x
One sentence, no adjectives: what you do, who for, and what changes because of it.
Say it out loud. If it takes more than one go, or you reach for "bespoke" or "end-to-end" or "helping businesses grow", the sentence isn't finished — and everything else in this week's theme is built on sand until it is.
One sentence, no adjectives: what you do, who for, what changes. If a stranger can't repeat it back, no page structure will save you — and neither will a model.
This is the last post of the week because it's the foundation, and foundations are easier to see once you've looked at everything standing on them. The page structure, the three quotable lines, the monthly check — all of them assume there is something clear to be found. If the underlying description is vague, structure just makes the vagueness easier to locate.
Not because the business is unclear in their head. Because the words on the website were chosen to sound established rather than to be repeated.
There's a real reason for that, and it isn't vanity. Specific language feels risky. "I do monthly bookkeeping for creative agencies of five to twenty people" closes doors — it visibly excludes the sole trader and the fifty-person firm. "We provide tailored financial solutions for growing businesses" excludes nobody, which feels like safety.
It isn't safety. It's invisibility with better manners.
Here's the test that makes it concrete. Take your current description and ask: could a competitor put their name on this sentence unchanged? If yes, it isn't describing you. It's describing the category, and the category is not what anyone is trying to find.
Category language fails in both directions now. A person can't repeat it to a friend, so you don't get referred. And a model has nothing distinguishing to work with, so you don't get named — it produces generic advice about what to look for, which reads like helpfulness and functions like exclusion. That's the failure mode this week opened with: described as a category rather than named.
Swapping three category words for three specifics. It usually takes an afternoon and feels worse before it feels better.
Name the actual job. Not "financial solutions" — "your monthly bookkeeping and a quarterly conversation about whether you can afford to hire".
Name the actual client. Not "growing businesses" — "creative agencies between five and twenty staff". Include the size, the sector, or the situation. Whichever one you actually screen for when a lead arrives.
Name the actual change. Not "peace of mind" — "you stop finding out about cash problems in the same week they happen".
Then the part that separates this from a branding exercise: add who it isn't for. One line. "Below about £150k turnover you want a bookkeeper, not this." It is the most credible sentence most businesses could publish, it's the one most likely to be quoted, and it's the same move as disclosing what AI does in your work before anyone asks — volunteering the limit is what makes the rest believable.
It costs you enquiries. Whether it costs you work is a different question, and mostly the answer is no.
The enquiries a vague description generates are disproportionately from people you'd decline, at a price you'd refuse, for a job you'd resent. Filtering those out earlier is not a loss; it's the same volume of work reached with less unpaid conversation. Anyone who has spent an hour on a call that was never going to close knows the arithmetic.
The honest caveat: this genuinely is harder if you're early, or deliberately generalist, or in a thin market where narrowing leaves nothing. In that case narrow on situation rather than sector — "businesses in their first year of VAT registration" — which is specific enough to be describable without shrinking the pool to nothing.
And a caution on the AI framing specifically, since it's been this week's lens. Don't write the sentence for a model. Write it so a person can repeat it to a friend at a dinner party, which is a harder and older test, and one a model happens to reward as a side effect. Copy engineered for extraction reads like a reference card and persuades nobody — which would be a strange way to lose, having spent a week getting found.
Being recommended has always required being describable. Referral worked that way long before any of this. What changed is that there's now a second thing doing the describing, it does it at scale, and it does it from whatever text it can find.
So the closing move of the week is the oldest one: say what you do, in words a stranger could repeat, and mean them.
For the audit that finds where your week actually leaks — free, 5-min read: The 10-Hour Audit.
Related Decodes
AI the Brief (@aithebrief) — AI advice you can act on and trust, for time-starved founders.
TL;DR: One sentence, no adjectives — what you do, who for, what changes. If a stranger can't repeat it back, no page structure rescues it. Most founders fail this not because the business is unclear to them but because the website words were chosen to sound established rather than to be repeated, and specific language feels risky because it visibly excludes people. The concrete test: could a competitor put their name on your sentence unchanged? If so it describes the category, not you — and category language fails twice over, because a person can't repeat it and a model has nothing distinguishing to name. The rewrite swaps three category words for specifics: the actual job, the actual client type, the actual before-and-after, plus one line saying who it isn't for. Narrowing costs enquiries rather than work, and mostly the enquiries you'd have declined. If you're early or genuinely generalist, narrow on situation rather than sector. And write it for a person to repeat at a dinner party, not for a model to extract — copy engineered for extraction persuades nobody.
