AI Selling Shovels — Earn a Million a Year?

OpenClaw sparks an industry

A small, crustacean-branded project called OpenClaw has unexpectedly become a commercial phenomenon, reportedly drawing some 230,000 stars on GitHub and dominating Chinese discussion boards. The headline is simple: tools that let AIs actually “do” things — not just chat — are creating a marketplace of services around deployment, configuration and maintenance. Who profits when AI becomes useful? Those selling the shovels.

What OpenClaw actually does

OpenClaw is a framework for building AI agents that can cross applications and systems to complete real‑world tasks: calling search engines, simulating clicks on airline sites, verifying dates and prices, and even trying to navigate payment flows and captchas. But it is not a language model itself; users must plug in APIs, set permissions and debug errors — a technical hurdle for ordinary people. It has been reported that on domestic second‑hand platforms remote deployments and on‑site installation services are bustling, with advertised starting fees around ¥300 and one freelance installer, identified as “Michael,” reportedly claiming $20,000 in one week and projecting six‑figure annual income from setup services. Overseas, some OpenClaw projects have even been listed for sale at prices reportedly reaching $300,000.

Markets, geopolitics and the “sell the shovel” logic

This wave follows earlier cottage industries around ChatGPT account registration and prompt‑engineering courses such as DeepSeek. Access patterns matter: many Chinese users initially turned to ChatGPT with aspiration but found barriers — foreign phone verifications and the need for VPNs — that helped spur local intermediaries and alternative services. At the same time, major large‑model vendors have reportedly introduced coding‑focused API plans to smooth one‑click deployments, making the tools easier to package and sell. Geopolitical frictions, export controls and differing regulatory environments are sharpening incentives for local installers and integrators to fill gaps left by cross‑border service frictions.

Labor effects and the big question

The broader worry is familiar: tools that boost output can also intensify work. Berkeley researchers have found that handing tasks to AI does not necessarily shorten working hours; people tend to expand what they produce, a phenomenon called “task expansion.” Manpower Group’s 2026 Global Talent Outlook reportedly shows regular AI use up sharply — to 45% — and many workers seeking stability amid rapid change. So will OpenClaw and its imitators free people from drudgery, or will they just create new forms of micro‑service labor and pressure everyone to work faster? For now, the market for shovels is booming — and the answer remains to be seen.