Sales copy and materials. All of it giving the same figures.

Spec sheets, web pages, presentations, emails, posts. Written in different years by different people, and between them they give three different weights for the same product. This piece produces them from a single set of facts approved once, and gets them signed off before they go out.

This page covers a single piece of the system. The other pieces, and how we choose which one to start with, are on the services page.

The problem isn't the writing, it's documents that contradict each other

A business that's been around for ten years has dozens of documents in circulation, and nobody knows which one is right. The net weight on the 2021 spec sheet, the units per case on the 2024 price list and the lead time in yesterday's email are three figures that should match, and often don't.

The first person to notice is always a buyer, and they notice at the worst moment: when they've already placed an order based on one figure and received another. At that point the problem stops being a typo and becomes a matter of trust.

A wrong document goes unnoticed until someone uses it to make a decision. Then everything shows up at once.

What changes in practice

The missing document stops being missing. Enquiries that stall today because nobody has time to prepare an English spec sheet or a page for a new product get moving again, because the draft is ready the same day.

The second change matters more and shows up later: every document says the same thing. Someone who reads about you in three different places finds the same figure three times, and doesn't have to ask you which one is right.

The third is that you know where the bottleneck is. The time from draft to publication is the metric, and in most businesses that time isn't spent writing, it's spent waiting for approval on someone's desk.

People read about a fifth of it

The first sentence carries the key fact, and there's a measured reason for that. Jakob Nielsen analysed 45,237 page views to work out how much text actually gets read: 28% of the words at most, even if people spent their entire time on the page reading.

The realistic figure is lower, and he says so in as many words: “More realistically, users will read about 20% of the text on the average page”. A sales document that puts the key figure in the third paragraph is hiding it from eighty per cent of the people who open it.

The practical consequence is a writing rule that applies to every document: the full answer in the first two or three sentences, figure included, and the reasoning underneath. People who read everything still find everything; people who read three lines leave knowing what they needed.

The set of facts is agreed once

It's the piece that makes everything else work, and it's built during the analysis. Every item has a value, the date it was last checked and the person who approved it, so an out-of-date figure stands out at a glance instead of surviving for years inside a PDF.

What goes into the set of facts	What happens when it changes	What you get
Product figuresweights, formats, cases	You change the item in one place, and the system lists every published document that still shows the old value.	No price list in circulation with last year's weight, and no wrong orders as a result.
Termslead times, minimums, Incoterms	The change goes to the person who needs to approve it before it appears in any document, because it commits the business.	The terms you promise in writing are the ones you can actually meet.
Certificationswith expiry dates	The expiry date sits next to the certificate, and the system flags any document that mentions it before it expires.	No spec sheet claiming a lapsed certification in front of a buyer who checks.
Voicehow you talk	We gather existing materials that sound like you, and the words you never use are left out.	Drafts come out sounding like you, and fixing them takes two minutes rather than a rewrite.

What the system does, step by step

Every document goes through the same five steps, whatever its format, and each step leaves a log entry recording what happened and who decided it.

Step	What happens	What you get
The briefwhat's needed, and for whom	You set out the document, the reader and the one thing it needs to achieve. A document with no stated reader doesn't get started.	No pages written for nobody, which are exactly the ones that never get used.
The draftfrom your facts	The draft is built on the figures in the set of facts and on the materials you already have, with the answer in the first few sentences.	Copy that contains what you know, instead of sentences that could be about anyone.
Figure checkautomated	Every figure in the draft is checked against the set of facts, and any that don't match are highlighted.	The kind of error nobody spots on a reread, a plausible but wrong figure, is caught before it goes out.
Approvalby a person	The document joins the approval queue of the person who needs to sign it off, and the system flags what's stuck and for how many days.	The bottleneck becomes visible, and the person in charge knows where it is instead of finding out at the end of the quarter.
Publicationwith labelling	The approved document goes out, with machine-readable marking wherever the content was generated by the system.	The transparency obligation is met at source, instead of being patched up later across dozens of files.

The review, which is what makes the rest fit to publish

A model that writes always produces text, even when it doesn't have the facts. It fills the gap with a plausible figure, and that figure looks just like a correct one: that's why the automatic check of figures against the set of facts comes before the human read, not after.

The human read is there for what a machine can't do: judge whether a sentence can be said to that customer at that moment. That's a commercial decision, and it stays with the person responsible for the relationship.

The approval queue is also where the service is measured. The agreed number is the time from a finished draft to publication, and it's measured on a real batch before we start, so the comparison afterwards actually means something.

AI-generated copy is labelled, and from August 2026 it's the law

Article 50 of Regulation (EU) 2024/1689 applies from 2 August 2026. For artificially generated content, it requires providers to ensure “that the outputs of the AI system are marked in a machine-readable format and detectable as artificially generated or manipulated”.

For systems already on the market before that date, the marking obligation alone is pushed back to 2 December 2026, while the article's other obligations apply straight away. What it actually means for a small business, with the scope and the four things to get in order, is on our page on Article 50 of the AI Act for Italian SMEs.

Our position is written down and doesn't depend on the law: we label it regardless, because a buyer who finds out on their own that they were reading generated copy stops trusting even the parts that were true. The full reasoning is on our page about the principles we build by.

How many are already doing it, and who's falling behind

It's no longer a niche. According to Eurostat, in 2025 20.0% of European enterprises with at least ten employees were using AI technologies, up from 13.5% in 2024. 8.8% were using it to generate written or spoken language, and 11.8% for text analysis.

The gap that matters isn't between countries, it's between company sizes. In the same survey, enterprises with more than 250 employees stood at 55.0%, and those with ten to forty-nine at 17%. Thirty-eight points apart, on the same technology, in the same year.

That gap closes on jobs like this one, where the expensive part is the time of people who, in a small business, have other jobs too. The survey only counts enterprises with ten or more employees, so the smallest aren't counted at all and nobody knows where they stand.

What goes out on its own, and what waits for a person

Everything that doesn't commit the business runs on its own: the draft, the figure check against the set of facts, flagging documents that need updating, and reminders about anything that's been stuck in the queue too long.

Nothing that gets published or sent to anyone goes out on its own. A price, a term, a delivery promise and every line carrying your name wait for a person to read and approve them, and every approval records who and when.

Same mechanism, different name in every trade

The mechanism is identical; the missing document isn't. It's worth looking at your own case, because that's where you see which enquiry is stuck today.

Sector	The missing document	Where you see it
Food and agricultureand export	The English spec sheet, with nutritional values and allergens, requested by a buyer who has been waiting eight days.	What goes into a spec sheet that doesn't waste anyone's time
Restaurantsand bars	The menu, updated in three different places, where the dish dropped in March still appears and the price is the old one.	A dinner service with the phone ringing unanswered
Hospitalityaccommodation and events	The document setting out what's included and what isn't, which today gets rewritten by hand inside every quote, and comes out slightly different every time.	The run-up to the season for a guest accommodation business

What this piece doesn't do

It doesn't publish on its own or send anything to anyone. It prepares, checks, queues and flags: the final click stays with a person, and that isn't a limitation to remove once the system matures.

It doesn't translate materials that need to go out in another language. Translation using the agreed glossary, reviewed by someone who knows the target market, is a piece in its own right: multilingual drafts.

It doesn't handle search rankings. Writing the material and getting it found are two different jobs: the second is visibility in search engines and AI assistants, and without the first it has nothing to show.

Questions and answers