Scaling customer support processes
Jordan Daley
Cuvva is a UK-based insurtech, creating a new type of temporary car insurance. You can insure yourself on someone else’s car in just a few taps—no long forms or call centre queues.
Cuvva has excellent in-app support.
They run a 24/7 live chat with an average response time of 1 minute. It’s not all plain sailing though:
Insurance is complicated. There are loads of different criteria and pricing metrics. And the app that has the final say.
Conversations are often handled by multiple agents. This allows for shift changes, lunch breaks, and team meetings.
Agents have varying levels of experience. And there’s not always a manager on shift.
The solution? A huge library of saved replies and support articles.
But when I joined Cuvva, support content was:
Not written by writers 🤷🏽♂️
So they often contained the type of insurance-y jargon that Cuvva really wanted to avoid.
Poorly organised 🤯
Agents couldn’t always find the responses they needed, so they’d wing it, or add their own to the library. This resulted in poor message discipline, and lower chat ratings.
Not up to date 🕰
Outdated and inaccurate information caused confusion and complaints.
Not a great source of data 📉
We were missing opportunities to track common topics and find ways to improve our service.
My process
Audit and rewrite 🧐
I started by having a look at support metrics, customer reviews, and article click-through rates
then I interviewed the support team, product team, and developers
I used this data to create process maps and conversation flows
then I rewrote every saved reply (and most knowledge base articles) according to plain language principles
A common misconception within the team was that support articles must be long and read like a story. In reality, shorter articles perform better — as long as they answer the question in the title and any related FAQs.
Scale the process 📈
I worked with our Head of Customer Support to create a CX content squad
I trained squad members on Plain Language guidelines and created a QA system to ensure that all new replies get checked for tone, accuracy, and FCA compliance
to keep the team motivated (and make sure they were rewarded for their extra work!) I pushed to make reply squad work count towards promotions and raises
I also asked PMs, designers, and underwriters to check which saved replies are affected by upcoming changes, and let Customer Operations know what needs to be changed
Improve taxonomy and hierarchy 🗂
I worked with the customer support team to create a naming system for saved replies
by sending out surveys and polls, I could work out which search functions and keywords agents used to find replies
Create resources 📄
finally, I made a style guide for all customer-facing content—support, saved replies, and social media responses
Impact
“OK, great… but why does it matter?”
Fewer conversations ⬇️
We saw a marked ticket reduction in areas where we improved our articles — especially when we collaborated with SEO to improve reach.
Lower response times ⬇️
support team could now find responses faster, and spent less time typing out replies as we had filled most of the gaps in our library.
Higher customer satisfaction scores ⬆️
Less confusion and mixed messages.
Better onboarding and training outcomes 🎓
New joiners needed less supervision online. They could use Saved Replies to learn more about our processes.
More data 📈
Our data team started tagging and tracking saved replies. They used the data to identify trending topics, measure the impact of projects, and even spot bugs in the app.
Like this project
Posted Sep 10, 2025
How I helped overhaul a UK-based Insurtech's support processes with improved support content, saved replies, and team processes.
