The one-pager you should write before every AI-assisted task

Two consultants are about to write the same proposal with Claude's help.

The first opens a fresh conversation and types "draft a proposal for a B2B SaaS company that needs help with their conversion funnel." Claude returns 800 words. The output is competent, generic, and identical to what every other consultant in the country would get from the same prompt. The consultant edits for 90 minutes to get something usable.

The second writes a 5-section brief — 90 seconds of work — before opening Claude. The brief names the prospect, the real problem (not the named one), the budget signal, the voice rules, and the one thing the consultant knows that the AI doesn't. Then she opens Claude with the same task, paste the brief, ask for the proposal. Claude returns 500 words. The output is specific, useful, and almost-shippable. The consultant edits for 15 minutes.

Same tool. Same time of day. Same task. Six-times difference in output quality and time saved. The variable is the brief.

This article is the brief format, why it works, and how to install the habit so it fires before every AI-assisted task.

Why generic output is an input problem

The instinct when AI returns generic output is to blame the model. The reality is that generic input produces generic output. A vague brief like "a B2B SaaS that needs help with conversion" matches roughly 50,000 companies. Claude doesn't know which one you're talking about, so it averages across all of them — and the average B2B SaaS doesn't actually exist. The output is competent because Claude is competent. The output is generic because the input was.

This is covered in more depth in why your AI prompts feel generic, which makes the broader case. The specific fix is the brief, and it works because it does the one thing the model can't do for itself: localize the task to this prospect, this engagement, this moment.

The brief is what good consultants implicitly construct in their heads before any meaningful client work. The advance writers, the experienced strategists, the operators who've shipped 200 proposals — they're doing this work mentally in 30 seconds while they walk to their desks. The one-pager is just the explicit version of that, written down so the AI gets to see it.

The five-section brief

The format that works for almost any consulting task:

1. Subject. One line. Who or what is this for? "Acme Widgets — restaurant SaaS — recovering conversion." Twelve words. Names the specific company, the industry context, and the situation.

2. Real problem. Two to three sentences. Not what the prospect said they wanted — what they actually need. "Priya named conversion recovery, but the real problem is an internal disagreement between her and engineering about the cause. The work they'll pay for is the diagnostic that settles the disagreement."

3. Budget signal. One line. "Indirect positive — 'we have some budget.' Likely in the $10-20k range based on company stage and stated priority." Or "Direct — they named $15k." Or "None — need to surface."

4. Voice rules. Two to four short lines. "Direct, peer-to-peer. No 'I'm excited to partner with you.' No exclamation points. Match the way Priya writes her own emails — short sentences, dry humor." This is the section that produces the biggest output-quality lift.

5. The thing the AI doesn't know. One to three sentences. The specific context that makes the work non-generic. "Priya's previous consultant pitched a copy rewrite and didn't deliver. She's wary of anything that sounds like that pitch. Avoid 'optimize' and 'improve' as verbs — they were in the previous consultant's deck."

That's the brief. Five sections, 80-150 words total. 90 seconds to write. The output Claude produces against this brief is several times more useful than the output it produces against a generic prompt.

The 90-second test

The brief format has a fast self-check. Read your draft brief and ask: could a different consultant pick up this brief and produce roughly the same work you would?

If yes, the brief isn't specific enough yet. Add the detail that makes it yours: the voice rule that's particular to this client, the context the AI couldn't infer, the gut read that comes from having been on the call. The brief should be unusable to anyone but you and Claude working together — that's the signal that it's localized properly.

If no — if the brief contains enough specificity that another consultant would diverge from your approach — you're ready to paste.

This check takes about 15 seconds. It's the second-highest-leverage step in the whole workflow.

What changes in the output

The five-section brief changes Claude's output in four predictable ways:

Specificity replaces averaging. Instead of "For a B2B SaaS facing conversion challenges," the output opens with "Priya — you've watched trial-to-paid drop from 18% to 9% over the last quarter, and there's an internal split on why." That's a different opening because the brief named the specifics.

Voice consistency. When the voice rules are in the brief, every output produced from that brief carries the same voice. Three proposals written for three different clients across the same week have three different substances but the same voice — which is what makes them recognizably your work product.

The right rebuttals show up automatically. When the brief names "avoid words the previous consultant used," Claude does. When it names the specific objection the client is going to raise, Claude pre-empts it. The brief is the cheat sheet for what the AI should avoid and what it should lean into.

Edit time drops. This is the practical payoff. The reason the second consultant edits for 15 minutes instead of 90 is that the output is already in the right shape, in the right voice, addressing the right substance. The edits are polish, not rebuild.

The free version: a brief that's also a debrief

The clearest worked example of the brief idea in production is the Free Discovery Debrief. It's a single prompt that takes raw discovery-call notes — the unstructured version of the five-section brief — and turns them into a structured debrief Claude can use as input for everything downstream.

The flow: you take messy notes during a discovery call (no formatting, no reformatting, just bullets). After the call, you run the Discovery Debrief prompt. The output is a one-page structured debrief that has all five sections of the brief format above (subject, real problem, budget signal, voice rules implied, the things the AI doesn't know). That debrief then becomes the brief you'd paste into Claude before writing the proposal, the scope doc, the follow-up email — any artifact downstream of the call.

It's the same brief format, generated automatically from inputs you already have. Drop the download into your workflow if you'd rather run the prompt than write the brief by hand for every discovery call.

The cases where the brief is wrong overhead

Three situations where writing a full five-section brief is unnecessary friction.

Truly simple tasks. "Rewrite this paragraph in a more direct voice." Doesn't need a brief. The input is the paragraph; the instruction is the voice. Brief overhead would be longer than the task.

Tasks inside a Project that already has the context loaded. If you've set up a Claude Project for the Acme engagement (per the article on client memory across Claude sessions), the Project's uploaded files already contain most of the brief content. The "brief" for a task inside the Project becomes a one-line task description.

Exploration, not execution. When you're using Claude to think out loud — what could this strategy look like, what are the angles on this problem — the brief locks you into a specific path too early. Use a looser opener for exploration; reach for the brief when you're producing a deliverable.

For everything in between — the actual production tasks where you're generating something the client will see — the brief earns its 90 seconds.

Pitfalls

Writing the brief in your head and not on the page. The mental brief feels like the real brief, but it isn't — Claude can't see it. The written brief is what the model uses as input. If it's not pasted, it doesn't count.

Padding the voice-rules section with abstractions. "Professional but conversational tone." That's useless to Claude. Specific instructions work: "Short sentences. No em-dashes more than once per paragraph. Don't use the word 'optimize.'" Concrete rules produce consistent voice.

Forgetting Section 5. The "thing the AI doesn't know" section is the highest-leverage one — it's the part of the brief that produces non-generic output. Most consultants skip it because the first four sections feel sufficient. The output difference between a 4-section brief and a 5-section brief is the difference between competent and shipped.

Using the same brief for different tasks on the same client. A brief calibrated for the proposal isn't right for the scope doc, even on the same client. The subject changes (proposal vs. scope), the real-problem framing shifts (sell the diagnostic vs. detail the deliverables), the voice rules may stay constant but the section-5 specifics differ. Write one brief per task, not one brief per client.

Treating the brief as overhead instead of leverage. Consultants who write the brief at the top of their workflow ship sharper work in less total time. The 90 seconds is not a tax; it's the savings.

What to install this week

Pick the next AI-assisted client task on your calendar. Before opening Claude, write the five-section brief. Time how long it takes (target: under two minutes). Paste it. Run the task. Compare the output to the last similar task you did without a brief.

The first time you do this, the difference is noticeable but small — because you're still calibrating what to put in each section. By the third time, the difference is substantial. By the tenth time, you stop being able to imagine working without it.

The brief is the workflow. The AI is the production engine. Together they produce work that signals specificity and voice in a way that generic prompts never can — and that's the whole game for solo consultants whose differentiation IS specificity and voice.

If you want the canonical worked example of the brief idea in action — the prompt that produces the brief from raw call notes, the Claude skill that auto-activates after a discovery call, the Acme Widgets walkthrough — the Free Discovery Debrief is the download. The Proposal-Closer Prompt Pack extends the same brief-and-execute pattern across the full eight-step proposal workflow. And when a task outgrows the one-pager — a whole deliverable handed over in one run — the Claude Workspace Kit ships six engagement-grade briefs built on the same equip-don't-just-instruct discipline, wired to a workspace folder, review passes included. The folder pattern those briefs live in — why files beat pasting for deliverable-shaped work — is its own article.

The 90 seconds upfront is what makes AI-assisted consulting work read like consulting work instead of AI work. Spend the 90 seconds.