The potential of intelligent design and large language models (LLMs) in digital transformation.
Joshua Healey, Principal Consultant at Hippo explores the emergence of artificial intelligence and large language models, and what it means for content design in this three-part blog series.
The relationship between artificial intelligence and content design
Content designers are a well established and sought after addition to digital teams.
They combine the power of language with a range of design patterns, tools and systems to help users:
- Do the things they need to do
- Complete actions and task
- Find and digest information
From making services accessible to spotting where small mistakes might have big consequences, a content designer plays an important role in user-centred decision making and improving positive outcomes.
It turns out that artificial intelligence (AI) may be able to do this too. And do it well.
When the latest adverts for laptops are now ‘powered by AI’ and our business messaging software can summarise your 14 unread notifications in an instant – it can all feel a bit futuristic and unnerving. But as a set of technologies – AI has been around for a while.
Like many pioneering innovations before it – over time there has been a shift in both its public visibility and operational maturity. With that shift we have seen the art of the possible grow exponentially; from automating manual processes to informing the prioritisation of tasks in clinical settings – when used responsibly there is a lot that it can do.
So, AI is great.
But is it ready to replace your designer?
The relationships between language and content design
Every content designer will tell you that they are not a writer – and that’s true. It’s really true. That said, there’s no escaping the fact that a content designer’s command of language is one of the biggest levers they can pull to meaningfully affect positive user outcomes.
Typically, the skill of content design in the context of building products and services is to say a lot without saying a lot. Unless you are writing poetry or prose, you need to help people understand what to do and then how to do it with as little cognitive effort as possible. Good services should feel effortless – gliding you from where you are to where you need to be.
For example, a content designer might recommend that a user experience could be improved by:
- Re-ordering the sequence of a journey
- Introducing an in-page calculator
- Using a check your answers design pattern
But you can also be assured that simplified language, which has been through dozens of rounds of research and usability testing, will be an ever-present constant along the evolution of the service design.
ChatGPT4: a large language model with great potential
I could explain what a large language model is – but I thought I would let one explain itself.
GOV.UK recommends to write as if your audience has a reading age of 9 and this has become a common benchmark across digital services. So, I asked ChatGPT4 to explain a large language mode to me like I was a 9 year old. This is what it said:
- ‘’Imagine you have a huge library full of books, and each book has a different story or fact inside. Now, let’s say you want to ask the library a question, like “Why is the sky blue?” A large language model (like me!) is like a super-smart librarian who has read every single book in that library.
- Instead of searching through each book one by one, this librarian has learned patterns in the stories and facts, so when you ask a question, it can quickly pull together the best answers from what it has learned. It doesn’t always know everything perfectly yet, but it tries to give a good answer based on all the stuff it has read.
- So, a large language model is like a really big brain that learned how to talk, understand questions, and give answers by reading a ton of books’’
Given that it took a second to produce that, I was quite impressed.
What do large language models and AI have to do with content design?
We’ve acknowledged that content designers deliver a lot more than words – but words are perhaps one of their most powerful tools . Given that AI powered LLMs like ChatGPT are becoming more popular it is hard not to observe their potential to operate in the same arena. Interestingly, in its own description of itself it doesn’t actually mention words that much. It talks more about interpreting and using common patterns to try and get the best answer.
‘’Instead of searching through each book one by one, this librarian has learned patterns in the stories and facts, so when you ask a question, it can quickly pull together the best answers from what it has learned. It doesn’t always know everything perfectly yet, but it tries to give a good answer based on all the stuff it has read.’’
This almost feels like how we use design systems. As human beings.
Typically, when we are trying to find the right solution for a user-experience challenge our first instinct isn’t to trawl through dozens of individual products and services looking for what might work in our use case. Our first instinct should be to look at tried and tested design systems within or adjacent to our area of work as there is a good chance all of (or some of) the problem has arisen, been worked on and answered before.
GOV.UK describes design systems as essentially a place which brings together lots of patterns and styles from across the digital landscape which have been developed upon and tested by many different teams.
So if you afford a little bit of poetic licence – how is an LLM like ChatGPT forming answers derived from the collective patterns, themes and answers of multiple related inputs any different from a content designer championing and using a design system born from the same principles?
- The Design System contains styles, components and patterns to help teams in government create user-centred digital services. It brings together the patterns and code found in the Service Manual, GOV.UK Elements and elsewhere.
We’ve developed the Design System by working with teams across the government. We’ve also been building a cross-government community of Design Systems users and contributors who will help to maintain and develop it.
An experiment to see if AI can replace your content designer
What started as a thought experiment quickly turned into something bigger: a practical exploration that’s now shaped a series of blogs.
In the next post in this series, I’ll look at a fictional NHS service as a case study to unpack a common challenge in digital design: what happens when a content designer isn’t involved early on? The Blood Test by GP Referral Only Service—a pretend service designed for illustrative purposes—was midway through its Alpha phase, juggling tight budgets, evolving requirements, and a need to realign with user needs. We’ll explore how missing roles, limited accessibility research, and gaps in content design left critical aspects, like the service’s name and start page, in urgent need of attention.
