Create your AI-driven Value Proposition Canvas

Fri, 08 Aug 2025 00:00:00 +0000

I forked a public GitHub Repo to generate interview data for sensitive topics. These data could not be gathered from humans, because the questions of interest were too emotionally loaded for the target segment. After generating the data, I coded the insights and illustrated them on a Value Proposition Canvas

In this article, I will share my playbook on how to use LLMs for interview data generation and illustrate insights on a VP Canvas.

Step 1: Generate Interview Data with LLMs

I forked a public GitHub Repo to generate interview data for sensitive topics. The repo is built to run on local LLMs. I used [Ollama] (https://ollama.com/) to run the LLMs locally. The repo contains a script that generates interview data based on a set of questions. You can customize the questions to fit your research needs.

After setting up the repo, I ran the application via Docker in my localhost to generate interview data. To vary the responses, I used different prompts. Here is an example of a prompt I used:

{
 "prompt": "Welcome to our discussion on how we envision a service that supports those who have just lost their partner — by helping them manage the many administrative tasks that follow, from paperwork and account closures to official notifications."
}

Demo Video:

Here’s a demo video of the application in action: Download and watch the video

I exported the interview transcripts as a .docx file, which I then used for further analysis.

Step 2: Analyze Interview Data

I used ATLAS.ti to analyze the interview data. I AI coded the transcripts to identify key insights and themes. The coding process involved reading through the transcripts, highlighting important quotes, and assigning codes to them. This helped me to organize the data and identify patterns.

Once done, I exported the coded data as a .xls file, which I then used to create the Value Proposition Canvas.

Step 3: Classify the Codes into VP Canvas Categories

I used the .xls file to classify the codes into the categories of the Value Proposition Canvas. The categories include:

Customer Jobs: What the customer is trying to achieve.
Pains: The challenges or obstacles the customer faces.
Gains: The benefits or positive outcomes the customer seeks.
Products & Services: The offerings that can help the customer achieve their jobs.
Pain Relievers: How the products and services can alleviate the pains.
Gain Creators: How the products and services can create gains for the customer.

Step 4: Create Wordclouds for the VP Canvas

I used this open-source Wordcloud Generator to copy paste the codes from the .xls file into the wordcloud generator. This helped me visualize the most frequently mentioned words in each category, making it easier to identify key insights.

Et voilà! I had a visual representation of the insights from the interview data on a Value Proposition Canvas.

Innovation | Dominic Santschi | Website

Create your AI-driven Value Proposition Canvas

Step 1: Generate Interview Data with LLMs

Step 2: Analyze Interview Data

Step 3: Classify the Codes into VP Canvas Categories

Step 4: Create Wordclouds for the VP Canvas

Did you find this page helpful? Consider sharing it 🙌