The system in three levels: the brand book containers at the bottom, the orchestrator in the middle, the assets produced at the top
AI Brand System

Without a system, AI pulls you towards the industry average. With a system, it pulls you towards yourself.

I take the identity you already have and turn it into the context your team and AI use to produce. The rules move out of a document and into the files the work is made from.

Already in production

OpenEconomics works this way every week. Presentations, reports, PDFs and web pages come out of the library, produced by the marketing team. A whole video came out of it too, without editing software.

See how it's built ↗

The problem

If three of these sound familiar, you need the system.

Not everyone starts from the same place. Some have an identity that is coming apart; others have a solid one but can't use it at the speed today requires. The system works for both.

  • 01

    Every presentation starts from the file of another presentation, which in turn was adapted from yet another one.

  • 02

    The website, the sales slides and the LinkedIn profile look like they belong to three different companies.

  • 03

    Changing a colour in a template means going back to the original file, every time.

  • 04

    Someone in the company has started generating images and copy with AI. And it shows.

  • 05

    The brand book exists, but nobody has opened it since it was delivered.

  • 06

    Design requests pile up, and meanwhile marketing slows down waiting for materials.

  • 07

    Recurring production has grown: new presentations, PDFs and pages every week, with the same decisions to repeat by hand each time.

  • 08

    The marketing team could handle repetitive materials on its own, if it had a tool that let it do so while staying on brand.

It isn't a matter of taste, and often not even of organisation. The rule exists, but it isn't where the work happens: it sits in a document while the work happens elsewhere — so every new piece starts from scratch.

Why it happens

Why the rules don't reach the place where work happens.

The brand decisions already exist, and they are usually well written: they live in the brand book, the guidelines, the templates, the materials already approved. The point is that they stay there, in separate documents, while the work happens somewhere else.

Someone putting a presentation together in a hurry picks the blue from a screenshot, because it's quicker than looking up the exact code. The documentation wasn't wrong: it just wasn't there when it was needed.

In a system, colours are values the files read on their own. Components are ready-made blocks with the correct spacing built in. Page templates already exist, and whoever uses them starts from there.

And artificial intelligence receives the same decisions in a form it can apply: named rules, good examples and rejected ones.

How it's built

Three levels, from a single colour to a finished page.

Each level rests on the one below. Change a value at the top and everything that uses it updates, with nothing redone by hand.

Level 01

The basic decisions

Foundations

  • Colours and tonal scales
  • Typography and hierarchy
  • Spacing and grids
  • Icons and charts
Level 02

The reusable pieces

Components

  • Buttons, tags, cards
  • Tables and figures
  • Forms and fields
  • Navigation and footer
Level 03

The finished pages

UI Blocks

  • Website openers and sections
  • Reports and analyses
  • Dashboards and data tables
  • Editorial modules and FAQs

How it works

What goes into the system and what comes out.

On one side, the material the company has today; on the other, what the system produces, every time within the same rules.

Brand bookrules
Existing assetsexamples
Design tokenscode
Componentspatterns
AI Brand
System
Social content01
Web pages and landing pages02
Presentations and reports03
Images and video04

The delivery

What you actually receive.

At the end of the project, this is what you keep, in a repository of your own, in formats any developer or designer can open.

  • 01

    A library that opens in the browser, where every element has a live preview and code to copy.

  • 02

    Colours, fonts and spacing as values the files read, instead of screenshots to sample by eye.

  • 03

    Components and finished pages for website, reports and dashboards, ready to fill.

  • 04

    The written context that artificial intelligence reads before producing anything.

  • 05

    A rubric for judging a result, written so that non-designers can use it too.

  • 06

    The archive of approved and rejected examples, each with the reason for the decision.

  • 07

    Usage documentation: when to use a component, how to combine it, which exceptions to avoid.

Explore the case studies

A real project, from start to finish.

A real case, from start to finish: the audit of existing material, the system architecture, quality control and the finished outputs.

The OpenEconomics design system open on a laptop

Case 01 · Economic analysis and advisory

OpenEconomics

An already solid identity — brochures, presentations, reports and digital products built over time — set up so the marketing team can use it without going through me. Today that system produces presentations, PDFs, web pages, reports and a whole video made without editing software.

Role
Design direction
Output
AI Brand System
Status
In production
Read the case study ↗

Want to see the system applied to your industry? Contact me.

Quality

How the output is checked.

A model that produces quickly also makes mistakes quickly. That's why I don't deliver only the rules: I define how a result is judged before it reaches a client. The judgement is written down, so it doesn't depend on the memory of whoever was there.

First level · Compliance

The rule is code

Colours, spacing and components are values and constraints, not suggestions. Anything that falls outside the system shows up immediately, and is fixed at the source instead of in every single file.

Second level · Quality

An explicit rubric

Hierarchy, legibility, balance, data treatment. Every result is assessed on named criteria rather than impressions. The same criteria apply to me, to your team and to the AI.

Third level · Taste

An archive of choices

Approved and discarded examples are kept together with the reason. That's how the system learns to tell what is correct from what is good.

How we work

Three steps. You can stop after any of them.

None of the three commits you to the next. If you decide not to continue after the audit, the document you have is still useful.

Starting point 01

The identity already exists

I start from the material you have: brand book, website, presentations, reports. I draw out the rules you already follow without having written them down, and turn them into a system the team can use. This is the most common path.

Starting point 02

The identity needs to be created or renewed

I design the visual language and build it as a working system from the start. This applies to a new product, a spin-off or a rebrand: the tools are created together with the identity, instead of arriving years later.

Step 01 · Diagnosis

Brand System Audit

Two to three weeks

I look at all the material you already have and tell you what can be systematised, what needs redoing and what is best left as it is.

  • Analysis of the visual assets
  • Recurring principles made explicit
  • Roadmap with priorities and timings

What I need from youAccess to the files and two hours with whoever knows the brand best.

Step 02 · Build

AI Brand System

Six to ten weeks

I build the library everything starts from, and the instructions artificial intelligence reads before producing, so it works within your rules.

  • The basic decisions, the pieces, the pages
  • Browsable, versioned library
  • Context and rules for the AI
  • Assessment rubric
  • Team training

What I need from youOne point of contact and about an hour a week for reviews.

Step 03 · Continuity

System Evolution

Monthly retainer

A system that stands still ages. Every month new formats come in, the archive of choices is updated and the team's production stays under control.

  • New components and templates
  • Quality control on outputs
  • Archive updates
  • Support for the team

What I need from youYour requests, as they come. No fixed meetings.

Inside the three steps: the five phases of the method

Phase 01Auditgather
Phase 02Readingunderstand
Phase 03Encodingencode
Phase 04Activationactivate
Phase 05Traininghand over

Good to know

Timings depend on how much material already exists and are set after the audit. Bespoke creative work, such as campaigns and one-off projects, stays outside the retainer and is quoted separately: the system frees up time for that work and doesn't replace it.

Who it's for

When it makes sense, and when it doesn't.

It makes sense if

  • You publish something every week, on more than one channel.
  • Someone inside the company already produces materials.
  • You have an identity to start from, even an imperfect one.
  • You want to use artificial intelligence without it showing.
  • You depend on an outside supplier even for minor changes.

It doesn't make sense if

  • You need a logo, and for now only that.
  • You have one campaign to run and that's it.
  • Nobody in the company produces materials, and nobody will.
  • You'd rather keep handing every single file to me.

For agencies

Do you work in an agency or a studio?

The system adds to the identity projects you already do, and extends what you deliver without new hires. The client stays yours: I come in as a specialist, in whatever set-up we agree on together.

Projects worth more

The identity you deliver becomes a system the client uses every day, and the scope of the project grows.

Clients who come back

An identity made years ago becomes a concrete reason to reopen the conversation, without waiting for the next rebrand.

Recurring revenue

Maintaining and evolving the system creates continuity, replacing the repetitive work that eats into margins.

Questions

What people ask me before we start.

We already have a brand book. Isn't that enough?

A brand book describes, and it will still be needed: it becomes the source the work starts from. What it can't do is prevent mistakes while someone is working in a hurry. The system can, because the rule lives inside the files that generate the materials.

Isn't uploading the brand book to ChatGPT enough?

It's the first thing people try; I did too. But the model reads a document designed to be looked at by a person: it guesses the colours from the images instead of having the exact values, and it weighs the example pages like everything else. From the third piece onwards it starts going its own way. What you need is the same content rewritten in a form a machine reads without interpreting: declared values, ready-made components, examples judged one by one.

Our team isn't technical.

It doesn't need to be. Whoever produces content opens a library and a back office: picks a block, writes, publishes. The code sits underneath and concerns me and the developers.

Doesn't artificial intelligence risk distorting the brand?

Yes, when it works without constraints, which is what happens today in half the companies that use it. Here it receives named rules, correct examples, rejected examples and assessment criteria. We decide the direction; it applies it.

Where do our materials end up?

The system lives in your repository, not mine. The files you send me stay in a shared folder that you close at the end, and I sign a non-disclosure agreement if needed. As for models: we use the ones your company has already approved, with settings that exclude training on your data. If your legal team has specific constraints, we review them before the audit so they don't become a problem halfway through.

How much does it cost?

It depends on how much material exists and how many formats need to go into the system, so any figure before I've seen your files would be made up. The audit is the smallest step and exists precisely for this: at the end you have a real number and you're free to stop.

How much time do we need to put in?

Two hours during the audit, about an hour a week during setup. After that, only the time to send me your requests. Most of the work stays on my side.

What happens if we stop working together?

The system is yours. Values, components, documentation and archive stay in your repository, in open formats any studio can pick up. The retainer is renewed because it's useful; you are never locked in.

About me

The path that led to this system.

I'm Alessio Granella, brand and graphic designer. Before working as an outside consultant I spent ten years inside a software company: I saw first-hand how presentations drifted apart and templates went un-updated.

Today I work with marketing departments and companies that need to communicate every week.

Since artificial intelligence entered production, I no longer deliver just files. I build the system that lets you produce more without me, and I teach that system to recognise when a result isn't good enough.

By your side, every step of the way.

Let's start

Half an hour to see if it makes sense.

Half an hour, no commitment. We look together at what you produce today and where time gets lost. If the system isn't the right answer for you, I'll tell you on the same call.

  1. 01You write to me and we book half an hour.
  2. 02I look at the material you already have.
  3. 03I tell you what can be systematised and what can't.
  4. 04If it makes sense, we start with the audit.

Drop me a line

I'll reply personally within one working day.

Prefer your own email? info@alessiogranella.com

Thank you, speak soon.

I've received your message. I'll reply within one working day, from this address: info@alessiogranella.com.

Not the right time for a call? Send me your website link and I'll reply with a couple of lines on what I would systematise first. You don't need to prepare anything.

Send me the link ↗