In 20min, I Conducted 12 Imaginary Customers Interviews For A New Venture Using AI For Free - MindfulCircles.Org

Disclaimer: If you are a venture architect, founder or a product manager, do not, and I mean DO NOT use such a method to validate an idea :) AI is generating content based on what was fed to it. The ability to innovate for humans by humans will always require humans participation, unless AI becomes mortal, has consciousness, can reproduce to survive, cares about its materialistic and ethical legacy AND is the decision maker in buying your service, only then, you may use such an approach to validate a new idea. But until then, validate your business idea with real customers. You may use such a method for inspiration and discovering blind spots in your thinking but never depends on AI to analyse what humans of tomorrow might like or dislike.

Ok that being said, let's come back to the venture in hand. It is called: MindfulCircles. The idea came to me while on a Meditation retreat last year. I always admired the people around me when on a retreat and the level of authenticity and calmness in every conversation we have. My only issue, being the science and tech nerd I am, is that the topics of the conversations were not really in my interest. At the same time I was always allergic to networking events in the startup scene. Don't get me wrong, I have met great people and even business partners in such events, but the type of conversations and the way people present themselves in these events is so not authentic that makes the value of going to such events for me close to zero. So back in October I started experimenting with a scalable mindful group format to bring professionals from the same industry together in a unique authentic way. People come as they are, meditate together to calm down and be present and mindfully share about their challenges, aspirations and dreams. Judgment free in a room full of compassion and loving kindness similar to the atmosphere you find in mindfulness retreat. 

To test this, startup professionals were my ideal target to start the first circle. Hence the first Berlin Mindful Startup Circle was born. I have been running it with a few friends since December and have been exchanging feedback on the format of the circle and how to improve it. So, no, I didn't depend on AI to create this.

Nevertheless as an entrepreneur and venture architect I love playing with tech gadgets, physical or digital :) So I was like why not give chat GPT a set of personas and ask it to conduct problem and solution discovery interviews with them and see what will happen. And I have to admit, I am positively surprised, yet definitely would take all of the wordings ChatGPT spits out with a grain of salt, almost always. But here how it went:

Step 1- Create Personas Using Syntheticusers

With the free version of SyntheticUsers you could simply describe your target customer, what their pains and needs are and how would your solution solves it for them.

Once added, click on Generate SUs to get personas together with a few characteristics that Syntheticusers gives. What I didn't like though is that the results are not always consistent even if prompting the same input and frankly speaking sometimes some of the characteristics don't even make sense. But my goal here is to unleash blind spots not really validate an idea, so I went on and created 6 personas. It took me less than 5 minutes.

Step 2 - Generate Problem Discovery Interview Questions

I then went to ChatGPT and I asked it to list for me problem discovery questions for the solution I have in mind. I used a long Promot where I gave it the target customers, their pains, needs and the proposed solution. I got quite interesting results. Those questions can be used for interviews with real humans. Good job ChatGPT. Here the results:

Step 3 - Ask Your Personas The Problem Questions

A simple prompt where I copied the questions and the 6 personas. The results were ok but nothing new that I didn't expect or learn before. I couldn't get the model to tell me something that is new or not thought of before. But for me as a non-native English speaker I am always in appreciation for LLMs better way of phrasing what I have in mind. It is always good to know what keywords or phrases to use when blogging about the venture for example, but definitely the answers are non-inspiring and should never be used in professional context as a replacement for what humans would potentially answer. Once again, never use this method to validate your idea ;)

However where I might actually use it in a professional context is to upload the real results from real customer interviews as input and continue with Step4.

Step 4 - Refining The Pains and Needs Based On Problem Interviews

Now that we have our problem interviews results :) We should refine what we originally thought the pains and needs are. So in the same manner using a simple prompt I asked GPT to summarise the Pains and Needs based on the interviews. And I really like the results. This could definitely go in the Value Proposition Canvas, though I am lacking quantifiable metrics. In real world I would add some indicators about how big or small each pain or need is in my interviews to help in prioritisation. Once again, this is to unleash blind spot, never use this method alone to validate your idea.

Step 5 - Generate Solution Interview Questions

Using the updated list of pains and needs, in the same manner I asked ChatGPT to generate Solution Interview Questions.

Step 6 - Conduct Solution Interview Questions With Your Personas

Also using a simple prompt where I copy-pasted the 6 personas with their characteristics and asked ChatGPT to interview them using the list of the Solution Discovery questions it suggested.

And results were not bad. And can be quoted actually on our landing pages 😉 (don't!).

Step 7 - Generate The Full Value Proposition (Customer, Pains, Needs, Gains)

This was my favourite part. Because with a simple prompt I actually got quite interesting results that can be used in professional context. I would imagine if I fed GPT the actual results of real interviews this part would be quite inspiring. Though what are the chances that the results of an interview with a human would be that structured. The aim in discovery questions is to ask open ended questions and get inspired in which direction you should take the idea or how you are describing it. The aim is not to get structured answers that can be fed to AI. So on a final note, once again, don't use this in your professional work as a method to validate a new idea :)