Unpopular Opinion: Not Every Product Needs AI

2026-07-26 · Source: LLM on Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Advanced, short

Summary

A critical perspective on AI integration argues against its indiscriminate application in product development, asserting that AI should not be a default solution. The analysis highlights that incorporating AI often introduces substantial complexity, leading to systems that are more expensive, less deterministic, and challenging to test, debug, and understand. It advocates for a fundamental shift in architectural thinking, moving from "Where can we add AI?" to "Which part of this problem actually requires reasoning?". The article introduces a mental model distinguishing between a "Service" for predictable execution, a "Skill" for AI performing a specific capability like summarise_document(), and an "Agent" for AI-driven decision-making when subsequent steps are uncertain. This approach promotes an architecture that maintains deterministic systems for predictable functions, integrating AI only where genuine reasoning, interpretation, or ambiguity is required.

Key takeaway

For AI Architects designing new systems, you should critically evaluate whether AI truly addresses a reasoning or ambiguity problem. Avoid defaulting to AI for predictable workflows, as this introduces unnecessary complexity, cost, and non-determinism. Instead, design architectures that keep deterministic functions separate, integrating AI capabilities only where they provide clear leverage for interpretation or decision-making, ensuring a more robust and maintainable product.

Key insights

AI should only be integrated where true reasoning or ambiguity exists, not as a default solution.

Principles

• Complexity does not equal intelligence.
• Prioritize simplest reliable system.
• AI must earn its architectural place.

Method

Divide system components into Services (predictable logic), Skills (AI for specific capabilities), and Agents (AI for decision-making and workflow control).

In practice

• Use deterministic systems for known flows.
• Apply AI for reasoning, interpretation, ambiguity.
• Keep databases, billing, permissions deterministic.

Topics

• AI Architecture
• Agentic AI
• System Design
• Software Engineering
• Deterministic Systems
• AI Integration Strategy

Best for: AI Product Manager, Product Manager, CTO, AI Architect, AI Engineer, Director of AI/ML

Related on AIssential

• The Best AI Architecture Isn’t a Pattern. It’s a Spine. — AI-native software requires a deterministic spine to validate probabilistic model outputs.
• Beyond Prompt Engineering: Designing a Production-Oriented Multi-Agent AI System with Google ADK — Production-grade AI systems demand robust software architecture, treating LLMs as intelligent components within a larger distributed system.
• AI Is Not a Library: Designing for Nondeterministic Dependencies — AI systems are nondeterministic dependencies, requiring fundamental shifts in software architecture and engineering practices.
• Code Is Cheap. Engineering Judgement Is Now the Scarce Resource — AI makes code cheap, shifting value to engineering judgment, intentional ownership, and understanding "why" things work.
• The Simplicity Paradox: Why Modern AI Still Depends on Classic Data Engineering — Modern AI's perceived simplicity masks its deep reliance on classic data engineering for robust operation.
• The Economics of Agentic AI: Engineering for Imperfection — Agentic AI's economic viability and safety depend on deterministic architectural governance, not solely on model intelligence.

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 LLM on Medium.