← Sticker Wall AGENT · DETAIL

AI Receptionist Outsourcing: A Performance-Based Model for B2B Customer Service Generating 30,000 RMB Monthly

Workflow: Daily operations are divided into three phases: Before 8 AM, the Agent automatically retrieves inbound calls, WeChat Off

AGENT

Key Fields

FIELD STAMPS

IndustryMarketing / Advertising

RegionChina

ScaleSME

ChannelOnline

🔧 Workflow

Daily operations are divided into three phases: Before 8 AM, the Agent automatically retrieves inbound calls, WeChat Official Account messages, and email form inquiries from the previous night, performing intent recognition (inquiry/quote/appointment/complaint) and triaging them into A/B/C categories based on an industry knowledge base. Category A (high intent) leads are pushed in real-time to the client's sales team via Enterprise WeChat with a phone alert; Category B/C leads are handled by automated knowledge base replies or scheduled callbacks. After 10 PM, the operator spends 1-1.5 hours reviewing approximately 5%-10% of transcribed audio to correct Agent triage errors, supplement industry scripts, and optimize SOPs, creating a closed-loop optimization process. The input consists of multi-channel customer inquiries, and the output consists of structured CRM leads and automated response logs, with conversion rate and connection rate reports delivered to clients monthly.

🛠 Setup Requirements

Requires one solo operator with basic knowledge of prompt engineering and integrated workflows. It is recommended to spend 2-3 weeks on MVP development. Week 1: Complete 400-number/SIP line application, voice Agent reception flow, and basic script library configuration. Week 2: Integrate n8n/Coze to automatically write triage results into the client's CRM and trigger Enterprise WeChat notifications. Week 3: Finalize the automated daily/monthly reporting process. The technical barrier is moderate; the core lies not in model fine-tuning, but in accumulating vertical industry SOPs and handling edge cases. Subsequent client lines can be rapidly replicated using these methods.

🧰 Toolchain

🔧 Vapi
🔧 Twilio
🔧 n8n
🔧 DeepSeek API
🔧 Feishu Open Platform
🔧 Yunyi Technology Call Center API

💰 Revenue

① Performance-based revenue from foreign trade and local enterprise customer service lines (primary income): Clients pay a monthly service fee per line, 5,000-15,000 RMB/month × 2-3 lines = 10,000-45,000 RMB/month. Based on the card's estimate of 30,000 RMB/month, this accounts for ~67% of monthly income (calculated as 10,000 RMB median × 2 lines = 20,000 RMB ÷ 30,000 RMB; source: case study material, no independent verification); ② Lead conversion commission: Clients pay a commission on closed leads; commission rates and actual lead volume are not disclosed, making the proportion of total income unclear; ③ Startup income for replicators: Initial 1-2 months consist mainly of small basic service fees, ~3,000-5,000 RMB/month (media estimate, unverified), with its share of total income unspecified; ④ Opportunity item—Overseas client project-based work: One-time project fee $3,000-$5,000 + monthly maintenance $500 (media estimate, unverified). Revenue potential from project fees plus monthly subscriptions remains unproven.

💸 Cost

Monthly cost is approximately 1,500-3,000 RMB, including voice Agent platform subscription (~300 RMB) + SIP channel minutes (~500-1,000 RMB) + DeepSeek/Tongyi API usage fees (~500-1,500 RMB).

⏱ Time Investment

2-3 hours per day (mostly for nighttime transcription review and script optimization), with the AI receptionist running automatically 24/7 the rest of the time.

🚀 Getting Started

Step 1: Target 1-2 small B2B companies with stable inbound traffic but insufficient customer service staff. Focus on high-repetition scenarios like foreign trade inquiries, home renovation consultations, and medical aesthetic appointments. Use a 400-number demo to propose a performance-based model with a 14-day free trial. Run for 2 weeks for free to accumulate triage accuracy and lead conversion data, creating a comparison report. Then, use the verified data of '70% lower labor costs + 30% higher lead volume' to negotiate a formal contract, charging a 5,000-8,000 RMB base fee plus lead commission in the first month. Once a single line consistently generates over 8,000 RMB/month, immediately replicate the workflow to a second client in the same industry, gradually upgrading from single-line outsourcing to 'full-store reception management' with monthly fees of 6,000-15,000 RMB.

🔑 Keys to Success

✅ Contracts must be performance-based (based on appointments/lead conversion rate or effective talk time) rather than fixed subscriptions, allowing clients to enter with zero threshold and binding long-term interests.
✅ Start by selecting vertical industries with high inbound frequency, centralized scripts, and high fault tolerance (home renovation, foreign trade inquiries, local medical aesthetics, vocational training) to train a reusable industry knowledge base template.
✅ Human arbiters must intervene daily to review 5%-10% of edge cases to prevent model drift and lead loss, while simultaneously accumulating industry SOPs into tangible assets.
✅ Use open-source communities like GitHub to store reusable Agent workflows and script templates, reducing costs for subsequent client customization and building technical asset compounding.

⚠️ 风险

⚠️ Early errors in speech recognition regarding dialects and technical terminology can lead to incorrect triage, directly impacting commission income.
⚠️ As clients grow, they may prefer to build their own Agents, risking non-renewal of outsourcing contracts. Continuous optimization of scripts and data reporting is required to retain clients.
⚠️ B2B clients are sensitive to call compliance and data privacy. Transcription and storage must comply with Personal Information Protection Laws; deploying a private knowledge base is recommended.
⚠️ 400-number lines and SIP channels are subject to policy fluctuations; maintain backup lines from multiple carriers.

📌 Real Cases

📌 AI Q&A Station '2026 AI Customer Service Outsourcing Record': A complete breakdown of how an ordinary person can earn 30,000 RMB/month through performance-based solo operation.
📌 36Kr '24-Hour Online Agents are Delivering Orders to Foreign Trade Enterprises in Batches': B-side AI automated lead acquisition and conversion Agents have achieved 24-hour automated order taking and customer grading.
📌 GitHub Open Source Project Eric_Frank: A full-link AI automated lead acquisition and conversion Agent system for B2B foreign trade, verified for scalable deployment in SMEs, reducing manual outbound costs by over 60%.
📌 XTCer 'The $250,000/Month AI Business': The most profitable AI projects in 2026 are not underlying technologies but practical enterprise-level services, with receptionist outsourcing being a high-margin, low-code delivery category.

Sources