本地生活服务 · 一双手的生意 × AI · 2026-07Local Services · The Hands Economy × AI · Jul 2026

AI 接管了找人、排班和收钱,却始终接不过那双手——于是替代顺序被倒了过来:Tony 老师,比他的老板更安全 AI has taken over finding, scheduling and collecting, and still cannot take over the hands — so the order of replacement flipped: the hairdresser is safer than his boss

渗透是一条 U 型曲线:两头深,中间空——获客种草 ~80%、匹配调度 ~85%、评价支付 ~80%、复购金融 ~70%;而手艺交付只有 ~10–15%。AI 吃掉的是店长、前台、调度员、办卡顾问;交付层唯一的真实存量替代,是一台美甲打印机。Penetration runs a U-curve: deep at both ends, hollow in the middle — seeding ~80%, matching and dispatch ~85%, review and payment ~80%, repeat and card finance ~70%; while the craft itself sits at ~10–15%. AI ate the manager, the front desk, the dispatcher and the card seller; the delivery layer’s one real replacement is a nail-printing machine.
位移的物理账:到店美甲师 8–10 单/天,上门只有 3–4 单/天(B);一笔「60 分钟」上门单真实占用 135 分钟(去程30+服务60+收拾15+回程30)。去掉门店没降成本——只是把可见的房租,换成了不可见的位移时间与利用率坍塌。The physics of the house call: an in-store manicurist serves 8–10 orders a day, a mobile one 3–4 (B); a «60-minute» visit truly consumes 135 minutes (30 travel + 60 service + 15 pack-up + 30 return). Removing the store cut no cost — it swapped visible rent for invisible transit time and collapsed utilisation.

主脊是一张服务工单的八段生命周期(需求→匹配→定价→调度→位移→交付→评价→复购办卡),渗透深度呈 U 型;第二版图是位移基因四型(客体不可移动/中央工厂/重设备到店/伪上门)——品类的命运由它决定。另有河狸家验尸报告、办卡金融与 2025 预付司法解释、产品指南(2C/2B)。姊妹分工:住所资产→home,一餐饭→dine,变美产业→beauty,零工宏观→work The spine: one service ticket’s eight-stage life (demand → matching → pricing → dispatch → transit → delivery → review → repeat & cards), its penetration a U-curve; the second map is the four mobility genes (immovable object / central factory / heavy-equipment in-store / pseudo-mobile) — a category’s fate is decided there. Plus the Helijia autopsy, card finance under the 2025 prepayment ruling, and a product guide (2C/2B). Siblings: housing assets → home, the meal → dine, the beauty industry → beauty, gig macro → work.

传统节点Traditional
AI 已接管Taken by AI
平台 · 调度与税Platform · dispatch & tax
那双手 · 硬骨头The hands · hard bones
办卡金融 · 风险Card finance · risk
3–4 vs 8–10单/天
上门美甲师 vs 到店美甲师的日均服务量(多源一致,B)——一个人一次只能到一个家:位移的物理刚性是利用率天花板,算法能省绕路,省不掉位移本身Daily orders: mobile vs in-store manicurists (multi-source, B) — one person reaches one home at a time: transit’s physical rigidity caps utilisation; algorithms trim detours, never the transit itself
4–10%
抖音 2024-07 上调后的丽人类软件服务费(原 3.5%,平台公示 B)。内容平台收税正常化——商家从此被美团+抖音双重收税;⚠️名义佣金与「实际流量成本 20–40%」(诉苦口径 C)是两本账Douyin’s beauty-category fee after the Jul 2024 raise (from 3.5%; platform-published, B). Content platforms normalise their tax — merchants now pay two landlords; ⚠️nominal commission vs the «real traffic cost of 20–40%» (grievance basis, C) are two ledgers
+40%
美团 AI 数字员工带来的门店夜间留资率提升(5 万家教培+丽人门店,5 秒应答;平台口径 C)——AI 接管的第一批完整岗位:客服、前台、排班员The lift in overnight lead capture from Meituan’s AI staff (50k tutoring and beauty shops, 5-second replies; platform basis, C) — the first full jobs AI took: service, front desk, rostering
150
天鹅到家注册认证劳动者(招股资料,B;注册用户 1600 万+)——家政平台的真壁垒不是 App,是招募/背调/培训/替换/保险的劳动者网络Swan Daojia’s certified workers (prospectus, B; 16M+ registered users) — the home-services moat is not the app but the worker network: recruiting, vetting, training, replacement, insurance
口径警告:本页综合四份深度研究交叉整理——四源已于 2026-07 重跑修订(旧版曾有截断、报错与引文缺失;修订版全部完整收束、带可溯源引文),修订版结论与旧版收敛,旧版孤例数字维持原分级。渗透深度%为编辑折算的展示值,非硬数据。⚠️两组「同一件事的两本账」必须带定语引用:到店佣金名义 8–12%(C)vs 商家实际流量成本 20–40%(诉苦口径 C);外卖技术服务费 6–8%(拆分口径 B)vs 综合扣点 15–22%(含配送 C)。河狸家「估值 50 亿」为宣称口径(C);「HACR 1:72」单源且报告方不明(C/D);「美团 2608 亿核心本地商业收入」口径未核,本页不用;「穿戴甲破百亿」高度存疑(D)。本图盲区:海外对照(Thumbtack/Angi/Urban Company 等为编辑补充,D)、宠物与搬家品类(文档零覆盖,按位移基因推演)。 Basis warning: cross-compiled from four deep-research reports — all four re-run and revised in Jul 2026 (the old set carried truncations, an errored run and missing citations; the revision closes complete with traceable links); the revision converges with the old conclusions, and old orphan figures keep their grades. Penetration percentages are editorial display values, not hard data. ⚠️Two «two-ledger» pairs must carry qualifiers: in-store commission nominal 8–12% (C) vs the merchant’s real traffic cost of 20–40% (grievance basis, C); delivery’s tech-service fee 6–8% (split basis, B) vs the all-in 15–22% (incl. courier, C). Helijia’s «¥5B valuation» is a self-claim (C); «HACR 1:72» is single-sourced with an unclear publisher (C/D); Meituan’s «¥260.8B core local commerce» is unverified and unused here; «wearable nails past ¥10B» highly doubtful (D). Blind spots: overseas comparators (Thumbtack / Angi / Urban Company are editorial additions, D) and the pet/moving categories (zero coverage; reasoned from the mobility genes).
中心悖论 · AI 什么时候能给你理发?The central paradox · when can AI cut your hair?
理发是图灵测试的反面The haircut is the Turing test, inverted
图灵测试问「机器能不能像人一样说话」;理发问「机器能不能像人一样动手」——它要求灵巧手+审美语义对齐+情绪价值+入户信任四项同时达标,是具身智能的试金石。四份文档同判:20 年内难脱离科幻。The Turing test asks whether a machine can talk like a person; the haircut asks whether it can work with its hands like one — demanding dexterity + aesthetic alignment + emotional value + at-home trust, all four at once: embodied AI’s touchstone. All four sources concur: twenty years from science fiction.
四项同时达标,缺一即败Four bars, all at once
灵巧手 × 审美 × 情绪 × 信任Dexterity × taste × feeling × trust
灵巧手:头骨凹凸、发质弹性、力反馈闭环,头发不可撤销、容错为零;审美:「稍微修一下」千人千义,没有 ground truth;情绪:Tony=半个心理咨询师,「安全距离内的陌生人社交」;信任:让机器拿刀靠近头皮——恐怖谷+安全焦虑,突破需以代际为单位。Dexterity: skull contours, hair elasticity, force-feedback loops — hair is irreversible, tolerance zero; taste: «just a trim» means a thousand things, with no ground truth; feeling: the barber is half a counsellor — stranger-sociality at a safe distance; trust: letting a machine bring a blade to your scalp — uncanny valley plus safety dread, a generation-scale barrier.
于是,替代顺序倒置Hence, the flipped order
Tony 老师比他的老板更安全Tony is safer than his boss
白领世界「体力在底、脑力在顶」的替代顺序,在这里被倒了过来:AI 最先吃掉的是店长、前台、运营、客服、投手、办卡顾问——「食利与流程」的那一层;交付层的手比管理层的嘴更稳。一人公司随之成立:AI=给手艺人配一支虚拟运营团队(HACR 1:72,⚠️单源 C/D)。选人时代的硬数据:74.4% 消费者更相信发型师而非理发店、71% 愿为心仪发型师多通勤 52.5 分钟(平台报告 B/C);现象级案例:发型师晓华 7 天走红、吸引超 20 万游客,「山城小栗旬」227 万粉带动门店客流翻倍。诚实修正:头部案例仍是「个人 IP 反哺门店」,完全脱店的一人公司还是早期萌芽,不宜夸大为既成现象。The white-collar order — muscle at the bottom, brain at the top — flips here: AI first ate the manager, the front desk, operations, service, ad-bidding and the card seller — the rentier-and-process layer; the delivering hand outlasts the managing mouth. The one-person firm follows: AI as the artisan’s virtual back office (HACR 1:72, ⚠️single source, C/D). The «choose-the-person» era in hard numbers: 74.4% of consumers trust the stylist over the salon, and 71% will commute 52.5 minutes longer for the one they want (platform report, B/C); the breakout cases: stylist Xiaohua drawing 200k+ visitors within a week of going viral, and a 2.27M-follower barber doubling his shop’s footfall. The honest correction: the head cases remain «personal IP feeding the shop» — the fully shop-free solo firm is still a seedling, not a fact to inflate.
渗透 U 型曲线 · 两头深,中间空The penetration U-curve · deep ends, hollow middle
需求 80%Demand 80%
匹配 85%Match 85%
定价 60%Price 60%
调度 75%Dispatch 75%
位移 40%Transit 40%
交付 ~13%Craft ~13%
评价 80%Review 80%
复购 70%Repeat 70%
读法:AI 接管了工单的两头(找人/排班/收钱/召回),中间的位移与那双手纹丝不动——第 6 段是全行业的护城河,也是「平台掌握第一次、手艺人掌握第 N 次」的地基。%为编辑折算展示值。Reading: AI took both ends of the ticket (finding, rostering, collecting, recalling); the middle — transit and the hands — never moved. Stage 6 is the industry’s moat, and the ground of «platforms own the first order, artisans own the Nth». Percentages are editorial display values.
Reading the MapReading the Map

从这张图带走的五条规律Five patterns to take away

立场声明:本页是批判性、祛魅的行业结构分析,用 A–D 角标区分司法解释/平台公示/招股资料与厂商口径、待核数字,「同一件事的两本账」处标 ⚠️ 并存。产品指南是市场地图,不构成经营、投资或消费建议;办卡与预付内容不构成法律意见。核心判断一句话:AI 接管了找人、排班和收钱,却始终接不过那双手——替代顺序在这个行业被倒了过来。 Stance: a critical, demystifying structural analysis; A–D badges separate judicial rules, platform-published rates and prospectuses from vendor claims and unverified figures, with ⚠️ where one fact keeps two ledgers. The product guide is a market map — not business, investment or consumption advice; the prepayment content is not legal advice. The core judgment in one line: AI took over finding, scheduling and collecting — and still cannot take over the hands; here, the order of replacement flipped.