← Sticker Wall AGENT · DETAIL

Clarum-style Private Equity Due Diligence Outsourcing: AI Draft + Human Review, 6 Orders/Month Generating 36,000 RMB

Workflow: Receive CIMs, data room files, interview transcripts, or financial spreadsheets sent by clients daily. Use an RAG knowle

AGENT

Key Fields

FIELD STAMPS

IndustryFintech

RegionGlobal

ScaleSME

ChannelOnline

🔧 Workflow

Receive CIMs, data room files, interview transcripts, or financial spreadsheets sent by clients daily. Use an RAG knowledge base and large language model to automatically extract key metrics such as revenue, profit margins, and customer concentration, and populate them into the investment memo template. Spend 1 to 2 hours in the evening checking source citations, refining the writing style, and adding risk markers before delivering the initial draft. The input consists of a batch of raw files, and the output is a memo deliverable with traceable citations. In the medium to long term, integrate deliverables into a self-updating knowledge base so that when clients in the same track place orders again, historical templates and conclusions are directly reused. Processing time per document decreases with accumulated cases, forming a compound asset rather than a one-time outsourcing task.

🛠 Setup Requirements

Requires basic financial analysis skills plus prompt engineering and retrieval setup capabilities. Tools can include GPT series or Claude linked with LlamaIndex-style RAG frameworks, cloud storage, and file parsing libraries. Getting templates and source alignment fully operational takes about 2 to 4 weeks. For long-term reuse, spend another week organizing historical deliverables into a structured knowledge base and industry-specific prompt templates. The overall investment requires no more than one month of spare time to start taking orders.

🧰 Toolchain

🔧 Claude or GPT series large language model APIs (for summary extraction and draft generation)
🔧 LlamaIndex-style RAG frameworks (for local file retrieval and source tracing)
🔧 Python document parsing libraries (pdfplumber, python-docx for parsing VDRs and financial statements)
🔧 Make or Zapier automation workflows (connecting file import, parsing, and notifications)
🔧 Notion or local knowledge bases (accumulating historical templates and industry prompts)

💰 Revenue

Priced per order at 5,000 to 30,000 RMB each. With proficiency, 5 to 10 orders per month yield a monthly income of roughly 20,000 to 80,000 RMB; this entry takes the median of 6 orders at approximately 36,000 RMB. Horizontally referencing Clarum's willingness to pay, its estimated annual fee per seat is 5,000 to 10,000 USD with an estimated 2025 ARR of around 330,000 USD, indicating that institutional clients are willing to pay continuously for such workflows, leaving room for upward pricing of individual contracting.

💸 Cost

LLM APIs and document parsing cost about 500 to 1,500 RMB per month, with platform commissions varying by channel. If purchasing finished products like Clarum, the annual per-seat fee is higher. The cost of a self-built system for individuals is almost entirely limited to APIs and subscription fees, and marginal costs continue to decrease as templates are reused.

⏱ Time Investment

2 to 4 hours per day, with a single delivery cycle of 3 to 5 days. During peak periods, two orders can be run in parallel with staggered delivery.

🚀 Getting Started

Step 1: Use public company financial reports and industry reports to simulate a due diligence memo sample, post on platforms like Xianyu, Xiaohongshu, or equity investment communities to accept low-priced trial orders, complete 3 orders to build your own template and prompt library, and then gradually raise prices. Step 2: Incorporate client file anonymization and confidentiality commitments into standard operating procedures, signing traceable citation and draft disclaimer terms before accepting orders.

🔑 Keys to Success

✅ Every conclusion must have traceable citations to the original file page number; this is the core differentiation from general AI ghostwriting.
✅ Accumulate an industry template library; orders in the same niche become faster to complete over time, creating compound value.
✅ Clients are primarily small and mid-sized PEs, FAs, and individual angels, acquired through referrals from Xiaohongshu and investment communities.
✅ Start with low-priced trial orders to accumulate real delivery cases, and use showcaseable anonymized samples instead of a resume to persuade the next batch of clients.

⚠️ 风险

⚠️ Errors in financial and due diligence conclusions may carry reputational or even joint liabilities; it must be clarified that deliverables are for draft services only and require human final review.
⚠️ Data confidentiality requirements are high; client files must be localized or anonymized to prevent leakage.
⚠️ LLM hallucinations can cause deviations in financial metrics and clause extraction; critical numbers must be cross-verified against original reports and contracts before finalizing.

📌 Real Cases

📌 Clarum is a YC W24 project whose official website states its AI agent can import data room files to automatically generate screening memos and investment committee memo drafts, with an estimated annual fee of 5,000 to 10,000 USD per seat and an estimated 2025 ARR of around 330,000 USD, confirming willingness to pay for this workflow.
📌 Community path recommendations in search results mention similar service capabilities quoted at 5,000 to 30,000 RMB per order on channels like Upwork and Xianyu, with an income range of 5 to 10 orders per month serving as a pricing anchor.
📌 Domestic counterpart evidence: Shizai Agent claims to complete enterprise due diligence in three minutes, covering automated report generation for business registration, judicial, and bidding public sentiment. Zhongke Baocheng launched an AI due diligence system covering statement analysis and capital demand calculation, both indicating that due diligence outsourcing demand has spilled over from private equity to credit and pan-financial enterprises, and individual contractors can benchmark against their capability checklists to package services.

Sources