← Sticker Wall AGENT · DETAIL

Dojo AI Testing Agent Operations Generating 15K RMB Monthly

Workflow: The input consists of the client repository's Swagger interface documents or code changes, which are processed by Dojo c

AGENT

Key Fields

FIELD STAMPS

IndustryMarketing / Advertising

RegionGlobal

ScaleSME

ChannelOnline

🔧 Workflow

The input consists of the client repository's Swagger interface documents or code changes, which are processed by Dojo combined with GitHub Actions to automatically generate test cases and run regressions. The output is a test report complete with failure root cause conclusions. Humans act solely as referees to verify whether the AI-flagged failures are genuine, and once approved, reports are sent back to the client on a project or monthly basis. A single project yields 3,000 to 5,000 RMB, and maintaining 3 to 5 clients simultaneously results in a monthly income of approximately 15,000 RMB (calculated based on figures within the card; the case metrics are yet to be independently verified). During peak seasons, a delivery cycle occurs every 2 to 3 days.

🛠 Setup Requirements

Requires familiarity with at least one backend language and CI pipelines, as well as the ability to read interface documents. The tech stack uses Dojo alongside self-built GitHub Actions or Jenkins triggers, allowing the first client project to be up and running in about 2 weeks. Start by clarifying the specific problem to solve rather than using AI for the sake of AI, maintaining a progressive introduction starting with assistance followed by automation.

🧰 Toolchain

🔧 Dojo AI Testing Agent
🔧 GitHub Actions
🔧 Swagger or OpenAPI documentation
🔧 Slack or Lark (Feishu) notifications
🔧 Postman or HAR import tools

💰 Revenue

① Small-team testing operations (primary revenue): clients pay service fees per project, ranging from 3,000 to 5,000 RMB per project. Managing 3 to 5 clients simultaneously yields a monthly income of about 15,000 RMB, accounting for roughly 100% of monthly revenue (estimated based on card figures; case metrics pending follow-up verification); ② Annual packaged subscription: similar organizations pay an annual fee, with 4 clients totaling 120,000 RMB annually (equivalent to 30,000 RMB/year per client; the exact number of signed clients cannot be verified, and their share of revenue is unspecified); ③ Custom integration projects: helping clients integrate Dojo-like agents into private repositories and CI, charging service fees per project (quotation, project count, and revenue share are all missing); ④ Opportunity item - Performance-based revenue sharing: after clients introduce AI testing, writing time is reduced by 60% and maintenance costs by 80% (case data, unverified independently). Commissions are charged based on saved work hours, though the commission rate and revenue share are yet to be determined.

💸 Cost

Dojo subscriptions cost several hundred RMB per month, and API calls are pay-as-you-go, keeping total tool costs under 1,000 RMB. If the client provides a private deployment environment, server costs are covered by the client, leaving the individual responsible only for subscription fees and minor cloud function expenses.

⏱ Time Investment

Investing 2 to 3 hours daily, focused on manual review and failure root cause determination, while leaving the rest to automated agent runs. A weekly client sync and report compilation is conducted, keeping total time under 15 hours.

🚀 Getting Started

Start by using Dojo in an open-source project to generate tests for an existing Swagger repository and run a successful regression, then record a short video or write a retrospective to post in tech communities. Secure the first paid trial client before discussing long-term services. Proceed from practical problems, having AI generate test case drafts first, conducting manual reviews of edge cases, and gradually transitioning to automated execution.

🔑 Keys to Success

✅ Humans act only as failure root cause referees rather than writing test cases for AI
✅ Starting from interface regression is more stable than UI and easier to integrate with CI
✅ Using public repository examples to prove capability and lower customer acquisition trust costs
✅ Building client dependency through continuous delivery of reports to prevent the loss of one-off projects

⚠️ 风险

⚠️ If clients build their own internal AI testing teams, the demand for outsourced operations will rapidly disappear
⚠️ AI hallucinations generating seemingly correct but invalid assertions, where missed defects lead to difficult-to-assign production incident responsibilities
⚠️ Test case quality degrades with context length, requiring sharded processing by category; otherwise, large-batch single-run generation results are unreliable

📌 Real Cases

📌 The 2026 Prism Space Guide points out that AI-generated test cases require manual review, and multiple teams have already used workstations like LangGraph plus Playwright to compress test case generation time from hours to 30 minutes, with operators charging 3,000 to 5,000 RMB per project.
📌 Alibaba Cloud Developer Community 2026 test engineer salary data shows that test engineers who can train AI command 800K, whereas those who only write test cases are submitting resumes, indicating a sustained paying market for outsourced operations and custom integration services.
📌 In a case mentioned by SegmentFault where an MCN agency built a fully automated workflow, 1 person's review replaced the 8-hour daily work of 3 operations staff, charging an annual fee of 120,000 RMB. The same service logic can be replicated in AI testing outsourced operations.

Sources