WTF is AI-Native Product?

In my last post, I talked about how focusing on a problem that couldn’t be previously solved without AI is the key to building a successful AI product. Tapping into an unfulfilled user need, so to speak. Essentially, it’s a call to create “AI-native applications” instead of “AI-enhanced features”. But what exactly does “AI native” mean?

Back in 2011, Albert Wenger, Partner at USV, described the term “mobile native” application in the context of the disruption occurring in mobile applications:

We are fascinated by the disruption underway in mobile applications. Carriers seem to have lost their role as gatekeepers for applications as smartphone sales are rapidly ramping and “app stores” or direct downloads are the new distribution models. This is exciting as it opens up a whole new arena for startups to compete in. Here is some of our early thinking about this with the goal of getting a dicussion going.

The challenge for startups (and investors!) has been identifying opportunities that are “native” to the new platforms. By “native” we mean opportunities that simply did not exist previously and cannot exist without the phone. For instance, we would not consider delivering breaking news to a mobile a native opportunity, as a startup rarely has a better chance of being “CNN for mobile” than CNN does. Native opportunities are the ones that make use of unique capabilities of mobile platforms.

In the mobile world, native apps hinged on the unique capabilities of mobile platforms, like real-time location services, touch interface, and multimedia input (audio, image, and video). These capabilities provided developers with new building blocks to innovate. For instance, Uber and Instagram would not have been possible without location services, connectivity, and multimedia input. But as Wenger observed, it was not the uniqueness of each capability that was important. Rather, the convergence of these technologies on a single device is what brought about a paradigm shift.

Such platform shifts often trigger changes in human behavior. For example, the widespread adoption of instant messaging facilitated by fast broadband access has rendered phone calls increasingly old-fashioned. Similarly, Instagram filters have made it socially acceptable—even trendy—to share pictures publicly. As users grow more comfortable with these behavioral shifts, startups have the opportunity to introduce solutions that will be adopted much faster and can become solid businesses.

New primitives for AI native applications

So, how can we apply this concept to AI? Let’s outline the capabilities that AI and LLMs have brought to the table:

  1. Natural Language Understanding & Generation
    • Improved understanding of human language results in more accurate text analysis and extraction of meaning.
    • Enhanced capabilities to generate human-like text enable seamless communication, content creation, and text translation.
  2. Multimodal Interaction & Generation
    • Integration of various communication modes (e.g., text, voice, images) to achieve seamless user interaction.
    • Generation of human-like multimedia content e
  3. Contextual Awareness & Reasoning
    • Better comprehension of the context surrounding multimodal inputs, allowing more relevant and accurate responses or actions.
  4. Personalization
    • Hyper-personalized and context-aware experiences, tailored to individual preferences and needs.
  5. Agent-Based Task Automation
    • Ability to execute tasks and achieve outcomes based on previously specified goals.

An AI-native application is one that relies on at least one of these capabilities to provide core value to its users—or combines several of them in a unique way. The most obvious example is an AI personal assistant, aka Siri that actually works. To provide true breakout service, it’ll combine 90% of these primitives, if not all.

AI influencers don’t pass on the opportunity to call AI and LLMs a new platform, similar to mobile. I disagree with such a description. Unlike mobile, AI and LLMs primarily serve as an underlying technology that enhances and transforms existing platforms, rather than being a standalone platform in and of itself.

Mobile and web provided a foundation for developing a wide range of applications, services, and experiences across various domains. AI, in its turn, is the ultimate Lego block for new emergent platforms like AR and VR, and calling it a platform is the same as considering any programming language a platform.

AI native vs. incumbents

When it comes to building native AI apps, it’s essential to first identify problems uniquely positioned to be solved with the capabilities it made available. As Elad Gil aptly wrote, every technological shift distributes value between the incumbents and startups differently. Understanding the dynamics of your target market and how value accrues there is crucial.

For instance, if you’re building a meeting notes summarizer, you’ll have a hard time capturing value since the incumbents control the platforms where meetings actually happen. But if you’re building an AI native experience from day one, and it turns out to be valuable, the incumbents will have a tough time copying your app. Why? Because they’d have to rebuild their services from the ground up, which means sacrificing their existing user segments and revenue streams — the innovator’s dilemma at its finest!

In short, the key to success in the AI product world is to identify those juicy, AI-native opportunities and exploit the unique capabilities that LLMs and other AI technologies have to offer.

Happy hunting, folks!

Leave a Reply

Your email address will not be published. Required fields are marked *

Up Next:

The Future of Social Media Monetization: Can Social Networks Thrive Without Ads?

The Future of Social Media Monetization: Can Social Networks Thrive Without Ads?