A Grey Literature Review of AI-Native Applications

2024-06-04 · Source: cs.SE updates on arXiv.org · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Advanced, extended

Summary

This grey literature review, analyzing 106 sources from December 2022 to May 2025, establishes a comprehensive understanding of AI-native applications. It defines them by two core pillars: AI as the central intelligence paradigm and their inherently probabilistic, non-deterministic nature. The study identifies 7 core elements and 8 key characteristics, such as "Automation and Intelligent Workflow" (44 studies) and "Multimodal and Conversational Capabilities" (41 studies). Critical quality attributes include "Reliability and Robustness" (73 studies) and "Usability and User Experience" (72 studies). An emerging technology stack features "LLM Orchestration and Integration Platforms" (66 studies) and "Frontend and UI Tooling" (86 studies). The review also outlines 9 opportunities and 7 challenges, proposing a dual-layered engineering blueprint to guide practitioners and future research in this rapidly evolving software engineering paradigm.

Key takeaway

For AI Architects and Software Engineers designing AI-native applications, prioritize mastering LLM orchestration frameworks and implementing hybrid cloud/on-device deployment strategies to balance privacy, latency, and cost. You must also establish robust AI-specific observability, moving beyond traditional debugging to continuously monitor probabilistic model behavior, data drift, and output quality, which is critical for long-term system integrity and reliability.

Key insights

AI-native applications fundamentally redefine software by centering AI as the probabilistic core, distinct from AI-assisted features.

Principles

• AI-native applications are defined by AI as the core intelligence, not an auxiliary feature.
• Their inherently probabilistic nature reshapes traditional software quality attributes.
• Continuous learning and adaptivity are essential for true AI-nativeness.

Method

The study proposes a dual-layered engineering blueprint for AI-native applications, offering actionable design guidelines and technical recommendations for practitioners and researchers.

In practice

• Master LLM orchestration frameworks like LangChain or LlamaIndex.
• Implement hybrid cloud and on-device deployment strategies.
• Establish AI-specific observability for probabilistic outputs.

Topics

• AI-Native Applications
• Large Language Models
• Software Engineering
• Grey Literature Review
• LLM Orchestration
• AI Observability
• System Quality Attributes

Code references

• llc202jy/GLR-on-AI-native-applications

Best for: AI Scientist, AI Architect, Software Engineer, Research Scientist

Related on AIssential

• A Practical Guide to Becoming an AI-Native Engineer — AI-native engineering requires orchestrating AI agents with disciplined practices and robust security, shifting focus from coding to verification and context.
• Model-Native Computing Architecture: Envisioning Future System Architecture Through the Lens of Computer Architecture — The Model-Native Computing Architecture proposes ICAM, a six-layer framework, to unify LLM system design through a computer architecture lens.
• How can you test your code when you don’t know what’s in it? — Testing AI-driven agentic workflows requires adapting to LLM non-determinism, moving beyond rigid assertions to intent-based validation.
• What actually makes a system AI-Native — AI-native systems integrate intelligence as a first-class architectural primitive, evolving beyond simple LLM calls.
• Another try at LLM workflow automation — Effective LLM automation requires extensive scaffolding and iterative tooling, not just smarter models.
• Beyond vibe coding: The five building blocks of AI-native engineering — AI-native engineering demands structured orchestration of agents, specialized models, methodologies, specs, and context for production-grade software.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.SE updates on arXiv.org.