Prompting as a Working Method
Imagine a new colleague: sharp, fast, absurdly well-read — and it is their first day. They know nothing about your company, your customers, or what your boss considers "done." If you hand them "write a summary of this account," you will get a competent answer to the wrong question. Brief them the way you would brief a capable new hire, and they are astonishing by lunchtime.
That is prompting. A strong prompt is not a magic phrase — it is a compact work brief.
Weak brief, strong brief
Here is the difference on a real task. Maya at Northstar Components needs a renewal brief for Helios Manufacturing, a customer up for renewal in six weeks.
The weak version:
Summarize this account.
Claude will comply — with a generic overview that mentions everything and prioritizes nothing, because Maya never said what the summary is for. Now the brief version:
Create a one-page renewal brief on Helios Manufacturing for an account
executive preparing a renewal call, so that the call focuses on retention
risks and expansion openings.
Use only the attached account notes and support-ticket summary.
Include: current products and usage trend, the two biggest renewal risks,
one expansion opening, and three questions to ask on the call.
Do not invent facts that are not in the sources. If something important
is missing, list it under "Unknowns" instead of guessing.
Format: headed sections, bullets, 300 words maximum.
Before finishing, check: does every risk cite a specific note or ticket?
Same model, same data. The second prompt produces something Maya can evaluate — every element traces to a source, gaps are declared instead of papered over, and a good result is recognizable at a glance.
The six parts of a working brief
- Task — the action Claude should take
- Context — situation, audience, and purpose
- Sources — the material Claude should rely on
- Constraints — boundaries, exclusions, tone, and length
- Output — the required structure or format
- Quality bar — what a good result must contain
As a reusable skeleton:
Create [output] for [audience] so that [purpose].
Use only [approved sources].
Include [required elements].
Do not [important exclusions].
Return it as [format].
Before finishing, check it against [quality criteria].
Not every prompt needs all six spelled out — a quick brainstorm needs two lines. But when a result will be reused, shared, or acted on, the brief earns its length many times over.
Show, don't only describe
When wording, structure, or tone matters, ten adjectives lose to one example. "Professional but warm" means something different to everyone; a pasted example of last quarter's best customer email means one thing. Include a short example and label what it demonstrates — "match this tone" or "follow this structure, not this content."
Iterate with evidence, not vibes
First drafts are the start of the loop, not the end. When output misses, resist "make it better" — point at the gap:
- "The recommendation is not supported by the supplied figures."
- "The tone is too promotional for an internal risk review."
- "You compared two options; compare all three against the stated criteria."
And when a task is fuzzy, flip the direction: ask Claude to interview you first — "Before drafting, ask me the five questions that would most improve this." It is often the fastest way to discover what your own brief was missing.
Placeholder: replace with a four-step loop: brief, draft, evaluate, refine.
Practice
Take a vague request you actually received recently and turn it into a working brief using the skeleton. Test it twice with different inputs. Keep it only if a colleague could run it and recognize a good result without asking you — that is the test of a reusable prompt.
Ready to move on
Your brief names the outcome, supplies approved context, sets constraints, and includes a quality bar someone else could check — and you improve results by pointing at gaps, not restarting.