AI·LABA Product Manual

爱喇叭（AI·LABA）产品手册

AI·LABA Product Manual

1. Product Overview

AI·LABA is an AI-native, pay-for-results task marketplace: it organizes fragmented demand, talent, and service providers into a business network where tasks can be published, executed, accepted, and settled.

What We Are Not

We are not Explanation
Not a survey tool Online surveys are just one task family; the system identifies whether an entry belongs to tasks, recruitment, talent referral, projects, or RFPs
Not a job board Recruitment and referral ultimately connect to executable work; the Platform clearly distinguishes hiring relationships from result tasks
Not a service directory Projects must enter workflows, evidence, acceptance, and settlement; complex offline work is broken into executable batches by Operators using Kayle workflows

2. Four Types of Participants

Role Who Core value
Demand Side Enterprises / brands / institutions / individual merchants Find the right people and service providers faster, see delivery process and results, and lower sourcing, management, and transaction costs
Performer Individuals / professional talent / flexible workers Take on tasks by capability, region, and time; zero platform fee on personal tasks; build a verifiable performance record for repeat business and higher-value opportunities
Operator · Agent Service providers / recruiters / channel partners Respond to RFPs, organize projects, recruit personnel, perform QA, and deliver reports — with an enterprise-grade delivery control tower
Ecosystem Partner Payment / KYC / insurance / compliance / data services Enter the transaction chain through standard interfaces, provide regulated services, and lower the cost of scaled collaboration

One person can switch between roles. "Referrer" corresponds to the talent referral scenario (see 3.5 and 9.4).

Note: The "four types of participants" above are the Platform's user-facing role categories. "Ecosystem Partner" refers to third-party service providers (payment/KYC/insurance/compliance/data) integrated via standard interfaces — not a standalone account-role enum in code. The code account roles are Demand Side (publisher) / Performer (performer) / Operator (operator), plus the talent-referral relationship.

3. Five Task Families

# Family Description
01 Online tasks Surveys, annotation, information collection, and digitally deliverable work
02 On-site verification Stores, POIs, events, merchandising, and on-site evidence (evidence-driven)
03 On-site errands Store visits, pickup/delivery, accompaniment, errands, photo/measurement, and on-site data collection
04 Recruitment listings Event staff, part-time, project team, and position recruitment
05 Talent referral Refer yourself or others; attribution and incentives follow the rules (self-referral allowed)

Complex offline work is never forced into a single task: Operators decompose it into executable batches through versioned Kayle workflows.

4. Three Entry Points (Demand Side)

Publishers only need to answer one question: What do you want to get?

Entry Path Use case
I want results Tasks (online tasks · on-site verification · on-site errands) Turn vague needs into purchasable results
I want to recruit talent Talent (recruitment listings · talent referral) Discover positions and match people; hiring and employment typically happen with third parties
I want a service provider to take it on Projects (RFP · bidding · award · Kayle execution) Operators take on projects, decompose tasks, recruit and dispatch, QA, and accept

AI identifies intent, fills gaps, and recommends a path; pricing, publishing, awarding, and payment remain human-confirmed.
The three capabilities (discovery / execution / settlement) can be used individually or fully combined — the Platform does not force migration of entire processes.

Underlying publishing paths are consistent with the code: Tasks (task / daigou / find-people), Talent (find_people talent referral), Recruitment listings (info / syndication, structurally fund-free), Projects (project), and Trustmatch. The demand side is summarized into the three main entry points above.

5. Pay-for-Results: The 7-Step Loop

Define Results → Confirm Quote → Review & Escrow → Recruit & Dispatch → Execute & Collect Evidence → Accept Results → Settle & Rate
Step Content
1 Define results Deliverables · criteria · deadlines
2 Confirm quote Accepted by an individual or Operator
3 Review & escrow Publishing review · funds in place
4 Recruit & dispatch Organize people and resources
5 Execute & collect evidence Submit evidence by milestones
6 Accept results Remediation if not passed
7 Settle & rate Pay after acceptance · build credit

6. Product System (Six Modules)

  1. Unified publishing & marketplace: Tasks, recruitment, referrals, projects, and RFPs share one entry point and identity system.
  2. Trust & funds foundation: Review, escrow, acceptance, and ledger; connects compliant payment, tax, and invoicing providers.
  3. Kayle workflows: Version complex offline projects into roles, milestones, evidence, and task batches.
  4. Operator workbench: Bidding, teams, batch recruitment, QA queues, reporting, and reconciliation.
  5. Fulfillment coaching: Monitors risk, forms recovery plans, intervenes after human confirmation, and outputs reports.
  6. Multi-terminal & open capabilities: Mini-program, Web, and mobile; APIs and webhooks open progressively.

7. AI Capabilities (AI OS)

AI is not a chat box — it is a business operating system that runs through the entire chain. Distinctiveness: AI calls real business tools, generates typed business artifacts, and continuously learns from transaction outcomes.

The Five Layers of the AI OS

Layer Capability
Dialogue & intent Understands goals, asks follow-up questions dynamically; users don't need to learn the product first
Knowledge & rules Normalized Knowledge Packs, business contracts, and permission boundaries
Tools & matching Retrieval, recommendation, pricing assistance, and orchestration of tasks and providers
Artifacts & confirmation Generates executable drafts; key actions must be human-confirmed
Traceability & learning Provenance, audit, deterministic degradation, and cross-tenant isolation

Role-Based Copilots

Third-Party AI Collaboration (Controlled Channels)

In-platform agents (e.g., Xiao Laba / WorkBuddy), authorized APIs, and Feishu (Lark): only the minimal toolset is authorized; write operations require idempotency, audit, and human confirmation; no full-database copying to third-party AIs.

Compliance: AI-generated content is labeled and traceable per GB 45438; all AI output is for reference only, and key decisions are human-confirmed.

8. Business Model and Fees

Principles: the Platform does not handle funds; certification fees, membership subscriptions, software project packages, and transaction take rates are independent of each other.

Item Fee Notes
Personal task platform fee ¥0 No commission (code-confirmed task.take_pct=0.0)
Guaranteed/escrow transactions Platform does not handle funds; handled by licensed payment/split-settlement institutions See Disclaimer funds clause
Project take rate Standard 5% / Escort 8% (by service tier, price snapshot at award); ¥199 per-project floor; legacy projects awarded before v3 still run at the original 20% Subject to Platform publication
Project software package (per-use) ¥999/project Pay-per-use task execution OS (code-confirmed saas_payg_project_fen=99900)
Operator certification (annual) ¥1,000/year Annual (code-confirmed operator.verify_fee_fen=100000)
Membership subscription Plus ¥299/yr · Business ¥1,999/yr · Enterprise from ¥30,000/yr Annual (code-confirmed)
Performance bond Default ~10% (set per project; no global hardcoded default), min ¥2,000, cap ¥50,000 Per project (code-confirmed bounds bond.min_fen=200000 / bond.cap_fen=5000000)
Execution/QA coordination fee Tiered by task type (current seed-sample values: survey ¥2 / photo ¥8 / field ¥15 / labeling ¥1, code-confirmed) Subject to Platform publication

Pricing note: a line-by-line code review (decisions #52 / #60, 2026-08-03 migrations) confirms the current rates above — project take Standard 5% / Escort 8% with a ¥199/project floor, Operator SaaS ¥999/project plus three annual tiers, Operator certification ¥1,000/year, membership Plus ¥299/yr · Business ¥1,999/yr · Enterprise from ¥30,000/yr. Rates with no code basis (the previously cited 2% transaction-protection fee, 1% supply-chain-finance fee, individual ¥299 / company ¥999 certification, and ¥25k/¥100k/¥500k external-settlement annual packages) were removed as they do not exist in the codebase. For official communication, always follow the Platform's latest published rates.

9. Platform Rules

  1. Real-name & deferred verification: L0 sign-in → L1 anonymous profile at first task → L2 real-name/KYC only at withdrawal, lowering the barrier to entry.
  2. Review gate: Task/project publishing requires review; nothing goes live directly; key actions (pricing, publishing, awarding, payment) are human-confirmed.
  3. Escrow settlement: Guaranteed transactions use fully online escrow; settlement occurs after acceptance. Off-platform transactions are permitted but are not covered by Platform guarantee (see the Disclaimer).
  4. Single-level referral incentives: A reward is attributed to only one valid referral relationship and settles after the agreed outcome node; multi-level rebates and duplicate attribution are prohibited.
  5. Anti-abuse & conduct rules: Free publishing is rate-limited to prevent spam; fake orders, sham transactions, and fraud are prohibited; violators face feature restrictions or bans.

10. Compliance and Security

Item Description
PIPL compliance Zero PII in logs/events; data stored within China; on-site evidence de-identified via 100m distance bucketing; faces and license plates blurred before visibility
Anti-pyramid-scheme by structure Referral is strictly single-level and paid upon completion; the schema precludes "referrals of referrals" and prevents multi-level rebates
Licensed escrow · avoiding illegal "second clearing" The Platform does not handle funds; funds are completed by licensed payment/split-settlement institutions, avoiding "second clearing" (二清) compliance risk
AI content labeling Follows GB 45438 labeling requirements; AI-generated content is labelable and traceable end-to-end
Filings ICP Filing No. (Shanghai): 沪ICP备11008428号-1; Public Security Filing No.: 沪公网安备31010102007660号

11. Typical Scenarios (Illustrative; not a statement of clients or revenue)

Note: The Platform's current demo data is illustrative (e.g., a "skincare habit quick survey" task). The industry scenarios above illustrate product capabilities, not recorded client cases or revenue statements.

  1. Market research · offline surveys and data collection: multi-city street interviews, intercept interviews, store observations; import quotas/cities/POIs/time windows into executable batches, delivering structured research data sets.
  2. FMCG · store verification and mystery shopping: one brand task decomposed into hundreds of acceptable results covering merchandising, pricing, promotion execution, stockouts, service experience, and competitor observation, outputting store scores and gap lists.
  3. Brand intelligence: turning scattered opinions into verifiable decision inputs (consumer voice, channel supply, market facts); AI topic analysis and contradiction detection, with raw evidence always traceable.
  4. Batch recruitment (F&B / logistics): define cities, shifts, licenses, experience, and start times; the enterprise confirms hires and settlement follows agreed outcome nodes.
  5. Talent referral: refer yourself or others; the reward belongs to only one valid referral relationship and settles after the outcome node (single-level, prevents multi-level rebates).
  6. Supply chain finance: project delivery → client acceptance → receivables confirmation → financing application → early payment → maturity repayment, completed by licensed institutions such as banks/factoring companies.

12. Role-Based Usage Guide

Demand Side (Publisher)

  1. Register and log in (business users are advised to complete certification).
  2. Choose an entry: "I want results" / "I want to recruit talent" / "I want a service provider to take it on".
  3. Describe your goal; the AI copilot fills scope, budget, evidence, and acceptance criteria and drafts the task/recruitment/RFP.
  4. Confirm the content and quote, submit, and publish after passing the review gate.
  5. Receive quotes/bids, select the performer; complete fund escrow for guaranteed transactions.
  6. Track progress and review evidence; the system settles per agreement upon acceptance; rate each other to build credit.

Performer

  1. Register and log in; complete your capability, region, and time-preference profile.
  2. Browse recommended tasks in the Task Plaza (personalized feed) and read the rules.
  3. Use the Performer Copilot to generate execution checklists; submit photos, location, and data evidence by milestone (quality pre-check).
  4. Get paid after acceptance (zero platform fee on personal tasks); accumulate a verifiable performance record for higher-value opportunities.

Operator / Agent

  1. Complete Operator certification (¥1,000/year).
  2. Respond to RFPs, confirm scope, quotes, and scheduling.
  3. Configure Kayle workflows, decompose projects, and recruit and dispatch in batches.
  4. Perform evidence QA, risk monitoring, and anomaly recovery; deliver acceptance reports; complete reconciliation and repeat business.

Referrer

  1. Share job or project opportunities (obtain candidate authorization before referring others).
  2. The Platform establishes one unique valid referral relationship to prevent duplicate attribution.
  3. After the agreed outcome node (interview / onboarding / retention / project completion), receive a single-level reward per the rules.

13. FAQ

Q1: Does the Platform charge?
Personal task platform fee is ¥0; guaranteed/escrow transactions are handled by licensed institutions and the Platform does not handle funds (see the Disclaimer); certification fees, membership subscriptions, and project software packages are separate (see Section 8 and Platform publication).

Q2: Is my money safe?
The Platform does not handle funds. Funds are held by licensed payment/split-settlement institutions and are settled only after acceptance.

Q3: Is the Platform an employer?
No. The Platform is a task marketplace and technology service provider; it does not create employment relationships. The transaction object is a business result, not working hours.

Q4: What if a task is not completed?
It enters remediation/review/dispute processes per the acceptance criteria; no settlement occurs without acceptance (subject to Platform rules).

Q5: Can I transact off-platform?
Yes, but off-platform transactions do not enjoy Platform escrow settlement or transaction protection.

Q6: Will AI make decisions for me?
No. AI only assists and drafts; key actions such as pricing, publishing, awarding, payment, and acceptance are human-confirmed.

Q7: How do I withdraw funds and get invoices?
Withdrawals, split settlements, invoicing, and taxes are completed by licensed institutions and the user, with the Platform providing process assistance; large enterprise settlements can be invoiced through the Platform (subject to finance and compliance rules).

Q8: Do referrals earn rebates?
Talent referral is single-level and paid upon completion; the Platform prohibits multi-level rebates.

14. Contact and Support

爱喇叭 AI·LABA · www.laba.com.cn