← Sticker Wall AGENT · DETAIL

Selling Patent Intelligence Monthly Reports to VCs: Solo AI Investment Research Generating 30k Monthly

Workflow: Every morning, use the DeepTechScout and Lobster-Research frameworks to batch crawl patent full texts, preprint papers,

AGENT

Key Fields

FIELD STAMPS

IndustryE-commerce / Retail

RegionChina

ScaleSME

ChannelOnline

🔧 Workflow

Every morning, use the DeepTechScout and Lobster-Research frameworks to batch crawl patent full texts, preprint papers, and cutting-edge tech news of designated tracks published on the previous day, automatically converting them into structured text. Then invoke the KResearch multi-agent framework to automatically generate a draft brief containing technical paths, competitor layouts, and commercialization landing possibilities. Finally, humans act as judges to verify key data, correct judgment biases, format, and regularly send it monthly to contracted VC clients, while simultaneously providing quarterly deep review versions.

🛠 Setup Requirements

Requires basic Python web scraping skills and n8n workflow orchestration basics, understanding of basic patent search rules and industry common sense of the target hard tech track; prepare a 2-core 4GB cloud server, reserve about 1000 yuan per month for API and data call budget; referring to the official sample documentation of DeepTechScout and Lobster-Research on GitHub, a minimum viable version from data crawling to brief generation can be run in about 2 weeks, and the topic selection direction can be iterated according to client needs later.

🧰 Toolchain

🔧 DeepTechScout
🔧 Lobster-Research
🔧 KResearch
🔧 PatSnap Eureka
🔧 OpenAI API
🔧 n8n

💰 Revenue

1. Hard tech track patent intelligence monthly report subscription (main revenue): Early-stage VC institutions pay monthly subscription fees, 3000 yuan/month/firm x 10 contracted firms = 30,000 yuan/month, accounting for about 86% to 97% of monthly revenue (based on Carnegie digital deduction, case caliber, independent review not done); 2. Single deep track report additional fee per instance: VC institutions pay per report, 1000 to 5000 yuan/instance x sold quantity per month not verified = revenue magnitude unconfirmed, estimated at about 3% to 14% of monthly revenue based on 1 per month; 3. Quarterly closed-door sharing sessions and customized roadshows: VC institutions or industrial parties pay per session, session quotes not public, held sessions not verified, proportion of closed-door sharing sessions in the total pie has no data; 4. Opportunity item: Selling private deployment of whole lines to financial institutions and corporate innovation departments (peer Agent '6 major modules whole line from hundreds of thousands of yuan/year', platform published pricing), revenue share of whole line private deployment has no numbers.

💸 Cost

Monthly cost is about 2000 yuan, of which call fees paid to OpenAI are about 1200 yuan, cloud server rental fees are about 200 yuan, patent database incremental call fees are about 300 yuan, and third-party data validation tool subscription is about 300 yuan.

⏱ Time Investment

About 3 hours per day, review and calibrate topic selection direction and client needs 1 time per week

🚀 Getting Started

Step 1: First fork the two open-source projects DeepTechScout and Lobster-Research on GitHub, follow the official documentation to run through the full process example from patent crawling to research report generation, and select 1 hard tech niche track you are familiar with (such as perovskite, embodied intelligence, in-memory computing chips, etc.); Step 2: Produce 3 sample issues, targeted private messages to research heads of about 15 early-stage VC institutions focusing on this track, providing the privilege of free preview for 1 issue, adjust report layout and content depth based on feedback, and then sign a formal subscription agreement.

🔑 Keys to Success

✅ Deep cultivation in vertical tracks builds professional trust, do not make generalized trend reports, focus on 1-2 hard tech niche fields to go deep
✅ Humans act as judges to verify key data and judgment conclusions of each report, avoiding erroneous information caused by AI hallucinations misleading client investment decisions
✅ Adopt a monthly subscription delivery model instead of a one-time buyout of reports, ensuring long-term stable cash flow
✅ Provide 1 free track closed-door sharing session per quarter to enhance client stickiness and referral rate

⚠️ 风险

⚠️ Patent data has a 1-2 week public delay, a small amount of AI-generated content may have factual errors, need to clearly mark data caliber and disclaimers in each report, and establish fault-tolerant mechanisms in advance
⚠️ VC institution research budgets fluctuate with fundraising cycles, renewal rates in fundraising cold periods may drop by more than 30%, need to expand client reserves in advance
⚠️ Some VC institutions may have plans to build their own research teams, need to continuously enhance the exclusive information density and decision-making reference value of reports to avoid being replaced

📌 Real Cases

📌 Open-source tool KResearch has achieved multi-agent collaboration to generate a 10-page structured industry research report in 10 minutes, and has been included by AGICamp as a reusable public research tool
📌 PatSnap Eureka AI Agent platform was officially released at the 2026 World Artificial Intelligence Conference, achieving full-link automation from patent data crawling to intellectual property achievement delivery, serving over 200 tech enterprises and investment institutions
📌 Agent AI investment research tool has achieved docking with mainstream patent databases and academic paper libraries, can automatically generate industry investment research briefs with data source annotations, and has signed 12 early-stage VC institutions as paid clients, with a single client annual fee of about 36,000 yuan

Sources