58 Tongcheng SMB Recruitment
An AI-Powered Reconstruction
Design Challenge for Workstream
The Ecosystem Dilemma
Balancing Monetization with Health
The platform initially expanded rapidly by leveraging the "demographic dividend" and massive information flow. However, as this dividend diminishes, the platform’s over-reliance on ad-revenue mechanisms (like paid listings) has exacerbated structural issues dominated by intermediaries. This has led to a dual churn of both SMBs (Small and Medium Businesses) and genuine job seekers.
The Agency
Acquire traffic through inflated salary listings
Provide high-touch offline support
Siphon platform users into private channels
SMB
Limited budget with no dedicated HR staff
Disadvantaged in ad bidding wars
Struggle to reach authentic candidates
Job Seekers
Difficulty verifying information validity
Rely on peer referrals and instant feedback
Easily swayed by agencies' high-touch services
The Trade-off: Traffic vs. Ecosystem
While labor agencies currently drive the majority of revenue, their dominance threatens the long-term health of the ecosystem. The platform must pivot to prevent further depletion of its user base.
Design Vision:
Empowering SMB Direct Hiring
for a Sustainable Ecosystem
Defining the Ideal Customer Profile (ICP) through Strategic Filtering
Selecting the vertical
Defining the Target Customer
美容娱乐 (勾)
Focus on sectors not yet fully saturated by agencies to avoid direct confrontation.
Manufacturing
Catering
Retail
Beauty
Entertainment
The Sweet Spot
SMBs with 10–50 Employees
This segment is in an awkward growth phase: busy operations and high turnover, yet unable to afford dedicated HR staff.
Management Gap
The Store Manager doubles as HR. They have the authority to hire but lack the energy to manage the process.
High-Frequency Need
High staff turnover creates a constant deficit of 3–5 positions.
Willingness to Pay
High acceptance of paid enterprise tools if they deliver efficiency. They prioritize the hiring results.
Labor Agencies
They contradict the design vision.
Key Accounts
(e.g., Haidilao)
They already possess professional HR teams and proprietary ATS (Applicant Tracking Systems).
Micro-Merchants
(Mom-and-Pop shops)
Their hiring needs are too low-frequency and casual.
Deep Dive Into Pain Points.
The "Supply-Demand Mismatch" Behind High Visibility
The traditional "bidding for top placement" model generates high-volume but low-quality exposure. Employers are exhausted by filtering invalid leads, while job seekers are silenced by "resume barriers."
美容娱乐 (勾)
The Job Seeker Experience:
The Feedback Black Hole
The Merchant Experience:
Paying for Noise
Vanity Metrics
Paid promotions bring high volumes of "irrelevant" or "mass-applied" leads, resulting in high exposure but no qualified hires.
Shifted Screening Costs
The platform focuses only on "casting a wide net," forcing busy store managers to spend excessive time on phone screening and verification.
The Messaging Dilemma
Faced with a flood of repetitive inquiries, managers lose leads if they don't respond, but lose business hours if they do.
Skill-Expression Mismatch
Many blue-collar workers have practical skills but cannot express them through "text resumes," leading them to be overlooked by systems and employers.
The "Feedback Black Hole"
Due to a lack of timely responses from busy merchants, job seekers resort to "revenge mass-applying," creating a vicious cycle.
Shift to Personal Networks
Struggling to distinguish real from fake ads on platforms, newcomers rely more on the credibility of "peer referrals."
Research Methodology: Constraints & Workarounds
Online Deep Dive:
Operated dual accounts (employer/seeker) on the platform to simulate the loop and conducted phone/WeChat interviews.
Offline Validation:
Visited overseas Chinese restaurants with similar operational models during off-peak hours (2–4 PM) to validate staffing pain points.
Unable to conduct physical site visits with 58.com merchants while overseas.
Design Strategy
From Resume Transfer to Value Matching
Recruitment is about identifying people, not screening text.
AI must act as an active connector, transcending text media to verify real capabilities.
美容娱乐 (勾)
Resumes are a "lossy compression" of a candidate, failing to capture blue-collar core values like physical stamina, attitude, and stability. Relying on text screening creates a "Garbage In, Garbage Out" loop.
The Medium Mismatch
Reconstruct The Goals
The Solution
De-Resume-ification
Stop forcing deskless workers to perfect text resumes. Instead, use natural language interaction to build user profiles.
Seeker Side:
Shift from "Man finds Job"
to "Job finds Man.
AI acts as a "Digital Agent," reaching out via WeChat to guide the conversation.
Merchant Side:
Shift from "Manual Hunting" to "AI Full-Service."
AI handles the first round of phone screening and repetitive replies.
Surface Goal:
Improve screening efficiency for merchants.
Fundamental Goal:
Facilitate efficient value exchange between merchants and seekers.
Proactive Service
User Insights
Decoding the Recruitment Ecosystem
In-depth interviews with two catering veterans with 10+ years experience to establish design standards.
Key Insights
Design Tactics
美容娱乐 (勾)
Referrals as the Mainstream
Established businesses rely on internal referrals, while new shops depend on platforms. Candidates hired through "peer referrals" undergo social screening and are generally more stable than direct platform hires.
Network Effects:
"Head Chef brings the team" is standard. Blue-collar retention is heavily tied to social networks within the team.
High Frequency for Low Barriers:
Low-barrier roles such as dishwashers and servers experience frequent turnover, while core technical positions like head chefs remain relatively stable.
Hybrid Criteria:
Hard constraints (Age/Certifications) vs. Soft constraints (Communication/Chemistry/Team fit).
High Digital Literacy:
Managers are educated enough to master complex backends (Meituan/Ele.me).
Use Enterprise WeChat AI to simulate a "Digital Fellow Villager."
Implement a "Gold Recruiter" mechanism to leverage industry connections.
Convert human intuition into algorithmic models for low-barrier roles.
Transforming human intuition into algorithmic models
A professional yet lightweight tool migrating C-side interaction habits to B-side.
美容娱乐 (勾)
AI "Digital Peer" Icebreaking Transforming unstructured colloquial speech into precise candidate profiles
Job Seeker Side
AI Recruiter Precision Delivery Calibrating recommendation algorithms through feedback loops
Merchant Side
Rapid job posting
Based on predefined merchant types and role templates to lower the barrier to posting roles and configuring AI interviews
AI-led interviews
AI conducts automated phone screening and outputs structured interview reports
Human-in-the-loop calibration
Review reports, refine evaluations, and manually follow up on long-tail questions
Profile iteration
Incorporate interview feedback and responses to refine preference models and improve recommendation accuracy
Private-channel conversion
Guide users to add enterprise WeChat support. Reduce conversational loss across platforms
Profile iteration
Continuously enrich information from conversations and interview records.Dynamically refine the job seeker profile model
AI Ice-Breaking:
Dialect-aware AI proactively places calls. Natural conversation generates a resume and recommends roles
Zero-Effort Onboarding:
No manual resume entry required after registration.
Why prioritize phone-based voice calls as the primary data entry channel?
Accessibility: Phone calls ensure maximum coverage of the blue-collar demographic.
Authenticity: Real-time dialogue reduces the ability to "polish" or fake information.
Visual Supplement: For image-critical roles, visual data (photos/videos) is requested after the initial voice screen.
美容娱乐 (勾)
Three Core Merchant Features
for Efficient Value Matching
AI Surrogate Interview
Automatically contacts candidates, cleanses resume data, and schedules offline interviews for those who pass.
Intelligent Communication Assistant
AI manages initial communication to mine candidate traits and prevent churn caused by slow replies.
Social Referral Mechanism
Digitizes the "referral" model, assigning high algorithmic weight to social connections ("Renqing" filtering).
美容娱乐 (勾)
Configuration
Low-Threshold Setup, Quantifying Intuition
User inputs job details, system generates JD, then proceeds to AI interview setup.
Multidimensional Standards
Dynamic Question Generation
Automated O2O Invitation
System presets reduce effort; supports natural language commands for AI modifications.
Merchants can input preferred answers to help the AI clarify scoring criteria, improving screening accuracy.
Automatically invites qualified candidates to the store, minimizing time gaps and drop-off.
AI proactively contacts applicants to handle data mining and screening.
Automated Outreach
美容娱乐 (勾)
A clear workspace
Designed for a coherent review experience
Before
After
Screening resumes, reviewing interview results, and replying to messages are handled across multiple pages and modules, requiring users to manually switch contexts to complete tasks.
Fragmented work modules
An intelligently consolidated workspace
Automated tasks are handled by default;
Manual tasks and review items are aggregated by priority, enabling efficient overview and review;
AI interviewer continuously screening and qualifying candidates on your behalf.
Subtle motion indicators show that AI is continuously running, reinforcing user awareness.
Interview results of the same type are aggregated for display, enabling faster and more efficient review.
美容娱乐 (勾)
Information Hierarchy
Adaptive Views for Different Contexts
Workbench View - Overview
Tracks recruitment progress;
aggregates similar results.
List View - Retrieval
For quick location of specific candidates;
shows key info only.
Detail View - Review
Deep dive into a single candidate with structured data.
Card View - Scan
For rapid browsing;
highlights key information and summaries.
Context-Driven Interaction Details
Action buttons on cards are collapsed by default to prevent accidental clicks while scrolling.
Folded Actions
Expanded cards reveal action buttons, ensuring clicks are intentional.
Intentional Interaction
Expand on Click
Clicking a list item expands the card in place, allowing rapid scanning without navigating back and forth between list and detail pages.
美容娱乐 (勾)
Interaction Patterns
Leveraging Short-Video Mental Models
Vertical Scroll: Switch between candidates, mimicking TikTok/Reels to lower the learning curve.
Horizontal Swipe: Switch between different data dimensions for a single candidate to avoid information overload on one screen.
High reading and cognitive load
Difficult to quickly identify key signals
Reduces single-screen density Allows users to focus on key content in stages
Paginated Display
Full-page Display
美容娱乐 (勾)
Integrate multi-channel data to help merchants form a comprehensive view of the candidate.
Social Endorsements
Ecosystem alignment: Align with the offline hiring norm of “referrals through trusted connections,” leveraging social ties to reduce decision-making cost.
Network accumulation: Grant recommenders honorary titles such as “Gold Talent Scout” to attract high-quality candidates while building long-term professional credibility for future job seeking.
Algorithm optimization: Treat relationship networks as high-weight signals to significantly improve recommendation accuracy.
Interview Audio Slices
Controlled playback: Default to muted audio with optional user activation, creating a non-intrusive playback experience.
Key extraction: Surface highlight segments from long recordings to reduce the effort required to access critical information.
Structured Overview
Visual priority: Address the widely observed “face-first” preference in user research by presenting large images to clearly convey the candidate’s real appearance.
Efficient screening: Use red and green labels to enable rapid evaluation of hard requirements.
Intelligent summary: AI interview assessments support the evaluation of soft skills.
Dialogue Analysis
Intent discovery: Analyze historical conversation data to accurately surface a job seeker’s underlying interests and estimate interview conversion probability.
Decision support: Incorporate “initiative in asking questions” as an evaluation signal, providing a more authentic reflection of core motivation than resume text alone.
Multidimensional Profiles
Reconstructing a Holistic Persona
美容娱乐 (勾)
Efficient Smart Replies
& Human-AI Collaboration
Before
After
Dialogue as Data: Conversation is not just for contact; it is an input source for refining the candidate's profile.
Pain Point Solved: Merchants avoid repetitive replies; candidates get instant feedback.
Scattered, repetitive manual replies to homogeneous questions.
Automated hosting and centralized processing.
Unified Interface
Rapid Voice Input
Self-Learning Knowledge Base
AI handles standard queries.
Unique questions are aggregated into a single view for human handling.
When a large volume of homogeneous inquiries arrives unpredictably, users must open and review each conversation thread individually to respond, resulting in low efficiency.
Reply directly from the aggregation page without page jumping.
Voice-to-text combined with short-video style interaction for speed.
Manual replies are saved to the library to train the AI's future logic.
美容娱乐 (勾)
Acknowledgments
AI-Empowered Design Process
🙏
Industry Research: Used Podwise to transcribe founder podcasts, integrated into NotebookLM to form a knowledge base, and used Gemini for brainstorming.
Competitive Analysis: Used ChatGPT-Atlas Agent to analyze HR SaaS tools and summarize value propositions.
User Research: Used Lark for transcription and NotebookLM/Gemini to synthesize interview insights.
Design Output: Leveraged Gemini and Figma Make to accelerate design deliverables.
View More Projects
© Qicheng Hsu. All rights reserved.