When faced with the challenge of inconsistent messaging across Global Player, I built a custom AI tool that transformed our team’s copywriting process—streamlining tone, aligning outputs across product squads, and optimizing the overall refinement and approval process.
In the fast-paced world of UX design, creating consistent, user-centered copy is crucial but often time-consuming. My custom GPT was designed to act as a UX Copywriter for Global Player. I armed it with the mission of helping our product and design teams craft high-quality UX copy, helping us craft messaging that resonates with our user’s diverse needs while staying true to each of our brands and their unique tone of voices. This case study dives into the journey of building, refining, and evolving the tool, its impact on our workflow, and the key lessons learned.
Global Player is the home of the UK’s leading Radio Brands, including Heart, Capital, Radio X, LBC and more. It supports 8 distinct brands voices within one app, each with their own identity and audience. This diversity created one of our biggest challenges: finding consistency in tone of voice and messaging across the entire app. The copy often felt disjointed with different product designers and product owners working on various features across all brands. Ensuring that all written content aligns with our overarching brand guidelines while resonating with our diverse user base became a struggle. The process of producing user-centric, accessible copy became a bottleneck in our design process.
To help us all out, I sought to develop a Custom AI UX Copywriter assistant, a custom GPT that could generate on-brand, user-centric copy quickly and consistently. The goal was to empower our team to align our voices across the product while significantly reducing the time spent on revisions and copy approval, by providing a strong and reliable starting point for crafting excellent and consistent UX copy.
I gathered some initial feedback from our team to get a clearer picture of the problem.
Here’s a couple of quotes I collected from my colleagues:
We often end up with different tones across different screens. It’s hard to maintain consistency when multiple people are involved in copywriting.
James T-B.
Senior Product Designer
The copy approval process drags on for too long, with multiple revisions needed. It really slows things down.
Daniel P.
Product Owner
I already had my personal reasons for building this. Still, it was clear that a Custom AI UX Copywriter would help to address these challenges and improve consistency and copy quality across the app.
I wanted to make a global impact across the company, aligning other departments with this tool to bring consistency across how we talk about our product outside of the app too. Sadly, that’s not yet happened, but I’m actively working toward building trust in it and raising awareness of its availability across the org.
To kick off the project, I followed my usual discovery process to clearly identify the problem areas. I knew the AI copywriting tool was something I initially needed for personal use, but if I was going to share it across the company and make a wider impact, it had to solve more than just my own challenges—it had to address the issues faced by other stakeholders involved in the copywriting process.
I took a holistic approach, identifying all the stakeholders who contributed to the creation of style guides and UX copy across different product squads. I collected the various resources these teams referenced and gathered insights into the pain points they were experiencing with inconsistent messaging.
Since there wasn’t any quantitative tracking in place, I focused heavily on qualitative data collection. I conducted informal interviews and gathered feedback from key team members to better understand how different voices and squads approached copywriting. The main problem was achieving alignment across multiple voices while maintaining consistency across the app.
Tools Used:
Chat GPT Pro License – This served as the foundation for building and training the custom AI UX Copywriter.
Miro – Used for mapping out workflows, stakeholder alignment, and visualizing the current state of our copywriting process.
Notion – For documenting the project journey, storing resources, and tracking iterations as the tool evolved.
Medium Articles and Industry Resources – To stay up-to-date with best practices in AI and copywriting tools and integrate those learnings into the development of the tool.
To ensure the tool addressed the needs of different roles within our team, I defined the following three personas each with unique pain points that my Custom AI UX Copywriter tool aimed to address, based on real feedback:
Focused on crafting the user journey but often spending too much time trying to match the tone and voice of the app when creating copy.
Balancing feature launches and stakeholder expectations, needing clear, on-brand copy without extensive rounds of revisions.
Ensuring all UX copy aligns with the brand, often bogged down in lengthy approval cycles.
Solo Experimentation
When Global announced it was piloting Chat GPT, I knew I had to jump in. Until then, I’d been paying for GPT Pro out of pocket, anonymizing company information just to safely experiment with it. Now, with an official Global account, I could dive in without those limitations.
Over the past quarter, I’d fully immersed myself in learning everything about LLMs, experimenting with AI tools, and testing various custom copywriting GPTs. None of them delivered what I was looking for. I didn’t just want a copywriter—I wanted a tool that understood the strategy behind the messaging and could evolve based on feedback. So, I took it upon myself to build exactly that.
I started by compiling all the guidelines and resources I typically rely on when crafting UX copy. Then I began testing GPT with basic conversations, pushing its limits by feeding it chaotic typing, vague product descriptions, and gradually familiarizing it with Global Player’s voice and style.
With each interaction, I refined the tool, ensuring its output consistently aligned with our brand guidelines. Over time, it got better at generating relevant, on-brand copy. That's when I began to build real confidence in its abilities and felt it was finally ready to be shared with my fellow product designers for feedback.
One of my side motivations for creating this tool came from working with two teammates who have dyslexia. Whenever they present mockups, there are almost always a few copy mistakes. Because one of my ADHD hyper-fixation quirks is spelling errors—it physically hurts my brain to ignore them. My constant corrections started to annoy them, though they understood I couldn’t just let it slide.
So, to create harmony amongst our neuro-atypical brains, I built this tool with them in mind, saving them from my “constant bullying” over copy mistakes. (All in good humour, I promise!)
Team Rollout and Refinement
Once the tool was delivering solid output, I introduced it to the wider product team. The goal was to create more consistency in how each squad produced copy, regardless of who was writing it. We went through several feedback sessions to fine-tune the tool, ensuring it understood our brand context and UX copy needs.
It’s like having a written design system in that it applies our brand guides into everything we write. The tool really speeds up our process, and the app messaging now reads more consistently.
Josh M.
Senior Product Designer
Continuous Feedback Loop
The tool quickly became a collaborative resource. Our team fed it direct stakeholder feedback, enabling it to improve with each iteration. This feedback loop drastically reduced the number of rounds needed for copy approval—from 4-5 rounds down to just 1 or 2. This helped minimise time spent on last-minute copy changes, saving valuable time and pushing products along quicker.
Over these months, we went from a general tone-of-voice setup to a refined, psychologically informed, inclusive, and user-friendly UX copy.
My key focuses have been:
Establishing a tone of voice that was conversational but clear.
Introducing psychological principles to influence user behaviour.
Refining error messaging to be more neutral and factual.
Fostering inclusivity and feedback loops to encourage user engagement.
Simplifying copy to reduce cognitive load and ensure clarity.
Introducing motivational and personalisation elements to encourage feature exploration.
Each refinement has made the copy more user-centric, clear, and empowering, ensuring Global Player provides a seamless, engaging experience across its platform.
1. Initial Refinement – Defining the Tone of Voice (May 2024)
Goal: Establish a foundational understanding of Global Player’s tone of voice (TOV) for UX copywriting.
Focus: Conversational, warm, friendly, knowledgeable, non-elitist communication style for UX copy. This stage defined how we would approach copy in a manner that aligns with Global Player's core values.
Outcome: We created initial templates for onboarding messages, feature notifications, and error messaging.
Example:
Onboarding: “Welcome to Global Player! Your home for live radio and podcasts.”
Error Message: “Something went wrong. We couldn’t load this content. Please try again.”
This was the first major phase where we integrated the Global Player TOV guide I provided.
2. Introduction of Psychological Principles (June 2024)
Focus: Conversational, warm, friendly, knowledgeable, non-elitist communication style for UX copy. This stage defined how we would approach copy in a manner that aligns with Global Player's core values.
We explored 106 cognitive biases (from the Psychology of Design document), such as:
Zeigarnik Effect: Emphasised uncompleted tasks to encourage users to return (e.g., “You stopped at [Podcast Name]. Want to continue?”).
Simplified complex tasks and avoided overwhelming users with too much information.
This stage added layers of psychological insights to UX copy to subtly influence user decisions, improving overall user engagement by ensuring the copy was both clear and motivating. This session marked the second stage of refinement.
3. Internal Team Launch & Error Messaging Evolution (July 2024)
Goal: Improve error messaging by making it clearer and more empathetic without being overly formal or technical.
Focus: Based on user feedback and our psychological principles, error messaging was refined to:
Be concise and non-technical.
Offer actionable steps for users without frustration.
Avoid formal and distant language.
Outcome: We moved from “Oops, hit a snag!” to clearer and more neutral messaging like “Something went wrong. Please try again later.”
Before
Oops, we've hit a snag!
After
We’re having trouble connecting to the server. Please check your connection or try again later.
This refinement session shifted how errors were communicated, moving toward a more neutral and informative tone, focusing on reducing cognitive load and user frustration.
4. Inclusivity and Feedback Loop Refinement (August 2024)
Goal: Ensure the UX copy is inclusive and encourages user feedback in an engaging way.
Focus:
Inclusivity: We worked on promoting shared experiences and ensuring that copy respects user diversity.
Feedback loops: Enhanced feedback prompts, making users feel like their input mattered, and providing reasons for giving feedback.
Outcome: The tone became more inclusive, focusing on making users feel like contributors to the Global Player ecosystem.
Before
Help us improve.
After
Help us shape the future of Global Player by sharing your feedback.
This session evolved the copy to feel more inclusive and community-driven, incorporating psychological principles such as Social Identity Theory to promote a sense of belonging.
5. Personalisation and Feature Rollout Messaging (September 2024)
Goal: Make messages more personalised and engaging, especially when introducing new features.
Focus: We refined copy for new features (like playlists, widgets, and navigation) using the Mere-Exposure Effect for familiarity, so users were more inclined to use new features.
Example: “Meet your new Global Home! Everything you love in one place—explore now.”
Outcome: We implemented motivational copy for new features, using clear, empowering language to encourage exploration.
This session added a motivational layer to the copy, where we introduced features with enthusiasm and encouragement.
6. Clarity and Cognitive Load Reduction (October 2024)
Goal: Ensure the copy is as clear and user-friendly as possible to reduce cognitive load and prevent user frustration.
Focus:
We honed in on simplicity and clarity across all interaction points, ensuring that users could easily understand the messaging.
Leveraged the Fitts’s Law and Cognitive Load Theory to structure the copy in a way that guided users smoothly through the app.
Outcome: The copy became even more streamlined, ensuring that users could complete tasks effortlessly.
Before
Please enter your date of birth.
After
Enter your date of birth (DD/MM/YYYY). For example, 31/12/1990.
This session solidified our focus on reducing friction and making instructions more specific and actionable.
Since introducing the new tool, the team’s copywriting workflow has transformed by giving us a strong start that is consistent with our parent brand’s tone of voice. I gathered some guerilla feedback from key team members:
Here are the overall improvements identified so far:
Consistency: The tool ensures that all UX copy follows the same tone, maintaining brand alignment across different features.
Efficiency: We’ve reduced the time spent on copy approval, allowing the team to move faster without sacrificing quality.
Collaboration: The tool has become an integral part of our feedback loop, learning from stakeholder input and improving its copy suggestions with every iteration.
Building the Custom AI UX Copywriter has been a transformative experience, allowing us to streamline the copywriting process, reduce time spent on revisions, and create more consistent UX copy across our app. This tool will continues to evolve the more it gets used, and future improvements will focus on refining its adaptability for niche content needs and improving its integration with other research and design tools.
Ultimately, this project showed me how AI can support internal workflows. No doubt comes with enabling teams to work more efficiently and produce higher-quality results. As with all LLMs, MLs, and AIs, I believe we as professionals have the moral obligation to refine our work with a human-centred approach. Avoid copying & pasting straight out of GPT, and make sure you’re challenging its output and telling it when it’s wrong.
What Went Well 😊
There wasn't any hard deadlines for this project, but it was something I set as a personal development goal. This allowed me to approach this during the freestyle time our department has given us every Friday from 2-5pm. This made the project really fun and purely experimental until I found a clear value that was worth sharing with others.
While building the Custom AI UX Copywriter, I encountered several challenges:
Earlier Testing: I initially focused heavily on functionality and spent way too long testing alone. I could have introduced structured testing with my teammates earlier on to catch issues related to tone and context.
Expanding Features Too Quickly: Expanding the tool’s functionality too fast, especially with A/B testing features, created complexity. Slowing down and validating each new feature more carefully would have resulted in smoother rollouts.
Handling edge cases: Early on, the AI struggled with niche content needs, requiring more manual intervention than expected. To address this, I had to feed it specific edge-case scenarios in response to its output, helping it recognize where it was falling short. This process allowed the GPT to learn from its mistakes, refining its ability to generate more thoughtful and accurate copy. By providing examples of existing discrepancies and broken paths in Global Player, the AI gained a better understanding of edge cases and adjusted its output accordingly.
I’ve been using the tool almost every day, now, discovering unique ways to leverage it—like workshopping A/B messaging experiments and ideas using applied psychology principles. I’ve also been feeding it new research assumptions based on our user data. It’s helped us communicate how our features work more effectively and yield some exciting conversion metrics from a recent in-app messaging campaign announcing a new customization feature (though those results will remain private!).
Thanks for reading. :)