How to discover opportunities for LLM enhancements
If you're curious about AI and think that an AI-chatbot-covered-in-purple-glitter-sparkles is perhaps a bit extensive for your product solution, I might have a tip for you. It's a quick and simple method that my team and I found useful while doing discovery for a project we're currently working on.
So, if you're anyone working in a product team and you've spoken to your customers a bit, you likely already have a good understanding of their context, their struggles, their hopes, and dreams. You're also likely working on defining some concrete problem statements to improve the experience based on your insights. This is a really good point in time where you can capture your user needs and the new technical advancements in one go and discover opportunities that may not have been obvious at first.
We know that genAI is really good at specific capabilities such as generating, analysing, summarising, suggesting, and translating text. Think of these as 'key verbs'.
What we found really useful was to use these during a small design sprint after we had spent some time working on empathy maps and understanding our user problems in depth. As we brainstormed on the 'How Might We...' statements, we focused on creating them with these key verbs in mind, so something like this:
How might we generate a draft of … ?
How might we suggest content for … ?
How might we summarise important highlights about … ?
So if your users spend a lot of time writing or processing text content in certain parts of the user journey, genAI could assist by generating a draft of a piece or suggesting potential additional content based on some relevant input data that is relevant to the context and available to you. This could be anything from text prediction or content suggestions in simple search or text fields to analysis of long text documents - if you have a product out already, you will likely already have these interactions to build upon.
There are a few things to keep in mind, such as ensuring that the user stays in full control and is aware that AI is assisting here, but a little help along the way with the mundane, repetitive, and predictable tasks will likely make a big difference for your user if you can get it right. This method is a simple addition to your usual discovery, and you can ensure that the AI enhancements you introduce are also aligned with the actual problems that your users need help with.
For more on design and genAI, make sure to check out designingwithai.ch.
