AI Native explores how artificial intelligence is changing the way we interact with software
AI Native explores how artificial intelligence is changing the way we interact with software
I created this resource to document the most unique and well-crafted AI-native design patterns to help inspire today's builders to push boundaries in the age of AI.
The products that truly shine aren't just using AI as a checkbox feature, they're fundamentally rethinking how users interact with technology. If we're going to build the next generation of transformative products, we need to break free from old patterns and embrace this new paradigm of design.
Most of the apps, products and services we will all rely on a daily basis over the next decade either don’t exist yet or are just beginning to take shape. Some companies will successfully adapt to this shift, while others won’t, and a wave of entirely new companies will emerge.
This moment presents a chance to rethink our approach to technology, an opportunity to shape the future through design.
AI-native vs AI-assisted
AI-native vs AI-assisted
- AI-native products are built entirely around AI, taking charge with a focus on independent operation, delivering complex and detailed results that often require more processing time. If you remove the AI from the platform, the product will not work and the customers won't buy it.
- AI-assisted products are designed to support human efforts, offering AI-driven features like quick responses or automation while still being usable without AI, providing faster but less in-depth outcomes.
There are some key differences between AI-first and AI-assisted products that shape how they function and serve users.
AI-native products are built entirely around AI, taking charge with a focus on independent operation, delivering complex and detailed results that often require more processing time. If you remove the AI from the platform, the product will not work and the customers won't buy it.
AI-assisted products are designed to support human efforts, offering AI-driven features like quick responses or automation while still being usable without AI, providing faster but less in-depth outcomes.
Essentially, AI-first products are AI-dependent, while AI-assisted ones use AI as an enhancement rather than a necessity.
Flipping the design process on its head
Flipping the design process on its head
Developing AI-first products requires a completely different way to work. While traditional software development and design process starts with a problem, explores possibilities first and then builds with technology, LLM-based design uses the technology itself as the primary exploration tool. Because LLMs are probabilistic rather than deterministic, this prioritizes experimentation, rapid prototyping and a strong demo culture to discover what's actually possible through direct interaction with the technology.
This isn't theoretical, take Perplexity as an example. They usually begin by testing LLMs with basic command-line prototypes. Only after validating the AI's capabilities did they move into designing the user experience.
Developing AI-first products requires a completely different way to work. While traditional software development and design process starts with a problem, explores possibilities first and then builds with technology, LLM-based design uses the technology itself as the primary exploration tool. Because LLMs are probabilistic rather than deterministic, this prioritizes experimentation, rapid prototyping and a strong demo culture to discover what's actually possible through direct interaction with the technology.
This isn't theoretical, take Perplexity as an example. They usually begin by testing LLMs with basic command-line prototypes. Only after validating the AI's capabilities did they move into designing the user experience.
Start with the AI
Start with the AI
Product teams with an AI-first mindset:
Test what's possible with prompts before jumping into mockups.
Build comprehensive output libraries to guide development.
Discover unique value propositions enabled specifically by AI.
Match AI capabilities to genuine user needs.
This approach acknowledges an uncomfortable truth: beautiful mockups mean nothing if the underlying AI can't deliver. Teams that validate technical feasibility first avoid the disappointment of designing experiences that can't be built.
Product teams with an AI-first mindset:
Test what's possible with prompts before jumping into mockups.
Build comprehensive output libraries to guide development.
Discover unique value propositions enabled specifically by AI.
Match AI capabilities to genuine user needs.
This approach acknowledges an uncomfortable truth: beautiful mockups mean nothing if the underlying AI can't deliver. Teams that validate technical feasibility first avoid the disappointment of designing experiences that can't be built.
Evolve & iterate
Evolve & iterate
AI-native product teams heavily rely on user feedback shape how the AI behaves.
Unlike traditional software with predictable behavior, AI-first products require continuous refinement based on real-world usage. Intercom demonstrates this in their support agent tools, they're still discovering what's truly possible with their AI implementation through ongoing experimentation and user feedback.
The development cycle never really ends with AI-native products. Teams must:
Test constantly with actual users.
Refine prompts based on real-world performance.
Stay up-to-date about emerging AI capabilities.
AI-native product teams heavily rely on user feedback shape how the AI behaves.
Unlike traditional software with predictable behavior, AI-first products require continuous refinement based on real-world usage. Intercom demonstrates this in their support agent tools, they're still discovering what's truly possible with their AI implementation through ongoing experimentation and user feedback.
The development cycle never really ends with AI-native products. Teams must:
Test constantly with actual users.
Refine prompts based on real-world performance.
Stay up-to-date about emerging AI capabilities.
Unlike traditional software with deterministic behavior, LLMs require extensive testing to understand their capabilities and limitations.
The probabilistic nature of large language models means you can never fully predict their output. Successful teams embrace this uncertainty through rapid prototyping and experimentation cycles. What starts as a rough demo often evolves directly into the shipped product, with refinements happening continuously in response to real usage.
Unlike traditional software with deterministic behavior, LLMs require extensive testing to understand their capabilities and limitations.
The probabilistic nature of large language models means you can never fully predict their output. Successful teams embrace this uncertainty through rapid prototyping and experimentation cycles. What starts as a rough demo often evolves directly into the shipped product, with refinements happening continuously in response to real usage.
A new approach for a new era
A new approach for a new era
AI-native product design represents more than an incremental change, it's a fundamental rethinking of the product development process.
The most successful teams have already adapted to this reality. They prioritize technical experimentation over pixel-perfect mockups, embrace the uncertainty of AI capabilities and build cultures where continuous refinement is the norm.
AI-native product design represents more than an incremental change, it's a fundamental rethinking of the product development process.
The most successful teams have already adapted to this reality. They prioritize technical experimentation over pixel-perfect mockups, embrace the uncertainty of AI capabilities and build cultures where continuous refinement is the norm.