Making Your Interests More Interesting: Playing With AI Part 1

Explaining the method behind our latest prototype for personalized experiences.


This article outlines the hypothesis, methodology, and development process behind our latest AI experiment. Which is just our way of saying we did a lot of work on something we’d love to share.

This is the first part of a series of prototypes we created. We recommend reading the article below to help you understand why we built it…

…but if you’d like to skip the formalities and dive right in, be our guest.

Demystifying AI: A Personalized Path to Understanding

In the ever-evolving world of technology, generative AI has captured the public's imagination with its ability to create highly engaging and seemingly human-like content. However, amidst all of the hype, many of us struggle to grasp what this technology can actually do for our daily lives. At Instrument, we recognized this disconnect and set out to bridge the gap by developing a prototype that makes AI education something it should be — personal.

Tailored Solutions, Real-World Impact

Our hypothesis was simple: when users share their occupation and hobbies, an AI system could take that info and generate specific examples of how AI could help them. This personalized approach would create a customized experience — akin to a website page tailored specifically to the user — which would help them better understand AI and how it could impact their passions, their work, and their lives.

A Streamlined, User-Friendly Experience

To test our hypothesis, we developed a prototype that utilizes a large language model (LLM) as a helpful assistant. Departing from the conventional chat interface, our experience offered something more straightforward (and in our opinion, easier). After users provided simple details like their occupation and hobbies, the AI generated practical and related use cases for leveraging generative AI.

Though we try not to pick favorites, if we had one for this prototype it would be the "Show Me How" button that accompanied each use case. When clicked, this button opened a new window with step-by-step instructions on how to get started using today's AI-powered chatbots. This functionality went beyond simply providing ideas; it guided users through the process and offered tips to help them effectively use it. What’s even cooler is that all of this was generated on the fly by the very technology the user was being taught to use.

Lessons Learned and Opportunities Ahead

Throughout developing this tool, we learned that prompt engineering — aka what instructions you give the AI — plays a crucial role in generating accurate and helpful use cases. This process required iteration and comparison to refine the tool’s intangible and intuition-derived results, or what we like to call vibe-driven development.

We also discovered that LLMs excel at breaking down tasks into manageable pieces, presenting information in a friendly and approachable way that helped guide users. One of the most powerful features of LLMs is their ability to generate bespoke, actionable, and hyper-specific content on the fly, setting it apart from the one-size-fits-all approach of traditional content creation.

At its core, this prototype illustrates an effective way to close the gap between generative AI as an abstract concept and generative AI as a valuable tool. We believe that practical and relevant education is much-needed in today's landscape so that everyone can discover AI’s true value for their lives.

I fully believe the potential for generative AI to positively impact our lives cannot be limited to opportunities that clients bring to us. That it cannot be bound to bottom line outcomes. That there is bigger picture work to be done.

Mike CreightonDirector of AI Research & Development

Integrating AI into the User Experience

In building this prototype, we recognized a larger opportunity — creating a seamless journey from learning about AI to actually using it. It might be an obvious conclusion, but we can leverage the AI itself to help people learn to use it. Yet we continue to find that it’s missing from so many of today’s products. More importantly, it keeps people from using those AI tools and services to their fullest potential.

This prototype also proved something else we've been thinking about: personalization is possible with minimal input. Personalized experiences are a promise that so many brands and services are looking to fulfill, and we feel that even today's LLMs are capable of making that a reality. With just a little bit of data about the user, we can surface and generate content that’s personally meaningful. This could be made even more effective when hooked up to a knowledge repository, further grounding the LLM in brand- or product-specific data.

So what does this all mean? Our clients should be rethinking what “personalization” means to them and their customers. Over the past decade, we've seen that consumers are willing to offer up all kinds of data if the value on the other side is sufficient — and this prototype provides proof of what that could look like. Because today’s data-to-value ratio is ready to be redefined, and AI is the tool to make it happen.

Feel like you’ve got a good understanding of generative AI?

Related Reading