Large language models (LLM) of AI, like Chat GPT 4, have been trained to research almost any subject and present it to users in a tailored way. 

The Government Digital Services’ definition of a content designer is that they ‘take a user need and present it in the best way possible’. 

On a basic level, LLM AI models share a similar approach to how we work as content designers. 

This might seem like cause for existential conversations about the impact AI will have on content and user-centred design practitioners, but at Hippo, we’re a little more pragmatic. 

We make decisions after a lot of research and investigation.  

Principal consultant Josh Healey has already explored LLMs in great detail in his three-part Hippo blog series. The series is essential reading for content designers, but the main takeaway is that the potential for aiding our roles is huge. 

Like people, LLMs require decent explanations to provide the best outputs. These are called prompts. ‘Prompt engineering’ is one of many in-demand human skills that AI requires. Essentially, this means shaping outputs by tailoring inputs. 

The good news is that content designers are well-practised at refining and shaping instructions for users to create desired outcomes. Only in this case, we’re treating the LLM as our user. 

A prompt engineering exercise

We wanted to put this practice to the test, so Josh offered to lend his expertise to one of our regular ContentWise Leeds sessions. These in-person sessions are for content and user-centred design practitioners and anyone with an interest in digital.

Our session, ‘Become a prompt engineer’ attracted content and service designers, along with lecturers, legal professionals and more. All the audience had roles that required a greater understanding of AI, and the mix of attendees demonstrates the large-scale impact of AI.

After an explanation of how LLMs work, we worked on a prompt exercise. Using a pre-determined brief, we had to create a message for rail customers, informing them that signal failure had caused a delay.

To create the best prompts, we had to use instructions that offered:

  • clairity
  • specificity
  • context
  • instruction
  • constraints
  • audience framing

It’s a long and potentially confusing list, but it helps to imagine your LLM is a very smart and enthusiastic person. So an equal mix of setting goals, such as channel, background and audience, and establishing parameters, like length and complexity, is essential. It’s a method of thinking that should be familiar to content designers. Often in our practice, it’s as much about what we remove as what we add.

The outputs

Despite similar objectives, it was fascinating to see the outputs. Some used a letter as a style guide, which was great for tone and brand voice, but the output was of letter-sized length. Our team specified a short message, and Chat GPT 4 reasoned we were creating a post that could be used on social media, offering to tailor it, amend it for email and alter the tone of voice.

Ethics and impacts 

We moved on to discuss the ethics of AI and LLMs. This covered copyright, intellectual property and the environmental impact – something fresh in our minds from our previous ContentWise event on planet-centred design.

In two hours, we barely scratched the surface of what AI could do and how it might be used. But our main takeaway was that AI is a tool, not a replacement, for human designers.

Despite being associated with the creation and editing of the written word, most content designers I’ve spoken to spend around 80% of their time talking, listening, asking questions and furthering their understanding of the environments they’re working in.

This involves person-to-person interactions, talking and listening, plus understanding needs, situations and problems. This is not something AI can currently replicate. So, for the foreseeable future, a healthy level of curiosity, pragmatism and open-mindedness is essential to maintain an understanding of AI and how it can partner, but not replace us.

ContentWise Leeds runs in-person every two months. Join the LinkedIn group to find out more.