Back to Projects

Back to Projects

AI-driven Concept

Available for Work

AI-driven Concept

Available for Work

Category:

AI Web App

Role:

UX Designer, Product Designer, AI Concept Development

Duration:

3 weeks

Client: Personal project

Summary

Min Grønne Oase is an AI-powered concept that helps users transform existing outdoor spaces into greener and more atmospheric environments. By combining image analysis, user preferences, and generative AI, the solution provides realistic visual recommendations that can be implemented in real life.

Summary

Min Grønne Oase is an AI-powered concept that helps users transform existing outdoor spaces into greener and more atmospheric environments. By combining image analysis, user preferences, and generative AI, the solution provides realistic visual recommendations that can be implemented in real life.

The Challenge

Many people want greener and more atmospheric outdoor spaces, but lack the knowledge, overview, and visual imagination needed to get started. It can be difficult to assess what fits the space’s size, lighting conditions, style, and personal level of plant experience.


The Challenge

Many people want greener and more atmospheric outdoor spaces, but lack the knowledge, overview, and visual imagination needed to get started. It can be difficult to assess what fits the space’s size, lighting conditions, style, and personal level of plant experience.


The Challenge

Many people want greener and more atmospheric outdoor spaces, but lack the knowledge, overview, and visual imagination needed to get started. It can be difficult to assess what fits the space’s size, lighting conditions, style, and personal level of plant experience.



Process

AI played a central role throughout the entire process. Used as an active collaborator, AI supported ideation, exploration, validation, and decision-making across all phases of the project, enabling rapid iteration while creating a documented trail of decisions, insights, and design rationale.



Brainstorm & Concept Development

ChatGPT was used as a structured sparring partner during the early concept development phase. Through an iterative dialogue, ideas were refined, challenged, and expanded step by step. This process helped identify opportunities to differentiate the solution from existing market offerings while creating a documented trail of decisions, assumptions, and design rationale that informed the subsequent development work.


Feedback & Validation

Claude as a critical reviewer. Claude reviewed the concept brief with a sharp focus on strengths, weaknesses, and pitfalls. Claude challenged assumptions, helping identify six critical improvements.


Testing the Image Transformation

I tested the image transformation in ChatGPT, and the outcome of the transformed images wasn't as I wished. Therefore it was necessary to establish system prompts for the transformation. Instructions:

  • AI must not change the room's basic structure, perspective, lighting conditions, etc. (extra strict instruction added)

  • Usability over wow factor: the success criterion is the most usable transformation — not the largest possible transformation

  • Experience level and style direction as concrete parameters to differentiate the outputs from one another


I tested the image transformation in ChatGPT, and the outcome of the transformed images wasn't as I wished. Therefore it was necessary to establish system prompts for the transformation. Instructions:

  • AI must not change the room's basic structure, perspective, lighting conditions, etc. (extra strict instruction added)

  • Usability over wow factor: the success criterion is the most usable transformation — not the largest possible transformation

  • Experience level and style direction as concrete parameters to differentiate the outputs from one another



Building the Web Application

After refining the image transformation process, the concept was translated into a functional web application. ChatGPT helped create a master prompt for Codex, while an early prototype was built with mock data to validate the experience. As the project evolved, I explored API integrations, tested multiple LLMs and image models, and ultimately selected a multimodal solution that delivered more consistent and cost-effective results.


Key learnings

Throughout the project, AI was not only the subject of the solution but also an active part of the design and development process. Working across multiple AI tools, models, and workflows provided valuable insights into both the strengths and limitations of AI-assisted product development, from prompt engineering and model selection to reliability, consistency, and decision-making.


Reflection


This project demonstrated how AI can support the entire product development lifecycle, from ideation and research to visual generation and implementation. It also highlighted the importance of validation, prompt engineering, and careful model selection when designing AI-powered experiences.



MVP Version

The current solution represents a first MVP (Minimum Viable Product) designed to validate the core concept and user experience. While the initial feedback has been positive, user testing has already revealed opportunities for refinement and several ideas for future development. These insights will help shape the next iterations of the product and further improve both functionality and user value.

The images below showcase the current MVP version of the application.