How we linked GPT3 with Google Sheets – and you can too

We’ve been experimenting with OpenAI for months, scaling up our efforts with the launch of ChatGPT. (More about about working with ChatGPT in our CRO agency, here.)

Open text analysis on data from user surveys, customer emails and interviews is a core part of our CRO methodology. While there’s enormous value in reading through thousand of customer comments, it’s time intensive. We use statistical programming languages like R and Python to analyse huge volumes of text and identify patterns in the feedback, top likes and dislikes, sentiment etc.

If your process is still mostly manual, GPT-3 can help you save time by automating parts of the analysis process. In this video, CRO specialist Dave Mullen explains how it’s done.

How Dave thinks about GPT-3:

Imagine you have a new assistant. They are unbelievably well-read. They also happen to have infinite time. They come back to you in an impossibly short amount of time.

On top of this, they’re so cheap that you’d be a fool not to try and put them to use.

But there is one big caveat: they are unquestionably overconfident. They will go away and have a stab at absolutely anything.

The way I’m thinking about GPT-3 is if I had this assistant with these qualities, what tasks might I give to them?

Chapters in the video:

• 01:42   Background
• 03:14   Overview of talk
• 05:01   A useful way to think about GPT-3
• 07:17   Basic example, overview of Google Sheets template
• 10:72   Why use Google Sheets for GPT-3?
• 11:59   Challenges with GPT-3 and how to deal with those
• 18:38   Clarity Score
• 21:06   Demo of customer review analysis with GPT-3
• 24:32   GPT-3 writes a tagline based on sentiment
• 26:42   Instruct AI to flag issues with a human
• 28:36   How to use the Google Sheets template

Interested to know about upcoming events? Click here to join attendees from companies like Meta, Spotify and Canon.