Narrow AI

Definition

AI systems designed for a specific task or domain, such as image classification or fraud detection. All commercially deployed AI today is narrow, despite the generality of modern LLMs.

Why it matters

Despite the hype around AGI, every dollar of AI revenue today comes from narrow AI. This is an important reality check for investors and executives: the value of AI is in solving specific problems well, not in achieving artificial general intelligence. LLMs are broader than previous narrow AI systems, but they are still narrow in important ways: they cannot learn from experience without retraining, they lack persistent memory across sessions, and they fail unpredictably on out-of-distribution tasks. Building a successful AI business means identifying specific problems where narrow AI delivers clear ROI, not waiting for general intelligence to solve everything.

In practice

The most commercially successful AI systems are narrow: Google Search's ranking algorithm, Tesla's lane-keeping system, Netflix's recommendation engine, JPMorgan's fraud detection. Even LLM-powered products are deployed for narrow tasks: GitHub Copilot for code completion, Grammarly for writing assistance, Harvey for legal research. Companies that try to deploy LLMs as general-purpose 'do anything' tools typically see low adoption and unclear ROI. Companies that deploy them for specific, well-defined workflows see much higher success rates. The pattern: narrow the use case, measure the outcome, expand from there.