You came here to look at AI.
Check this out.
Exhibit A: How to use AI in your digital strategy, in seven steps
This is anonymized response data from a survey on economic well-being. I analyzed 321 responses using IBM Watson NLU. From there I uploaded the analyzed data into a custom GPT from ChatGPT. I used that LLM for data processing, topic modeling and content creation. I also used the data to create visualizations in MongoDb. Click here for the full dataset.
IBM Watson turns survey responses into data on five emotions, entities, relations, and more. Click here to see the full dataset.
ChatGPT is by far the most powerful tools out there. I took all of the NLU datasets and used them to create two reports, one for semantic analysis and the other topic modeling. The semantic analysis report finds the top five themes found in the words and concepts in the data. Topic modeling reveals the top unique topics found in the words of people's responses. Together, they help us see the feelings underlying the opinion in their quantitative responses.
Above are the first few lines of an Emotions Report, created by our custom LLM. This is one of six reports I generate as part of a regular analysis. These in-depth reports provide usable insights for understanding the impact and influence of emotion in the responses. We can use the data to create visualizations, data picture stories that provide insights visually. Click here for the full report.
These are the most popular and relevant terms found in the sentence subjects of people's responses. In this case, the responses are filtered by the respondents' income level. Similarities across income levels are as fascinating as the range and intensity of subjects. For example, "a living wage" is least relevant to "not reported", then poor, then the working poor, middle class, upper class and the working class (not surprisingly). It's also the only subject shared by almost all income levels. Another thing easy to see that topics important to the working class are really, really important to them. Click here for the full portfolio.
I used the semantic analysis to create personas and topic analysis for persona variants. By personalizing the top five themes and various subtopics, it becomes easier to contextualize them. For example, you and I can frame the findings as personal to Rachel or people like Rachel. Or we can think about a punch list of demographic and psychographic values. It's easier to write and advocate for a person.
First, an observation. There is no way a human is going to glean this by reading 321 responses. The machine analyzes and retains at scale. Second, is knowing that the primary creative document is supported by a firm foundation of data; presented in a practical way. Click here for the full creative brief.
Exhibit B: A custom GPT is the new user support manual
Use the bot to explore Project 2025. Start with the display questions; or ask your own. Single-subject or task tools like these can replace the current ecosystem of app and product support resources. Imagine a link to one of these instead of a printed owners manual or a pdf. In any language.
Using AI for comms is just one way I can help
If you answer the same set of questions a lot. I can show you an automated, new and highly accurate way to answer their questions
The best reasons to use new technology lie in solving the daily problems facing your team. I can help you find organic uses for ChatGPT and AI.
I can help you build a significantly better owners manual, support page or how-to guide. No pdfs and is built for digital.
AN AFFORDABLE ALL-IN-ONE SOLUTION
Recent Work
A geo-powered site for union members looking for information on endorsed candidates. The site is geo-powered, meaning a user will only see those candidates that represent their area.
A voter guide to showcase the candidates that had taken a pledge to support a bill tightening the rules on government lobbying.