assumes you know the shape. Here you do not, and the point of the first pass is to find out cheaply, from a description you can sanity-check against any real example you can lay hands on.

Read the answer critically and then go looking for two or three genuine examples of the form. The description gives you the vocabulary to read those examples properly, which is a considerably faster route than staring at them and hoping the pattern emerges.

The reader is the part of the specification that is usually missing

Forms exist because a particular reader needs a particular thing, and once you know who that is, most of the conventions stop looking arbitrary. A board paper is short because the reader has six of them. A specification is dry because it will be argued over line by line. A grant application follows the assessment criteria because somebody is scoring it against them.

So ask what decision the reader is making and what they are permitted to do about it. That question does more to fix the register, the length and the ordering than any amount of instruction about tone, and it is the question people skip when they are anxious about an unfamiliar document.

Where the form is assessed against published criteria, those criteria are the brief and nothing else is. Read them first and structure to them, even where that produces a document you personally find repetitive.

Write the substance yourself and let the form carry it

Once the conventions are clear, the labour splits cleanly. The content — what you are proposing, what happened, what you are asking for, what you know that the reader does not — is yours, and it is the part that decides whether the document works. The arrangement is the part where help is cheap and low-risk.

The reliable sequence is to write your substance in whatever rough form comes naturally, then ask for it to be arranged into the conventional structure, then check the result against the specification you gathered earlier. Nothing has been invented at any stage, which removes most of the ways this goes wrong.

Expect the arrangement pass to smooth away things you meant. Read the output next to your rough version and put back anything specific that quietly became general, because the specifics were the reason the document existed.

Where the conventions are local, and where the stakes rule this out

Many forms vary by institution, jurisdiction and year in ways no general description can capture. The internal report format your organisation uses, the way a particular funder wants budgets presented, the letter a specific court expects: these are local knowledge, and a general account of the form will be confidently approximate about them.

The correction is to get one real recent example from the place it is going, and to treat it as authoritative over anything you were told in general terms. One genuine internal example is worth more than a page of description.

And there are forms where drafting help is the wrong instrument entirely, because the document has legal or clinical consequences and its conventions exist to prevent specific harms. A will, a consent form, a regulatory submission, a formal complaint that starts a clock running. Learn what they contain by all means. Have the actual document prepared by somebody who is accountable for it.