58.com SMB Recruitment
An AI-Powered Reconstruction
Design Challenge for Workstream
The Critical Dilemma between
Traffic Monetization and Ecosystem Health
Upfront Business Research
Job Seekers
Limited capability to verify information
Rely on social referrals and instant feedback
Platform
Grew rapidly driven by demographic dividend and internet trend
Revenue highly rely on recruitment advertising
SMBs
Limited budget and time
Hard to reach authentic seekers
Labor Agencies
High salary ads and customized service
Funnel users into private channels
The platform achieved rapid early expansion by relying on demographic dividends and Internet trends; however, as these dividends faded, an over-reliance on advertising monetization mechanisms (like paid listings) amplified structural issues dominated by labor agency ultimately leading to a bilateral loss of SMBs and job seekers.
Empowering SMB Direct Hiring to Build a Sustainable Ecosystem
Design Vision
Selecting the Sector
Defining the Target Customer
美容娱乐 (勾)
Choose sectors not heavily penetrated by labor agencies to avoid direct confrontation
Manufacturing
Catering
Retail
Entertainment
SMBs with 10–50 Employees
They are in an awkward stage: busy operations and high turnover, but unable to afford full-time HR staff.
Management Gap
Store Manager doubles as HR. Has hiring decision power but no professional bandwidth
High-Frequency Need:
High staff turnover creates a constant deficit of 3–5 positions.
Willingness to Pay:
High acceptance of paid enterprise tools. They prioritize the hiring results.
Labor agencies and the platform are in a symbiotic relationship; while the labor agencies contribute the majority of revenue, the platform must avoid exhausting the long-term health of the ecosystem.
Labor Agencies
They contradict the design vision.
Key Accounts
(e.g., Haidilao)
Have professional HR teams and self-built ATS systems
Micro-Merchants
(Mom-and-Pop shops)
Their hiring needs are too low-frequency and casual.
The Hiring Dilemma in
Scale, Efficiency, and Quality
ICP’s Key Pain Points
Limited Outreach Scale
Manual screening restrict in a small candidate pool.
Low Execution Efficiency
Store managers act as HR
Hard to balance.
Low Profile Quality
Text resumes hard to reflect value
Garbage-in, garbage-out
Alternative Research Methods
Diving into ecosystem and Design Principals
Key Insights
User Insight
Online Experience
Hired and applied on the platform, Interview through phone and WeChat.
Offline Verification
Visits to Chinese restaurants with similar business models at 2–4 pm.
Expert Interview
In-depth interviews with experienced restaurant operators
Social-driven culture
Key individuals bring teams
Relationships affect retention
Mixed hiring criteria
Hard & soft criteria
Trust and cultural fit matter
High digital acceptance
Open to complex tools
Actively leverage tools
From Resume Transfer to Value Matching
Recruitment is about identifying "People," not screening "Text."
AI should act as an active connector, crossing single-medium dimensions to verify a applicant's real capabilities.
美容娱乐 (勾)
Resumes are a lossy compression of a applicant's ability, unable to carry the core values of deskless workers groups, such as physical strength, attitude, and stability. Relying on resume text for screening is essentially processing distorted information, leading to a "Garbage In, Garbage Out" loop.
The Medium Mismatch
Reconstruct The Goals
Using Voice AI instead of text resume
Surface Goal
Improve screening efficiency for merchants.
Fundamental Goal
Facilitate efficient value exchange between merchants and seekers.
Design Deduction
Maximize Data Source: Phone calls have no entry barrier, allowing broad access to deskless worker groups.
Intuitive Dialogue Reflects Reality: Real-time dialogue reduces space for thinking and polishing, capturing the real personality.
On-Demand Visual Supplement: For roles that need a certain appearance, ask users to upload a photo or video after the call.
美容娱乐 (勾)
Parallel strategies on both ends.
Driving high-quality information flow and continuous algorithm refinement.
AI digital fellow villager Ice-breaking
Convert unstructured spoken language into precise profiles.
Job Seeker Side
AI Recruiter with Targeted Delivery
Calibrate recommendation algorithms using feedback
Employer Side
Simple Posting
Based on similar employers and position presets, low barrier to publish jobs and AI interview settings
AI-conducted interview
AI automatic phone screening. Output structured interview reports
Human-Machine Calibration
View reports and correct bias. Manually reply to long-tail questions
Profile iteration
Based on interview feedback and replies. Refine preference models to improve recommendation accuracy.
Private Domain Retention
Guide users to add Customer Service Enterprise WeChat, reduce cross-platform communication loss.
Profile iteration
Continuously complete information based on dialogue and interview records, dynamically iterate the applicant user profile.
AI Dialect Ice-breaking
AI calls proactively in the local dialect. Generates resume and recommends jobs through natural dialogue.
Simple Entry
After registration, no need to manually fill out a resume.
Systematic Solution
An All-in-One Solution
with Enhanced Scale, Efficiency, and Quality.
Employer-side Solution
AI Interview
Smart Assistance
Holistic Profiling
Expand outreach scale
Improve hiring efficiency
Restore informational quality
AI Interview and Auto-Screening
Enlarge Outreach Scale
美容娱乐 (勾)
Multi-dimensional Standard Configuration
Dynamic Question Bank Generation
Auto Interview Invitation
Built-in presets reduce setup effort. Supports AI and manual edits via natural language.
Allows employer input of preferred answers; helps AI clarify scoring criteria to improve subsequent screening accuracy.
Reduces time lag from manual intervention, lowering the risk of applicant drop-off.
Automates the tedious work of information mining and initial screening.
AI Auto-Contact Applicants
Provide default options, but allow precise adjustments through natural language
Start with AI Interview Configuration.
Simple to begin. Precise to refine.
美容娱乐 (勾)
Auto Screening in Multiple Dimensions.
Effortless to get qualified applicants.
Hard Criteria Alignment
Q&A Analysis
Tone Analysis
Dialogue Analysis
Smart Assistance
Improve hiring efficiency
美容娱乐 (勾)
Auto-Reply Messaging.
Save Employer Time. Retain Applicants.
Before
After
Maintain Info Input: Dialogue is not just a way of communication, but a key source for iterating applicant profiles.
Solving Dual Pain Points: Employers don't need to reply to repetitive info at random times; applicants get timely feedback, reducing churn.
Fragmented and Repetitive Replies
Automatic Hosting & Centralized Processing
Aggregated Processing Mode
Rapid Voice Input
Self-Learning Knowledge Base
AI handles replies automatically, while long tail questions beyond the knowledge base are collected into a unified interface for centralized human handling.
Employers had to open individual chat windows one by one to handle a constant flow of similar questions arriving at unpredictable times, leading to inefficiency.
Respond directly within the unified view without switching between separate chat windows.
Combine voice-to-text input with short-video style swipe interactions to efficiently handle massive message volumes
Store key information from manual replies into the knowledge base to enable editing and management, and to optimize AI dialogue logic.
美容娱乐 (勾)
Auto-Scheduling
Restore Control for Store Managers
Use dialogue analysis to predict show-up probability.
Batch low and medium intent applicants for interviews automatically.
美容娱乐 (勾)
Clear Review Experience
for Massive Content
Workbench View - Overview
Quick overview of the hiring process through aggregated similar results.
List View - Retrieval
Quickly locate a specific applicant, displaying only key information.
Detail View - Review
In-depth view of a single applicant profile, presented structurally.
Card View - Scan
Rough browsing, exposing key information and summaries.
Under the list view, clicking the right-side area of an item opens a preview card, allowing users to briefly view an applicant’s profile without switching repeatedly between list and detail pages.
Provides quick preview to reduce page switching
Action button exposed
User has manually tapped to confirm intent
Low risk of accidental taps
Action button collapsed
Prevents accidental taps during scrolling
List view
Single item expansion
Card view
Design Details
Holistic Profiling
Restore informational quality
美容娱乐 (勾)
Structured Overview
Hard criteria labeled for quick recognition
Soft skill summarized from multi dimensions
Smart deduplication reduces duplicate application noise
Social Endorsement
Leverage social referral networks
Bring high-quality network onto the platform
Platform endorsement for high-quality talent
Interview Q&A snippets
Tone analysis reveals attitude and energy
Edit AI scores at the question level
Dialogue analysis
Uncover hidden signal
Predict offline interview show-up probability
美容娱乐 (勾)
Unified framework adapts to different entry scenarios
Seamless Switching Experience in Detail Page
Design Details
Detect whether the summary is already shown by entry context, then set the default tab on the detail page.
Continue scrolling at the end of the detail page to switch directly to the next profile.
美容娱乐 (勾)
A More Sustainable
SMB Hiring Ecosystem
Larger Scale
Higher Efficiency
Better Information Quality
Special Thanks
AI Assistance in Design
🙏
Business Research: Used Podwise (podcast-to-text) to collect dialogues from industry founders, integrated into Notebook LM to form a knowledge base, and connected to Gemini for deep brainstorming.
Competitor Analysis: Used ChatGPT-Atlas Agent to visit and analyze multiple HR SaaS tools’s website and summarize value propositions.
User Research: Use Lark to transcribe audio, then input into Notebook LM and Gemini to consolidate multiple conversations into clear, structured user insights.
Design Output: Leveraged Gemini and Figma Make to accelerate design deliverables.
View More Projects
© Qicheng Hsu. All rights reserved.