Rethinking GenAI Development: From Products to Valuable Experiences Mindset

By
Arun Giri
June 11, 2024 3 minute read

Why Developing GenAI Experiences Feels Different: Unlearning the MVP Approach

Remember the excitement of those early GenAI demos? The sense that we were on the cusp of something truly revolutionary. Yet, as projects progress, a familiar pattern can emerge. We fall back on habits of traditional software development, focusing on building a comprehensive product instead of delivering the unique value GenAI promises.

Have you noticed this shift in your own GenAI journey? Have you found yourself asking, “How can we break free from the MVP mindset that’s worked for other projects?” As I pondered these questions and delved into concepts like Minimum Viable Experiences (MVEs), it hit me that to succeed in GenAI projects, we need to challenge deeply ingrained approaches and embrace a new development paradigm.

Rethinking GenAI Development: From Products to Minimum Valuable Experiences (MVE)

With traditional products, we tend to focus on a checklist of features. But with GenAI, customers crave solutions, outcomes, and experiences that those features enable. Here is the secret: the ‘magic’ of GenAI lies not in the technology itself but in what it allows the user to achieve.

The value of an idea lies in the using of it.

Beyond Rapid Feedback: The GenAI Experimentation Cycle

Rapid customer feedback is crucial, but GenAI demands more. We need a continuous experimentation cycle woven into the fabric of our development. Consider this: selecting models, tuning parameters, refining prompts, cleaning data, and retraining on the fly — none of these fit neatly into the traditional MVP model. GenAI necessitates an environment where the art of scientific exploration meets the discipline of engineering rigor.

The Art and Science of GenAI

Success in GenAI means finding the perfect balance between the art of scientific experimentation and the discipline of engineering rigor. ML Engineers bring their scientific expertise to explore different models, prompts, and data configurations. MLOps ensures these experiments are conducted efficiently and effectively throughout the development lifecycle.

Why MVEs Matter for GenAI

1. Customer-Centricity

MVEs force us to put the customer at the forefront. It compels us to ask, “What is the essential experience our GenAI technology can provide to address a specific need?”

2. Rapid Experimentation

MVEs enable a fast experimentation cycle essential to the dynamic world of GenAI. We rapidly iterate on data, models, prompts, and training processes to optimize performance and deliver the best possible value. For example, consider a business team struggling to extract insights from a sprawling internal knowledge base. An MVP approach might prioritize building a comprehensive search tool with advanced filtering options. In contrast, an MVE approach would focus on providing a simple natural language question interface (e.g., “What were the top three performing products in the Midwest region last quarter?”) that returns a concise answer and supporting visualizations. This highlights the shift from building a feature-rich product to iteratively crafting a valuable customer experience.

3. Overcoming Inertia

Embracing MVEs means breaking free from traditional development cycles burdened by inertia. We learn to value quick iterations and prioritize delivering tangible value over perfecting a distant vision.

The MVE Mindset Shift

Adopting the MVE approach requires a conscious mindset shift, viewing GenAI solutions as a means to deliver experiences that customers genuinely find valuable, and integrating a culture of rapid experimentation within the development process.

leader
Arun Giri
Head of Platform Engineering, Architecture & Lead Enterprise Engineer Union Pacific Railroad

Arun Giri is a visionary technology leader with a decade-long track record of designing and leading groundbreaking platform engineering efforts that have set new standards for scalability, security, and availability. As the architect behind an industry-leading API platform, Arun’s work has laid the foundation for one of the largest Transportation Management Systems (TMS) in North America—NetControl. Over the past 15 years, Arun has taken on increasing levels of responsibility, leading...

GenAI

Responsible Usage & Development of Generative AI Tool

Read more
Data Strategy

Generative AI: Building a Data Strategy for Businesses

Read more