← Sticker Wall AGENT · DETAIL

Factory AI Software Factory Full-Process Automation Generating 150k Monthly

Workflow: Receive customer software requirement tickets daily, dispatching multiple Droid agents to respectively execute coding, c

AGENT

Key Fields

FIELD STAMPS

IndustrySaaS / Enterprise Software

RegionUS

ScaleSME

ChannelOnline

🔧 Workflow

Receive customer software requirement tickets daily, dispatching multiple Droid agents to respectively execute coding, code review, testing, and deployment tasks while processing multiple projects in parallel. Humans only handle requirement clarification and final acceptance, with the system automatically generating deployment versions and pushing to CI/CD pipelines. After each Droid produces code, another Droid performs cross-review, automatically fixing compilation errors and test failures, ultimately merging into the main branch and triggering containerized deployment. Ticket queues for multiple projects are automatically allocated by priority, with human arbiters only intervening at critical decision points.

🛠 Setup Requirements

Requires proficiency in the Factory AI platform or its open-source implementation, along with basic DevOps and CI/CD configuration capabilities. Core investments include platform subscription fees and API call costs, with setup taking about 2 to 4 weeks, starting with small internal project trial runs. The configuration phase requires defining Droid role templates, task splitting rules, code review standards, and deployment trigger conditions, while preparing access tokens for GitHub repositories and container image registries. It is recommended to first validate the process using an open-source software factory framework before migrating to commercial platforms to control early costs.

🧰 Toolchain

🔧 Factory AI Platform
🔧 GitHub Actions
🔧 Docker
🔧 OpenAI API
🔧 VS Code
🔧 Kubernetes

💰 Revenue

① Replicator tier - automated delivery for SMEs billed per project (main revenue): SMEs pay 30k to 80k RMB per project, delivering 2 to 4 small projects monthly yielding 60k-320k RMB, netting about 150k RMB/month with a profit margin of around 70%. Nearly 100% of monthly revenue comes from this stream (calculated based on unit price × number of projects, case-by-case nature, no external independent verification seen); ② Recurring revenue - customers shifting to monthly subscription procurement for continuous iterative services, with unit prices and contract numbers not yet publicly disclosed, and the proportion of total revenue unspecified (case retelling, pending independent verification); ③ Efficiency moat - single Droid conversations average about 8 minutes, with 60% ending within 15 minutes, while Missions+Droid conversations generally take 2 hours and 37% of tasks exceed 4 hours, running unattended continuously for 40 days, allowing one person to run multiple projects simultaneously (this portion's contribution to revenue is not broken down, media estimates without independent review); ④ Opportunity item - licensing self-hosted software factory solutions to other development teams, with pricing and revenue potential yet to be determined, and proportion left blank.

💸 Cost

Platform subscription fees and model API costs are approximately 10,000 RMB monthly, fluctuating based on the number of parallel Droids. Open-source solutions can reduce infrastructure costs by 70%, but require an additional investment of about 2,000 RMB monthly for self-hosted server fees. GitHub Actions free tier is usually sufficient, and container image storage costs depend on project scale.

⏱ Time Investment

Invests about 20 hours per week in requirement clarification, quality spot-checks, and client communication. About 8 hours are spent writing and splitting tickets, 6 hours on code review and acceptance testing, with the remaining time handling client feedback and process optimization. No manual monitoring is required during automatic system runs, allowing Droids installed at night and weekends to produce continuously.

🚀 Getting Started

Step one: register for the Factory AI Enterprise Edition or deploy the open-source software factory framework, running a real internal development ticket through the entire process from coding to deployment. After confirming stable output quality, take on small software outsourcing gigs on overseas freelance platforms to validate willingness to pay. Start by cutting into a single vertical field, such as e-commerce backend management systems or data visualization websites, to more quickly build a reusable template library and improve delivery speed.

🔑 Keys to Success

✅ End-to-end automation reduces marginal delivery costs
✅ Multi-Droid parallel processing increases output per unit of time
✅ Humans act only as arbiters to safeguard quality standards
✅ Combining per-project billing with subscriptions creates multiple repurchases
✅ Template library accumulation builds cross-client reusability

⚠️ 风险

⚠️ Unstable generated code quality leads to rework or client claims
⚠️ High platform dependency where official API policy changes could drive up costs
⚠️ Enterprise clients have high code security and audit requirements, and automated processes may not meet compliance standards
⚠️ Droid concurrent execution conflicts lead to merge failures and deployment rollbacks

📌 Real Cases

📌 Factory AI officially disclosed that its Droid system can automatically complete the coding, testing, and deployment closed-loop from requirement tickets, with enterprise clients purchasing this capability via subscription
📌 GitHub open-source project owainlewis/factory demonstrates the feasibility of building a personal software factory using AI agents, with multiple developers setting up self-hosted automated delivery systems and undertaking outsourcing based on it
📌 36Kr reported that a Silicon Valley Agent worked continuously for up to 40 days, implying that similar Droid systems can run stably in long-term unattended scenarios, supporting high-frequency parallel delivery

Sources