Let’s Get to the Point:
Playing With AI Part 4

How to transform data into personalized, actionable insights with an AI-driven prototype.


There’s simply too much information out there. Whether you call it a sea, an ocean, or a supermassive black hole, we’re all trying to sift our way through it. So, we built something to help. This is the fourth of a series of prototypes we created — with an article that breaks down how we got here…

…but if you’d like to cut to the chase, go right ahead and check out our prototype.

The Challenge of Information Overload

Across every aspect of our lives, we're constantly exposed to a massive amount of data and information. Whether shopping online, managing our finances, streaming content, or simply browsing the web, the volume of information can be staggering. Even more so, it can make it hard to focus on what matters and decide where and when to take action.

Many systems today lack personalization and context, presenting a one-size-fits-all approach. And why does this matter? Because it makes it difficult to surface the most relevant and valuable information tailored to each person's needs and circumstances.

A Hypothesis for AI-Driven Personalization

So, we set out to find a solution. By leveraging generative AI to derive data-driven insights and customizing the experience to the specific context of each user, we hypothesized that we could make information far more understandable, actionable, and valuable.

First, we needed to center on generating rich insights from data using LLMs (large language models). By prioritizing and categorizing insights based on a user's profile, we could quickly home in on what's most important to them and their situation.

But to get there, we needed to look at what we already knew: UX Design. We applied this lens and came to the conclusion that offering multiple modes of interaction — like visual dashboards, audio summaries, content cards, and conversational chat interfaces — could help us create the engaging, personalized experience we desired. In other words, a chat interface alone wouldn’t cut it.

Bringing the Concept to Life

To validate our hypothesis, we developed a prototype that generates insights from data sources based on user personas we predefined and the tasks they aimed to accomplish. Our prototype then intelligently organized these insights into clear categories, presenting them in a clean, easy-to-scan list.

Within the prototype, users can click on any insight to dive deeper into the details, viewing related content and data visualizations. An integrated chat interface suggests contextually relevant follow-up questions, allowing users to explore insights through a natural conversation. For situations where users are on the go or need a screen-free experience, the prototype can generate audio summaries that share key insights in a friendly and easy-to-understand format.

We used the latest and most powerful LLMs from Anthropic, the Claude 3 family of language models, to process structured data inputs and generate relevant insights. We experimented with different data ingestion and generation techniques, including synchronous and streaming outputs. For audio narration, we leveraged OpenAI's text-to-speech capabilities which yielded natural-sounding audio experiences.

Key Learnings and Discoveries

Throughout our process, we found several valuable learnings:

  1. AI Excels at Data Simplification: The language models demonstrated a remarkable ability to distill complex data into clear, human-friendly insights and summaries that captured the most essential points.
  2. Context Drives Better Interactions: Providing a dedicated space to explore each insight made it easier to surface relevant content. This framing also made conversations feel more focused and natural.
  3. Configurability Through Prompts: While initially scoping for a single industry use case, we realized these AI-driven experiences could be rapidly adapted simply by adjusting prompts and data inputs. This is why our prototype presents two scenarios: health & fitness and hospitality.
  4. Need for New UX Patterns: As we continue experimenting with generative AI, we're clearly entering uncharted territory where established user experience patterns may no longer apply. This means an iterative, prototype-driven approach is crucial.

Exploring a New Frontier for User Experiences

The potential of generative AI to transform user experiences across industries is immense. We can help users make better decisions and take more meaningful actions by empowering them with personalized, context-driven insights through intuitive multi-modal interfaces.

As we continue exploring this new frontier, we have to embrace experimentation and rapidly prototype ideas to understand what truly resonates with users. Ultimately, there's a vast opportunity to innovate new AI-driven experiences that enlighten, empower, and bring value to digital experiences.

Related Reading