Product Planning in the Age of AI: Finding the Right Balance
When to think deeply and when to prototype quickly
As I watched AI transform my product idea into working code in under five minutes, I couldn’t help but wonder: are we entering a new era of product development, or are we just finding new ways to validate timeless principles?
This question hit me recently while reading a blog post that made me pause and reflect. Chris Kiehl, observing software development over the last ten years, made a statement: ‘Most programming should be done long before a single line of code is written.’
Most programming should be done long before a single line of code is written - Chris Kiehl
That is a profound statement for many reasons.
When you build a house, do you start without a plan or idea?
Even if you build a simple shed, you start with a plan.
If you cook a recipe, you have some idea of what you want to make. Sure, you can throw ingredients together, but you have a basic recipe in mind, even if you are making it from scratch.
When you build a software product, you have a basic result in mind. Yes, even for a component.
This applies to engineering, design, marketing, product management and more.
You always start with an idea, even if it is a simple concept.
With hand-written code, you start with an idea.
With AI-generated code, you start with an idea, and a prompt.
As humans, we are great at generating ideas.
AI is getting great at generating ideas, but there was still a spark of an idea from a human, at least currently, that prompted the seed of the idea.
Back to Chris’ statement around how long before a single line of code is typed/generated, the programming idea should have started well before.
I couldn’t agree more!
The question really is, what is the pre-work or idea generation that is adequate or needs to happen to drive this from A to B, and then C to D, and so on.
Where do product ideas originate?
Some ideas for products are intuitive and some don’t really present themselves until later as an iteration.
Some are visionary and might take years to come to fruition. Take Steve Jobs vision for the App Store in 1983.
Others can be realized in a day.
AI prototype example: From idea, to prompt, to prototype
Some ideas can go from thought to prototype in minutes. Companies and platforms such Replit are doing this. They just launched a service where you can suggest through a prompt to the Replit Agent what to build and the Agent builds your initial idea into a workable initial product prototype.
Let’s try this out… I will give Replit Agent a prompt with a spur of the moment idea.
Initial idea
It would be cool to have a product idea generator. Usually, product ideas come from existing customers/market and discovery about what opportunities exist, however, what if you have an initial idea and want to brainstorm a little more with AI assistance?
Initial prompt
Build a product idea generator. I can give you an OpenAI API key to add AI generation capabilities.Initial prototype
Let’s head over to Replit and try out this initial prompt.
1. Sharing the initial prompt idea with Replit Agent
2. Optimizing the prompt idea with AI assistance
3. Replit Agent suggests the initial prototype to be built, along with other initial feature ideas to consider
4. The initial prototype is ready for testing
Quick demo video showing the finished prototype in action
Now that we have a working prototype that was built with AI agent assistance, we can move on to testing, and then further iteration if desired.
Here is a summary of what takes place in the following video demo of the ‘AI Product Idea Generator’.
First the product idea is typed into the form field and then the ‘Generate Idea’ button is clicked and with AI assistance a product idea is generated.
The output includes:
Proposed product name
Initial proposed product description and positioning
Market Potential
Target Audience
Key Features
Debrief on this AI prototype example
How long did this AI prototyping process take overall? Less than 5 minutes to go from idea, to prompt, to prototype.
Let’s pretend this AI prototype was tested out in the market, got some positive feedback, and we wanted to take it on its way to a production build?
The great part of generating this initial prototype is now the initial code can be shared with those with engineering and design skillsets, and it can be taken to the next level.
Product ideas can originate from multiple sources
Typically, product ideas can come from a variety of sources.
Working within a product team those in the ‘product trio’1 are often those that are pulling product ideas from various sources. These could be from internal ideas (founders, executives, and teams), customer feedback, market demand/shifts, discovered opportunities from customer discovery, and more.
With a vision painted, the why considered, and the layers of the what and how becoming clearer, what do we do with these ideas?
Considering the agentic AI example above with Replit Agent driving much of the underlying idea to prototype work, how can we ensure this idea we originally had is something worth exploring more?
Prototype experiments
With the ability to spin up a prototype within a matter of minutes, as briefly touched on in the debrief above, we can go from prototype to experiment to production build - if an idea shows promise.
I’ve thought about the signal vs. noise of how easy it is to spin up a prototype and how the quality of the output compared to a well thought out, researched, discovered, and planned prototype could be.
Will the market be flooded with AI built prototypes and market demand show simple and quick prototypes aren’t as effective?
Or will these AI prototypes be a gateway into faster discovery and feedback loop cycles with product development teams and customers?
I think time will tell, and the end results will be shown by what goes from a simple product idea to a product with product market fit.
Wrapping it all up
Drawing inspiration from Chris’ original thought of thinking through an idea before generating code, whether traditionally or AI-driven, one can conclude that taking some time still to ponder, research, and validate an idea first, before jumping right into building, is still very valid.
At the same time, it behooves one also to experiment with fast prototypes and ideas to ensure speed to market, taking an intuitively innovative idea and building it faster than available resource streams, could benefit the end customer in the end while driving business outcomes.
As with anything in product and tech, experimenting with old and new ways of doing things and testing the ideas is always the best path forward.
Traditionally a product trio is a cross-functional group typically consisting of a product manager, a designer, and a software engineer who collaborate closely throughout the product development process. (More in-depth definition here)