Like some of my colleagues in Content Strategy and Development, I’ve been experimenting with generative AI apps (”Gen Apps”) like ChatGPT and MidJourney to figure out how they can be helpful at work.

Through the prompt library in the OpenAI Discord channels, I’ve picked up tips on crafting and layering effective prompts to get more specific outputs from Midjourney. A great tip has been to use ChatGPT to generate detailed prompts for Midjourney. (It’s AI prompting AI!)

These learnings came in handy recently when I had to prototype user flows for a client—an energy company in Asia. 

I quickly discovered that some of the flows didn’t have existing content. As our focus was to nail down the customer journey and experience, I needed to generate some content for the user flows quickly. 

Instead of tapping my Content Strategy coworker Geraldine to write on short notice, I wondered if I could turn to Gen AI by getting it to “clone” her. 

Cloning my coworker with AI

Here’s what I did to train ChatGPT before I gave it some tasks:

• I uploaded the voice and tone guidelines that Geraldine had created for the client a couple of years ago.
• I gave ChatGPT context on our client by uploading publicly available information from their website.
• I created templates based on the patterns I was designing for (see example below).

Now that “GeraldineGPT” knew what the template for a Service Landing Page looked like, I could prompt it to create mock content: 

What I got:

The results

I found that GeraldineGPT was great for:

• Creating structured content like “Help” articles and step-by-step how-to's
• Ensuring that the titles, tone and voice, and micro copy adhered to the brand guide and content strategy rules (e.g., sentence cases for titles, and verbs for “Call to action”)

Overall, it managed to get 90% of the copy correct but it would “guess” in areas where it did not have sufficient context. Hence, someone who has context of the content would still need to vet it carefully. 

I noticed too that it struggled with the understanding and therefore the use of phrases and terms specific to the client. For example, the phrase “move out” which the client prefers to use instead of “cancelling electricity service”. 

Exciting times ahead

There’s no doubt that Gen AI tools will improve with new plug-ins added almost daily. 

Already, there are tools and programmes that help to generate screens based on prompts. It will be a matter of time before AI produces full flows built on: 

• Components from a design system
• Content and design from design and content guidelines 

This is an exciting time when the barrier to design has never been lower. This has already gotten me to think about how our role, as designers, is going to shift from designing interfaces and interactions to a much deeper understanding of how and what our designs should be achieving for our clients.

Note: We'll be running more #PRexperiments on AI and other emerging tech to see how they can help us deliver our best work for clients. If you would like to explore using AI in your organisation, do connect with us by emailing info@pebbleroad.com.