AI Research Techniques

In the past few segments, we’ve shared a variety of ways to do research the old fashioned way. But over the past year, new technology has come online that makes doing research even faster. So for the next few minutes, we’ll share a couple of ways to use AI to put your research process on steroids. We’ll explain the ideas briefly here, and then include some additional demonstrations of us using an AI for research. All of that is included here in this course.

Starting with AI Analysis [20]. To use this technique, you’ll take the data and insights you’ve accumulated in your documents and feed it to one of the AI models for analysis. ChatGPT is very good for this and if you have access to the GPT plugins, you can drop a link to your Google doc with instructions on what you’d like GPT to do, and get back all kinds of analysis. Claude 2 is also a great option for this kind of work as it allows for larger amounts of input.

You’ll want to play around with this, but for example, to run a competitor analysis, upload the document with competitor data and ask the AI to identify the most common themes from the data. You can ask it to find the most common features and benefits. And you can ask it to identify new features and benefits that a user might be interested in, but are not present in the data.

Depending on how much data you upload, you could ask the AI to summarize websites and sales pages so you can easily see the highlights. 

Depending on the AI you select, you may be able to simply ask the model to conduct a competitive analysis on the list of companies you provide. But it is almost always better to be very specific about the output you want, specifically marketing/copy strategy and ideas. Otherwise you may get output about manufacturing or warehousing and other business issues that are beyond your scope of work. Check the included demo videos for some examples of how we would do this.

Another way to use AI is sentiment analysis. Start by uploading the quotes you’ve pulled from your mining exercises and ask the AI to identify the overall sentiment of the data. Ask it to identify underlying emotions and beliefs that these users have about the product and their chances for success. You may even want to try the same exercise using different AI models and the outputs will be different and may provide additional insights and ideas to use.

One of our favorite research techniques is using an AI model to create a focus group of users [21] to learn what they would say about the product. To do this, you would create a list of potential users. Include as much information about them as possible—their pains, struggles, psychological profile, and personal information for each person in your focus group. The more you can tell the AI about each person, the better their responses will be. Then tell the AI about the product and ask it to conduct a focus group where each person talks through their struggles, desired outcomes, and anything else you want to learn.

This usually takes more than one prompt to do… so string together several prompts asking different questions, exactly as you might do in a real focus group. What the AI provides for you will often be useful in your copy. But the quality of the output depends entirely on the quality of the input. The more good information you provide, the better the output will be.

Be sure to check out the over-the-shoulder demos where we’ll show you the prompts and process for using AI to do research.