Getting Started

A few years ago, I worked with a startup building a SaaS product that tried to solve a problem that many large enterprises have in putting structure on the early stages of the product development/discovery process.

It was like a Jira for Product Managers, but instead of focusing on tickets, the system took a user through a workflow of activities that helped them build out a presentation deck.

For users, it gave a guided approach for how to do product development; for managers and stakeholders, it provided a consistent "pitch deck," so that products could be compared apples to apples.

It was a nice idea, and through the customer validation work, I could see that it met a need in the market. However, we did not have the right sales approach and could not get enough traction, so eventually I had to close the company.

But now, with AI, I wonder if I could build an AI Product Bot that could do the same as the SaaS product but do 90% of the hard work for me?

Could the AI even run customer/user interviews?

It's an interesting thought and a nice little experiment to play and learn more about #GenAI and develop my #VideoCoding skills.

Experiment 1

The way I like to build software is to build out the minimum working pathway, which for me starts with some unit tests. I want to build a suite of unit tests that take some product ideas, push them through an LLM to see the results.
This approach is to do some basic experimentation with Prompt Engineering and testing with some different models.

I have an old product playbook that I will try to get the LLM to implement. One question is, how does a single LLM perform vs a group of Agents tuned for specific tasks/activities?

The outcomes of this experiment will inform the core architecture of the app.