Scaling content beyond myself with AI
I was the only content designer in the department and the content queue was growing faster than I could clear it. I built an in-product UX writing tool and a self-serve campaign workflow so designers and commercial teams could draft and ship their own copy, with a review step for anything high stakes.
260+
chats in the in-product writing GPT during first six months
20+
colleagues using the tool: most of the designers plus the commercial team
1–2 hours
saved per campaign, estimated
Most used
internal GPT at ClearScore in 2025 and 2026
Problem: The content bottleneck
I was the only content designer in the department. Every squad wanted copy from me, on top of BAU work within my own squad. The requests were arriving faster than I could clear them.
Designers and commercial colleagues had started using AI tools to write their own. The output ignored our tone of voice, content principles and the rules we work under as a regulated financial firm, so it came back to me anyway, now with a rewrite attached.
Opportunity: Build an AI tool for colleagues to write high quality copy themselves
Instead of finding ways for me to write more, I focused on how I could use AI to allow others to write high quality copy. I built two AI tools loaded with our tone of voice guide, content style guide, UI component rules and compliance guidance, so a colleague gets high quality copy that already fits the component. I also managed a Contentful workflow that alerted me for any high priority copy that needed my review.
A tool nobody knows how to use is a tool nobody uses. I wrote the documentation, trained the eight-person commercial team in person and demoed it to the product designers, so people had seen it work before they were asked to rely on it.
A screenshot of the GPT tool I created with files to reference tone of voice, our system design components, content style guide and more.
How we would merchandise financial offers and how it works now
Before
A commercial colleague would brief me for a smaller, one off credit card, loan and car finance offer. I wrote the merchandising card copy, built the Contentful entry, reviewed with the compliance team and then publish. Every one-off campaign in the business queued behind me and would take a few hours.
Now
Commercial colleagues pick from a bank of approved, ClearScore tone-of-voice copy for credit card, loan and car finance merchandising. If they need something the bank doesn't cover, they draft it in the writing tool. They build the Contentful entry themselves, and the workflow flags it to me for review before anything goes live. Three colleagues have shipped 11 campaigns so far in 2026.
What makes it a system rather than a shortcut
The quality gate didn't disappear, it moved. I still see every entry before publish, but as a reviewer rather than the author, which takes minutes instead of days. The copy bank means most campaigns never need new words at all, and the tool means the ones that do start from our voice rather than a blank page.
This process I have built and managed has been a key factor in scaling content beyond myself.
What I chose and what I ruled out
Rejected
Keeping or 'improving' the bottleneck
Continuing to own every piece of copy manually would have slowed down the output from our department and kept me from developing my role into strategic work as an IC4.
Rejected
Full autonomy without review
There are some areas where content is too high stakes for output to not be reviewed, especially as finance is heavily regulated. We needed a speed increase, not a risk increase.
Chose
Move to AI for scale
I built it in ChatGPT first because that's what designers were already using, and launched in September 2025. By January adoption was high enough to move the same framework to Claude as a skill, so designers with Claude licences could use it too.
Chose
Keep the bar high for quality output
Faster production doesn't mean lower standards. I continuously tweaked the system prompt and tested on new models as they were released, to ensure output is met the bar again. The tools are maintained, not launched and left.
What I shipped
The in-product writing tool. Designers paste a screen or a brief and get copy in ClearScore's voice, with the component rules already applied.
Launched in ChatGPT in September 2025, moved to Claude in January 2026 once adoption was high. It carries our tone of voice guide, style guide, component rules and compliance guidance, so the first draft already sounds like us.
In-product writing AI tool
The Notion agent for the Content Design System. Colleagues ask it a question instead of reading multiple pages to find the answer.
In-product writing AI tool
Notion AI agent for Content Design System
Launched in ChatGPT in September 2025, moved to Claude in January 2026 once adoption was high. It carries our tone of voice guide, style guide, component rules and compliance guidance, so the first draft already sounds like us.
I made the case to IT for a Notion Custom Agent with a deliberately tight scope: answer questions from our own documented guidance, nothing else. Colleagues ask whether they can use emojis or how to structure headings and get an answer grounded in our system.
Self-serve campaign workflow
A bank of compliant, on-brand copy for card, loan and car finance merchandising, plus templates for building a Contentful entry safely. Commercial colleagues build and I review before publish if necessary.
Results
The copy that reached me was now the high stakes work the Contentful review flow was built to catch. The rest shipped without me, which freed me for strategic work such as campaign efficiency and the research repository working group.
I saw over 260 chats within the internal GPT tool, adopted by 20 colleagues in the first few months, saving the company 1–2 hours per campaign. Nothing has shipped through the tool that compliance later raised or questioned.
"It effectively demonstrates you think in enablement and operational scale: reducing bottlenecks, improving quality across teams and enabling people to self serve."
"Jordan was always ready and available to support me at any moment, ensuring it was a seamless experience. I now know and feel confident in how to do it and can pass this knowledge on myself."
Anne Marie Greig
Head of Content Architecture
Ruairi Walker
Commercial Executive
What I learned
The build was the easy and exciting part. It was a fun way to upskill myself in AI and also see positive results.
Getting 20+ people to change how they ask for copy took longer, and most of it happened in conversations, training sessions, documentation and not in the tool itself. If I did this again I'd launch the training on the same day as the tool, not after it in an async Slack message.