Translating Brand Standards Into AI Workflows

AI image tools arrived faster than our visual standards could keep up with.

I turned Xello’s photography standards into a guided AI-assisted system, with prompt structures, visual governance and creative direction, so teams could create on-brand imagery with confidence.

Industry

B2B EdTech SaaS

What I did

AI workflow design

Visual governance

Photography direction

My role

Senior Brand and Marketing Designer

Timeline

Strategic initiative, 2026

Worked with

Director of Brand and Communications

The challenge

Everyone could make images. Not everyone could make on-brand ones.

As AI image generation became more accessible across marketing teams, keeping imagery consistent got harder. Xello needed a structured way to guide non-designers through AI image creation while protecting its established photography direction, emotional tone and visual language.

The challenge

Everyone could make images. Not everyone could make on-brand ones.

As AI image generation became more accessible across marketing teams, keeping imagery consistent got harder. Xello needed a structured way to guide non-designers through AI image creation while protecting its established photography direction, emotional tone and visual language.

The challenge

Everyone could make images. Not everyone could make on-brand ones.

As AI image generation became more accessible across marketing teams, keeping imagery consistent got harder. Xello needed a structured way to guide non-designers through AI image creation while protecting its established photography direction, emotional tone and visual language.

What I noticed

The problem wasn’t generating images.

It was defining the structures, behaviours and visual parameters that let teams move faster while still creating work that felt like Xello.

What I noticed

The problem wasn’t generating images.

It was defining the structures, behaviours and visual parameters that let teams move faster while still creating work that felt like Xello.

What I noticed

The problem wasn’t generating images.

It was defining the structures, behaviours and visual parameters that let teams move faster while still creating work that felt like Xello.

What we built

Rather than unrestricted image generation, the system focused on governance and consistency.

1

A prompt framework.

A modular structure that turns brand principles into repeatable visual direction: subject, action, environment, lighting, mood.

2

A brand photo assistant.

A custom GPT trained on the photography toolkit, helping anyone describe, create or check an on-brand image.

3

Audience-specific visual systems.

Emotional tone, environments and storytelling tuned to different learner journeys.

4

A scenario library.

Curated scenes, from counselor conversations to skilled trades workshops, so teams start from something that already works.

Five parts, each with a job: who, where, how it feels, how it looks and who it represents.

Five parts, each with a job: who, where, how it feels, how it looks and who it represents.

Brand Photo Assistant Lizzie, built on the photography toolkit with starters for people who don't know where to begin.

Brand Photo Assistant Lizzie, built on the photography toolkit with starters for people who don't know where to begin.

Eight audience cohorts, from district leaders to early career learners, each with its own tone, setting and palette. Needs the full width to stay readable.

Eight audience cohorts, from district leaders to early career learners, each with its own tone, setting and palette. Needs the full width to stay readable.

Core scenes, from counselor conversations to skilled trades workshops.

Core scenes, from counselor conversations to skilled trades workshops.

 Moments across the student journey, with the visual language and colour accents built in.

Moments across the student journey, with the visual language and colour accents built in.

What it included

The system

AI visual governance

Prompt systems

Photography direction

Brand consistency framework

What I delivered

AI prompt library

GPT assistant framework

Photography standards

Reference image boards

Skills I used

AI workflows

Visual governance

Brand systems

Creative direction

How it landed

Teams gained a structured, scalable way to create AI-assisted imagery that stayed consistent across marketing outputs, and could adopt new tools with confidence without fragmenting the brand.

Framework imagery in use on the homepage, across desktop, tablet and mobile.

Framework imagery in use on the homepage, across desktop, tablet and mobile.

What worked

Building the system around structure instead of individual prompts. Once the five parts of the framework were set, the assistant, the audience cohorts and the scenario library all came from the same logic, so a good result was repeatable rather than lucky.

What worked

Building the system around structure instead of individual prompts. Once the five parts of the framework were set, the assistant, the audience cohorts and the scenario library all came from the same logic, so a good result was repeatable rather than lucky.

Looking back

What worked

Building the system around structure instead of individual prompts. Once the five parts of the framework were set, the assistant, the audience cohorts and the scenario library all came from the same logic, so a good result was repeatable rather than lucky.

What I'd do differently

Test it with more people before scaling. The pilot proved the approach with one reviewer. Next I would run it with marketers who had never seen the toolkit, track which prompts still drifted off-brand, and refine the assistant from what they struggled with.

Looking back

What worked

Building the system around structure instead of individual prompts. Once the five parts of the framework were set, the assistant, the audience cohorts and the scenario library all came from the same logic, so a good result was repeatable rather than lucky.

What I'd do differently

Test it with more people before scaling. The pilot proved the approach with one reviewer. Next I would run it with marketers who had never seen the toolkit, track which prompts still drifted off-brand, and refine the assistant from what they struggled with.

Looking back

What worked

Building the system around structure instead of individual prompts. Once the five parts of the framework were set, the assistant, the audience cohorts and the scenario library all came from the same logic, so a good result was repeatable rather than lucky.

What I'd do differently

Test it with more people before scaling. The pilot proved the approach with one reviewer. Next I would run it with marketers who had never seen the toolkit, track which prompts still drifted off-brand, and refine the assistant from what they struggled with.

What I took away

As AI scales, governance matters more, not less.

What I took away

As AI scales, governance matters more, not less.

What I took away

As AI scales, governance matters more, not less.