The Applications Layer
Why the AI application layer defies broad generalisation and requires individual evaluation of each company's value creation potential.
The application layer is too broad to make a sector call. Some AI applications will generate extraordinary value; others will not. Unlike the chips or infrastructure layers, where a small number of dominant players can be identified, the applications layer spans every industry and use case.
Evaluation Approach
Assess Individually, Not as a Category
Treating "AI applications" as a single investment theme is a mistake. A medical imaging AI company and an AI-powered scheduling tool have almost nothing in common beyond the label. Each application company must be evaluated on its own merits: the problem it solves, the willingness of customers to pay, and the durability of its advantage.
Look for Genuine Productivity Gains
Many companies are adding "AI" to their marketing without fundamentally changing what they do. The distinction that matters is whether the product creates genuine, measurable productivity gains — or merely repackages existing functionality with an AI label.
The Agentic Shift
The shift from chat interfaces to agentic systems represents a genuine transformation in what applications can do. Applications that automate multi-step workflows — not just answer questions — create the most value. An application that can independently gather data, analyse it, draft a report, and iterate based on feedback is categorically different from one that responds to individual prompts.
The fundamental question for any AI application: does it create value that users will pay for repeatedly? One-time novelty is not a business. Recurring value creation is.
Changing Industry Economics
Senior engineers managing AI agents instead of junior developers changes the economics of entire industries. A small team with effective AI tooling can now produce output that previously required a much larger organisation. This dynamic favours application companies that enable this leverage.
The most promising AI applications tend to sit at the intersection of high-value tasks, structured workflows, and clear ROI measurement. If a customer can calculate exactly how much time or money the application saves, adoption and retention will follow.
Related
- AI Stack Framework — How the applications layer fits within the broader AI value chain
- AI Agents — The shift from chat interfaces to autonomous multi-step systems
- Inference Scaling — Falling inference costs that make more applications economically viable