The gap between a throwaway AI answer and consultant-grade output is the brief, not the model. Three prompts — positioning, pricing pressure-test, weekly numbers review — with the context recipe that powers them.

The consultant was never magic. The brief was.

Think about what you're actually buying when you hire a consultant: not intelligence — structured attention. They ask about your business, your market and your constraints, then apply a framework you could have Googled. The £2,000 wasn't for the framework. It was for someone finally putting your specifics through it.

That's the part you can now do yourself. Not because AI "replaces consultants" — for genuinely hard problems it doesn't — but because for the standard 80% (positioning check, pricing sanity-test, monthly numbers review), the framework plus your specifics gets you most of the way. The catch: the model only has your specifics if you hand them over.

The decoded idea: the brief is the product. A weak prompt ("how should I price my services?") gets the blog-post answer, because you gave it blog-post context. So build a context block once — five lines: what you sell, to whom, your capacity, your current prices, your constraint — save it, and paste it at the top of every serious prompt. Same one page every time. That block is what turns a generic model into your analyst.

The three prompts. Paste your context block first, then:

1. The positioning check:

"Act as a positioning consultant. Given my context above: (1) who is my sharpest-fit customer, described in one sentence they'd recognise; (2) what am I implicitly competing against, including 'do nothing'; (3) what's the one claim I can make that competitors can't credibly copy? Challenge me — if my current positioning is fuzzy, say so and show the fuzziest phrase."

2. The pricing pressure-test:

"Act as a pricing analyst. Using my context: lay out three price structures (current, value-anchored, premium-tier), and for each: who says yes, who walks, and what I'd have to believe about my market for it to work. Then list the three questions I'd need real data to answer before changing anything. Do not tell me to 'know my worth' — give me testable reasoning."

3. The weekly numbers review:

"Act as a part-time finance director. Here are this week's numbers: [paste revenue, costs, pipeline]. Tell me: (1) the one number most out of line with the last four weeks; (2) the likeliest boring explanation before any dramatic one; (3) the single question I should answer before next week. Two hundred words maximum."

Notice what all three share: a role, your context, a structure for the answer, and — the underrated part — an instruction to push back. Consultants earn their fee when they disagree with you. Ask the model to do the same or it will politely agree you into a wall.

Where the ceiling is. These prompts produce the analysis; they don't produce the judgement. The model doesn't know your market's mood, your relationships or your appetite for risk — the final call on positioning and price stays with you. That's not a limitation to apologise for; it's the job description. Analysis: delegated. Decisions: never.

The responsible-AI note. Your context block and your numbers are commercially sensitive — use them only in a tool you've vetted, with training off. Verify any market "facts" the model asserts before acting; it will state plausible figures with total confidence. And since we're talking pricing and money: this is analysis structure, not financial advice — your numbers, your accountant, your call.

A worked example, so this isn't theory. Take a freelance web designer using prompt two. Context block: bespoke sites, service businesses, capacity three projects a month, £2,400 average, constraint is lead flow. The value-anchored structure the model proposes will likely anchor on what a site earns the client, not hours — say £3,500 with a results-review call built in. The useful output isn't that number (don't trust it; it doesn't know your market). It's the "what you'd have to believe" list: that clients buy outcomes, that your portfolio proves them, that losing the most price-sensitive third of enquiries is survivable. Each belief is testable this month — raise the price on the next two proposals and count what happens. The prompt didn't set your price. It converted a vague anxiety ("am I too cheap?") into an experiment with a sample size.

Make it a cadence, not an event. These three prompts work best on a rhythm: numbers review weekly (Friday, ten minutes), positioning check quarterly, pricing pressure-test twice a year or whenever capacity is full — full capacity is the market telling you something, and prompt two is how you work out what. Save all three, with your context block, in one note. The system only compounds if the prompts are one paste away, not one memory away.

One decoded AI move, every week. Follow @aithebrief on Instagram for the daily version, or subscribe free on Substack for the deep one.