The Output layer says what the finished document looks like: the sections, their order, what has to appear, what stays separate, how long. Get it wrong and you receive correct information in a shape somebody reformats by hand every time, until the workflow quietly stops being used.
Key takeaways
- Name the sections and their order. "Professional" and "well structured" describe nothing.
- Format drift is the quiet killer of AI workflows, because the saving disappears into manual reformatting.
- State what must stay separate: GST, exclusions, assumptions, questions.
- State the length. Assistants pad by default.
- The fastest way to write this layer is to copy the structure of a document you already send.
What the layer holds
| Element | Example on a quote |
|---|---|
| Sections and order | Scope, itemised table, subtotal, GST, total, exclusions, questions |
| Must appear | Every line item with its own price |
| Must stay separate | GST as its own line. Assumptions listed, never merged into scope |
| Length | Scope in one line per item. No introduction |
| Tone carrier | Match the two example quotes supplied |
| What to leave out | No marketing language. No closing paragraph about being happy to help |
Steal from your own documents
You already have the structure. Open the last document of that type you were happy with and list its sections in order. That list is the Output layer. Ten minutes, and it beats any description of style you could write.
Why format drift matters more than it sounds
A workflow that produces correct content in the wrong shape gets reformatted by hand. That takes five minutes. Five minutes on every run erases the saving, so after a fortnight the person goes back to writing from scratch and reports that AI did not help. Nothing was wrong with the content. The Output layer was vague.
Common mistakes
Adjectives instead of structure. Professional, clear, engaging and concise are opinions. Sections, order and length are instructions.
No length limit. Default behaviour is padding, and padded documents get edited down by hand.
Forgetting the questions section. A short list of what the assistant needs clarified turns a draft into a working document, because the gaps are visible instead of hidden.
Merging separate things. Assumptions folded into scope, GST folded into totals. Both cause arguments with customers later.
Where it sits in the Stack
Output is the top layer, above Context, Task, Source and Rules. Read the full Context Stack.
Build yours from a document you already send: a full day at Avani Mooloolaba Beach Hotel, A$495 excluding GST, capped at 15 people. See the workshop, or enquire about onsite team training.
Frequently asked questions
How detailed should the Output layer be?
Detailed enough that the draft needs no reformatting. Sections, order, what stays separate, and a length limit.
Why does AI output come back too long?
Padding is the default behaviour. State the length and what to leave out, and it stops.
How do you get the tone right?
Supply a document you already send as a tone example in the Source layer. One real example does more than a paragraph of adjectives.
Should you ask for a specific file format?
Ask for the structure. File conversion is a separate step and it is rarely the bottleneck.
What is the questions section?
A short list at the top or bottom of anything the assistant could not resolve from your material. It turns a hidden assumption into a visible decision.
Why did the workflow get abandoned even though the content was right?
Almost always format drift. Manual reformatting on every run cancels the time saving. How to measure whether it saved time.