Future of Products with AI

Disclaimer: This piece is speculative. While AI-based systems are evolving rapidly, the concept of "learning applications" as described here is still largely theoretical. Real-time AI-driven software adaptation faces significant challenges, including computational constraints, security risks, and regulatory concerns. Additionally, while AI can automate aspects of software development, it is unlikely to fully replace explicit programming in the foreseeable future. The role of product management will also remain crucial in defining objectives, aligning AI-driven adaptations with business goals, and managing associated risks. This article explores a potential future but does not claim that these advancements are imminent or without limitations.

• Will AI replace software engineers??

In the short term, AI-based coding tools will not replace software engineers but will enable much more development with a significantly less workforce. However, the biggest shift in the product development process may not be workforce reduction but rather a fundamental change in how software is created and evolves.

The real question is:?

• Will software require explicit programming as it does now?

This is something we need to be talking about. AI, particularly generative AI, has the potential to enable applications that evolve and adapt autonomously. While the specific nomenclature for such applications is still unclear, I refer to them as "learning applications."

Learning Applications:

Imagine installing an application on your device. Initially, you download a basic version, but over time, the app evolves by learning from your interactions and potentially from other applications on your device.

This transformation does not occur through centralized updates or external interventions. Instead, the app contains a generative AI-based learning model that enhances its own codebase by adapting to your preferences-even those you haven’t explicitly stated.

More importantly, this metamorphosis happens independently of an internet connection. The adaptations are entirely unique to your device, meaning no two copies of the application will have exactly the same features, UI, or codebase.

Consider a broader example: Suppose you install an operating system that includes core applications. Over time, instead of requiring third-party apps, the OS itself writes applications tailored to your specific needs. This represents a form of on-device hyper-personalization that is potentially more secure and private since your data never leaves your device.

How different from federated learning:?

One may ask: "How is this different from federated learning?"

The key distinction lies in the scope of adaptation. Take the example of federated learning in text prediction. Federated learning enables the addition of new words to the application’s database based on your typing habits, but it does not change the application's structure, UI, or functionality.

In contrast, a learning application with an embedded AI model could go much further—it would dynamically modify itself, adding new features and interfaces based on real-time usage data.

Implications for the product development process:?

So, what are the implications for product development?

Potentially, they are drastic.

This approach could reduce not only development work but also the need for extensive user research, feedback loops, and data analysis. There would be less need to hypothesize about user behavior since applications would autonomously adapt to meet real-world usage patterns. Developers would no longer need to track feature requests in the same way because necessary changes would be implemented dynamically by the software itself.

This could significantly shorten the traditional iterative product development cycle. However, it does not mean human intervention will become obsolete. Interventions will still be necessary for security, regulatory compliance, and ethical oversight.

One thing is certain - product management will shift from a primarily people-management role to a system-management role, overseeing how AI-driven applications evolve rather than directly dictating product iterations.

While this vision remains speculative, technological advancements may eventually catch up to make it a reality. The question is not whether AI will replace software engineers, but how AI will reshape the way software is created, deployed, and maintained in the future.