Written by Mel Tangonan

Introduction

AI has surged in popularity the past few years, with everyone eager to hop on the wave and see if it really is a big deal for the future of work. It’s become the biggest buzzword in every industry, seen as glamorous and full of promise to revolutionize the way we operate. Like everyone else, we wanted in. The thing was, we were so accustomed to our usual methodologies that we weren’t exactly sure where or how to incorporate AI. Even a with popular and easy-to-use platform like ChatGPT, it wasn’t obvious how to integrate it into our usual way of work. Many in this space are still figuring out the potential of AI and how to best leverage it. Our team was no different—until a recent project with the Chicago Park District gave us the perfect opportunity.

The Park District brought in Clarity to lead a website redesign aimed at better serving its users and the Chicago community in general. This project wasn’t just about updating the site’s look—at its core, it was about transforming it into a tool that easily connects users to programs, events, local parks, and much more.

From the beginning, there was a big emphasis on making sure that the redesign aligned with user needs. Multiple efforts were made to understand these user needs, such as analyzing survey responses, conducting stakeholder interviews, and interpreting website analytics. Part of Clarity’s task in this project was to take this diverse mix of data, both qualitative and quantitative, and turn it into meaningful insights that could drive the redesign.

The Challenge

Gathering data from these multiple sources gave us a lot of valuable information to work with. But with such diverse formats and types of data, it quickly became a heavy challenge to interpret all of it.

1. The surveys alone produced over 7,000 open-ended responses. Surveys like these always offer the richest insights, but sifting through thousands of responses line-by-line was simply too time-consuming and inefficient.
2. With the stakeholder interviews, we documented the stakeholders’ responses to our interview questions. There were a variety of answers for every question and each stakeholder had their own unique take on what mattered most. The challenge was capturing all these different perspectives and distilling them into something coherent.
3. In addition to survey responses and stakeholder feedback, we had other valuable data sources to consider, including web traffic data from Google Analytics, a Functional Requirements document, a Pain Points document, a Benchmarking Analysis, and the Park District’s Strategic Plan. We needed to connect all these sources and link them to the broader picture of what users needed from the site.

With all this data in front of us, it was overwhelming. Still, within this mess, there was something valuable and we needed to use it. So we asked ourselves: how do we process and interpret all of this and turn it into something meaningful and digestible?

That’s when we turned to AI—specifically ChatGPT—to help us tackle this question.

How We Used AI

The idea was to have ChatGPT sift through the data to identify key themes, frequently mentioned ideas, and recurring pain points and desires. With AI handling the grunt work, we could spend more of our time refining and interpreting these findings, turning them into meaningful recommendations for the website redesign.

We started by feeding data from various sources into ChatGPT in batches, then prompting it to work its magic based on the type of data we were dealing with.

Survey Responses

Let me walk you through an example of how we tackled these survey responses. One survey question asked, “What’s one thing you’d miss most on the website?” The aim of this question was to figure out what aspects of the site users valued most.

First, we uploaded an Excel file with all the responses into ChatGPT. Then, we carefully prompted it to analyze the responses in the way we needed. That’s really where the the true power of AI lies. Here’s how we did it:

1. We started by asking ChatGPT to read the whole Excel file. It responded with basic info about the file: the number of sheets, column names, and a few sample responses. This was an important first step to make sure everything was in place and that we were diving into the correct data.
2. Next, we asked ChatGPT to focus on the sheet with responses to the question, “One Thing You’d Miss Most.” It gave a quick breakdown of the data—column names, total number of responses, some sample entries. Again, this confirmation step was important in making sure that the AI was working with the correct data.
3. Then, we prompted ChatGPT to do a deeper analysis, asking it to “find common themes and make the data digestible and meaningful.”
   1. ChatGPT identified several key themes, based on the frequency of words in the responses. For example, it noted that the word “Programs” (or variations like “program”) appeared 430 times. That told us right away that users placed high value on things related to programming.
   2. It also picked up common phrases, like “Online registration,” and this suggested to us that maybe it was important for users to have a smooth registration process for events and programs.
4. After identifying key themes, we prompted ChatGPT to visualize the results by creating a few specific charts.
   1. “Create a word cloud to visually represent the frequency of words in the responses. Larger words indicate higher frequency, providing a quick visual of common themes. Remove filler words such as ‘a’, ‘the’, ‘of’, etc.”
      
      
   2. “Create a bar chart to show the distribution of the most mentioned themes or categories. Here are the categories I want you to use: Parks, Programs, Facilities, Jobs, Events, Memberships, Permits & Rentals.”
      
   3. “Create another bar chart to show the frequency of the most common words. Again, remove filler words such as ‘a’, ‘the’, ‘of’, etc.”
      
      

To do all this, ChatGPT generated Python scripts to analyze the file. This step was really just a sanity check for us—to make sure ChatGPT was handling the data correctly. Once we saw that everything was in order, we could confidently move forward, trusting the insights that followed. Here’s a snippet of the code it used:

In no time, ChatGPT was done. We now had a clear understanding of what users valued most—and the data to back it up. What could’ve taken days of manual analysis was done in a fraction of the time. But, of course, the job didn’t end there. We didn’t just let ChatGPT run the analysis and call it a day. We still had to interpret these results and turn them into actionable insights. For instance, from the results above, it became obvious that Programs and Parks were top priorities for users. This data led us to focusing heavily on these areas in the redesign. This where the human intervention comes into play—making sense of the AI’s findings and translating them into data-driven decisions.

And that was just for one question. Now imagine how powerful this is when applied across multiple questions and responses. The efficiency at scale here is something we couldn’t achieve on our own.

Stakeholder Interviews

Now done with the survey responses, we moved on to the next source of data, the stakeholder interviews. These interviews brought a new level of complexity. Each stakeholder had their own priorities and perspectives, and it was our job to take these diverse viewpoints and piece them together into something cohesive.

The approach we used for analyzing all this feedback wasn’t all that different from how we handled the survey data. We really just focused on identifying where feedback most overlapped and where it diverged. Here’s how we tackled it:

1. We fed ChatGPT the raw and messy notes we took during the stakeholder interviews.
2. Then, we prompted it to sift through the data, looking for common words, recurring categories, and key themes—much like it had done with the survey data. The groundwork we’d already laid with the survey data made this process a lot smoother.

Although the process was straightforward, the results were powerful. ChatGPT synthesized all the stakeholders’ feedback into something digestible—something we could actually make sense of. It revealed common themes and showed us where there was the most alignment. This data gave us what we needed to make informed recommendations. As a simple example, one of the most frequently mentioned ideas was Improved Navigation and Search Functionality. Across all the interviews, there was a significant mention of the need to enhance the site’s navigation and search capabilities. Based on this, we knew the redesign needed to focus on making the website more intuitive and more user-friendly.

Other Data Sources

We also had valuable information from other sources, such as website traffic data from Google Analytics, our Functional Requirements document, a Pain Points document, a Benchmarking Analysis, and the Park District’s Strategic Plan. These sources held just as much weight in steering the direction of the redesign.

The process for analyzing these didn’t demand as much AI intervention. Unlike the survey responses and stakeholder interviews, which were more messy and qualitative, these sources were pretty straightforward. The real challenge here wasn’t breaking them down but tying them all together. It was another, deeper layer of synthesis. Now we had an abundance of insights, not just data, and our job was to turn them into something concise and actionable.

In this final step, we used AI really to cross-reference everything. For instance, we had ChatGPT match up user behavior patterns from Google Analytics with the themes we identified from the surveys. AI really helped us pinpoint where the insights aligned, where the recommendations overlapped, and what truly needed our attention. It helped us figure out the big question: what should we really focus on in this redesign? It helped us prioritize the changes that mattered most and ensure that this new website delivered exactly what the Park District wants and what Chicagoans need.

AI’s Role

AI helped us turn a mountain of information into something digestible, something actionable. ChatGPT took care of the grunt work; synthesizing and analyzing the raw data, and giving us the high-level insights we needed to make informed decisions that would shape the redesign.

Our experience using AI in this project showed just how much value it can bring to web development projects, especially in the early phases of research and discovery. By automating the tedious parts, AI saved us a lot of time. Not only can this speed up the project timeline, but it can allow us to focus more on the strategic side—interpreting the insights and aligning them with the project’s goals.

This effort also made one thing clear: AI is already a powerful disruptor in how we work, but it’s not replacing us. It’s more like robot arms—they let you do a lot more than you could on your own, but you still need to be in control. While ChatGPT processed and analyzed the data with amazing efficiency, human judgment is still crucial. After all, we were the ones telling it what to do and how to do things. We were still the ones refining the insights, making the strategic decisions, and ensuring the recommendations fit the bigger picture. AI is an incredibly powerful tool, but its true potential comes when it’s paired with sound judgment, strategic thinking, and expertise.

The question is no longer about whether AI is going to reshape work or not—it’s about how it will. We get to decide how to leverage AI to make work more productive, more interesting, and ultimately more meaningful. Looking ahead, AI’s role in our future projects will only continue to grow. For us, this project was just the beginning of discovering how AI can enhance our approach to web development and beyond.

Mel Tangonan is a versatile Consultant at Clarity. He has a wide range of project experience, spanning management consulting, project management, web development, and Anaplan software implementation.