My Process: From Business Pain to Working Module with AI

As the founder of formspace.design, I've come a long way from manual accounting in the 90s to independently developing complex systems with artificial intelligence. This journey began with acute pain: a lack of effective tools for managing landscape projects and business in general. Today, I'm building formspace.design, using AI as my main partner, transforming years of experience into functional modules accessible to everyone.

How Business Pain Becomes a Module Idea

My business experience began in the 90s with NPF Alkar, where, as deputy director, I faced the need to manage accounts and projects almost manually. In those years, we created software for training, but internal processes required constant attention and huge time expenditures. Later, at Dom i Sad (Home and Garden), a company I founded in 2000 and managed for over 20 years, the problem of accounting and automation became even more acute.

This constant routine, which drained energy and prevented me from focusing on creativity, became the starting point for creating every new module in formspace.design. I aimed to reduce the time spent on estimating and client management by 70%, and to automate plant selection for a 15-are plot. If you're interested in how it looked at the beginning, you can read about it in the article "The 90s: How I Manually Managed Accounts at Alkar-Invest Without a Single Program".

AI as a Bridge to Implementation Without a Team of Programmers

For a long time, I tried to solve the automation problem using classical methods: hiring programmers, assigning tasks. But, not being a programmer myself, I often encountered difficulties explaining my ideas and requirements, which led to misunderstandings and delays. I talked about this experience in the article "How I Hired Programmers Without Understanding What Exactly to Explain to Them". At 61, when I began actively exploring the capabilities of AI, I realized it could be my chance to build the tool of my dreams independently.

For me, AI is not just an assistant, but a full-fledged partner. It allows me, a person with deep domain expertise in landscape design and enterprise economics, but without formal programming education, to translate my ideas directly into code. AI handles the routine of writing syntax, while I focus on logic, functionality, and user experience, which provides tremendous acceleration and eliminates barriers.

My Process: From Idea to Prototype with AI

When an idea for a new module or an improvement to an existing one arises, my process looks roughly like this:

1. Clear problem and goal definition: I formulate what specific pain the module should solve and what result I expect.
2. Logic and data design: I draw diagrams, describe algorithms, and determine what data will be used and how it will interact. My experience as a business consultant is invaluable here.
3. Formulating detailed prompts for AI: This is a key stage. I break down the task into small parts and write the most precise requests for the AI, specifying the programming language, desired code structure, and necessary functions. For example, instead of "write code for an estimate," I'll write "create a React component for an interactive estimate table, including fields for 'item name', 'quantity', 'unit of measure', 'price per unit', and 'total cost', with automatic calculation of the grand total, the ability to add/remove rows, and save data in JSON format."
4. Code generation and analysis: The AI generates code, which I then carefully analyze. I don't blindly copy; instead, I check every line, compare it with my logic, and look for possible inaccuracies or suboptimal solutions.

Secrets of Effective Prompts and Iterations

The effectiveness of working with AI directly depends on the quality of the prompts. I've found that the more structured and specific the request, the better the result. Using examples, specifying the desired output format, and even asking for an explanation of the generated code's logic significantly increase accuracy. For instance, when developing a module for calculating earthworks, I clearly described volume formulas, soil specific gravity, and compaction coefficients so that the AI could generate an accurate algorithm. Without these details, the result would have been too general and unusable.

An iterative approach is my main principle. I don't try to get perfect code on the first try. Instead, I create a small but working fragment, test it, identify shortcomings, and then adjust the prompts for the next iteration. This allows me to move forward quickly, correcting errors in the early stages. Sometimes this means 5-10 refinements for one small component, but ultimately, it saves weeks of work compared to traditional development.

Common Mistakes and How to Avoid Them

The most common mistake when working with AI in development is excessive trust and insufficiently detailed prompts. Many expect AI to 'guess' what they need, but AI is a tool that requires precise instructions. For example, if I asked AI to generate a design concept for a plot without specifying its size (e.g., 10 ares or 50 ares), climatic zone, or client preferences, the result would be too abstract and useless.

Another mistake is ignoring the need for manual review and debugging. AI can generate syntactically correct but logically flawed code. Once, I received code for calculating the area of an irregular shape that looked flawless, but with certain input data, it produced incorrect results due to a non-obvious mathematical error. Only thorough manual review and testing across different scenarios can uncover such 'hidden' bugs.

How I Use formspace.design to Validate Concepts

Although I am the developer of formspace.design myself, I actively use its principles and even module prototypes to validate my own concepts and ideas. This allows me to view the product through the eyes of a user and continuously improve it.

For example, when I think about a new feature, such as an advanced zoning tool, I first use the approach from our module 1.x "Plan/Zones/Objects" to mentally or on a sketch determine how users will interact with this feature. I imagine what zones they will create (e.g., a 20 m² recreation zone or a 3-are vegetable garden), what objects they will place, and how this will affect the overall logic. Then I can apply the principles of 2.1 "AI Concept" to quickly generate basic ideas for the user interface or even code snippets that implement this zoning logic. This approach significantly reduces prototyping time and allows me to be confident that each new module solves the real problems of our users.

FAQ

Is it possible to create a complex product without a team of programmers?

Yes, I am convinced it is possible, especially with modern AI tools. My own experience with formspace.design confirms this. The main thing is deep domain expertise, a clear understanding of the task, and the ability to effectively interact with AI as an intelligent assistant, not just a code generator.

What is the role of AI in your development process?

For me, AI is a tool that automates routine coding tasks, suggests solutions, and accelerates prototyping. It frees me from needing to memorize syntax and language intricacies, allowing me to focus on architecture, logic, and user experience, drawing on my many years of experience in landscape and business.

What are the most common mistakes when working with AI in development?

The most common mistakes are insufficient specificity in prompts and blind trust in the generated code. AI requires precise instructions and subsequent thorough verification of results. Without critical analysis and testing, one can end up with a functionally incorrect or suboptimal solution that will require much more time to fix.

How long does it take to develop a new module?

Development time varies greatly depending on the module's complexity. A simple component can be created and tested in a few hours, whereas a large module integrating several functions (e.g., an estimate module with material and labor calculations, including about 50-70 different items and formulas) can take several weeks of iterative work. The main thing is constant forward movement and readiness for refinements.

Do you need to be a programming expert to use AI for software development?

No, you don't need to be an expert in the traditional sense. My journey is a clear testament to this. However, it is extremely important to have deep expertise in the field for which you are creating the software. It is this domain knowledge that allows you to correctly set tasks for AI, evaluate the quality of its responses, and guide the development process in the right direction.

What is the most important quality for a solo developer using AI?

The most important quality, in my opinion, is discipline and critical thinking. Discipline is needed for consistent work and iterations, while critical thinking is for evaluating AI results and independently correcting errors. The ability to self-learn and constantly master new technologies is also crucial, as the world of AI changes very rapidly.

Can you give an example of a specific business pain that formspace.design helped solve?

One striking example is the automation of estimate calculations. At Dom i Sad, creating a detailed estimate for a landscape project, including materials, labor, and equipment, took up to 3-4 hours of manual work for a 15-20 are plot. With the formspace.design module, which I developed with the help of AI, this process now takes 15-20 minutes, including the ability to quickly adjust items and prices. This allowed us to significantly increase the speed of processing client requests and improve overall efficiency.

My journey to creating formspace.design is one of continuous learning and applying new approaches. I believe that everyone who encounters inefficiency in their work can find a solution by arming themselves with modern tools and their own expertise. I invite you to explore the capabilities of formspace.design and see how technology can simplify your life and work.