Peter Clark

5 June 2023

Generative AI Journeys


We wanted to hone our efforts with generative AI on helping with the "cold start" problem — it's daunting having to create anything from blank, and if generative AI can assist with producing something that is 60-80% of the way there, it's much much easier to finish it off.

We've all been in the situation of opening Google Slides, writing a heading slide, and then thinking "what the hell do I write now"

Topics + Prompts

You can now ask Journey to create a Journey about any topic, via a prompt.

We generate 5-12 sections of content, structure it (bullets, paragraphs, headings) and also generate relevant consistent imagery.

Underlying Prompts

When you're then editing a Journey, you can obviously edit the text like any other Journey, but you can also view the actual underlying prompt that the AI used to generate that section. This not only helps you understand what the AI was trying to convey, but also allows you to quickly edit the prompt to enhance or adjust the content.

As an example, this prompt was:

Generate a list of 3-4 sentences explaining what polar bears are and why they are important.


We also offer the same capability for imagery — we send these prompts to Stable Diffusion:

As an example, this prompt was:

Generate an image of a polar bear using its adaptation. The image should show the polar bear swimming in the ocean, using its webbed paws to paddle efficiently.

Maybe you're dissatisfied with the image for some reason, you can easily tweak the prompt until you're satisfied.

Why we're doing this (because it's automagic ... and cool)

We've found this generation works shockingly well. We wanted to experiment with AI in a thoughtful way, and this seemed like a reasonable endeavour, but as we worked and tweaked this concept, we have been really impressed by the quality and depth that the AI is capable of.

We feel that this is a starting foundation of a really exciting foray into generative AI, allowing users to leverage the best of AI (the amazing output) whilst minimising the downsides (of it behaving like a blackbox)

We'd love to hear your thoughts on this experiment, and of course see what you come up with!

View this blog post as a Journey →