Skip to content
Prospicience Consultancy
E10In production

Brand rules written so AI tools follow them

Used across our client projects

Prospicience writes a short brand rules file that AI tools can read and follow: audience, purpose, colour, type, accessibility and what the brand must never look like. The same file briefs a human designer and an AI tool in identical words, and rules are checked on every output, so content is on brand from the first draft.

rules per brand
6 areas
things the brand must never look like
listed
works for
people and AI tools
How it works
  1. 01

    Wrote a short rules file per brand covering who it is for, what...

  2. 02

    Turned rules such as "no dashes in copy" or "only one orange...

  3. 03

    Kept the file in plain language

The gate

Rules are checked on every output rather than applied as a correction afterwards, so work is on brand from the start. Naming what the brand must never look like keeps AI tools inside the lines more reliably than description alone.

The challenge

AI made content drifts off brand, because the brand usually lives in a designer's head and a PDF that no AI tool can read.

Why it mattered

Off brand output has to be caught and redone by hand, which erases the time AI was meant to save. For an agency running many brands at once, that clean up multiplies with every brand and every asset.

What we built

  • Wrote a short rules file per brand covering who it is for, what the page must do, colours, fonts, accessibility and what the brand must never look like.
  • Turned rules such as "no dashes in copy" or "only one orange button per screen" into checks run on every piece of output.
  • Kept the file in plain language, so it briefs a designer and an AI tool the same way.

The result

  • Content on brand from the first draft, not fixed afterwards.
  • One brief shared by designers and AI tools, so both work from the same rules.
  • A reusable format now used across our client projects.

Built for trust

Rules are checked on every output rather than applied as a correction afterwards, so work is on brand from the start. Naming what the brand must never look like keeps AI tools inside the lines more reliably than description alone.

AI at work

AI design and writing tools read each brand's rules file before they draft, so every page, post or asset starts on brand. Checks run on every output and catch a broken rule before a person spends time on it.

What it produces

  • on brand first drafts
  • rule checks on every output
  • one brief shared by designers and AI tools
  • a named list of what the brand must never look like

How it works, day to day

A brand's rules file sits alongside the work. AI tools read it before drafting, the checks run on what they produce, and a designer reviews the result against the same rules.

Where it saves time, and gets better with use

  • Off brand drafts are stopped at the check instead of redone by hand.
  • The same file briefs every AI tool and every designer, so a new tool or a new person starts from the rules, not from scratch.
  • A rule written down once is reused on every asset and every project for that brand.

The technology

  • A short rules file per brand
  • Covers who it is for, what it must do, colour, type and accessibility
  • Says plainly what the brand must never look like
  • Briefs a designer and an AI tool the same way
Technical notes
  • Six rule areas per brand, expressed as machine readable constraints alongside the prose
  • Named anti references, which constrain generation more reliably than positive description
  • Rules checked against every output rather than applied as a correction afterwards

Questions we get asked

Why do AI tools drift off brand?
Because the brand usually lives in a designer's head and a PDF, and a model can read neither. Written rules, including what the brand must never look like, keep AI output on brand from the first draft.
What goes in a rules file?
Six areas: who the work is for, what the asset must do, colour, type, accessibility, and what the brand must never look like. That last list does the most work, because naming what to avoid narrows the output far faster than describing what to aim for.
Does this replace a designer?
No, it briefs one. The same file that guides an AI tool gives a designer a clear starting point, which is why we keep it in plain language rather than as configuration.

Related systems

Have a process that should run itself? Let us build it.

Tell us the problem. We will show you what we would build, how fast we can ship it, and what it will deliver, from a first pilot to a full platform.

Book a consultationcontact@prospicience.in