AI Wealth Advice:投顾边界与最佳利益架构
以下官方来源作为学习锚点。本文只把公开材料转成产品、流程、架构和证据设计语言, 不得被解读为法律意见、合规结论或适用性判断。正式项目的准确适用范围由 Legal / Compliance / Risk / Supervision 结合机构注册类型、产品类型、渠道、司法辖区和内部政策确认。
AI Wealth Advice / Robo-Advisor / Best Interest Boundary Architecture 解读
面向对象: AI Product Architect / Wealth PM / Robo-Advisor Product Lead / Senior BA / Solution Architect / Model Risk Partner / Conduct Risk Owner。 核心问题: AI 财富建议系统的难点不是“能不能讲清投资概念”, 而是能否把 education、recommendation/advice、portfolio construction、execution、human escalation、disclosure、evidence 和 supervision 做成可控边界。 学习目标: 建立 AI wealth advice boundary architecture, 把客户画像、风险承受能力、approved investment universe、组合推荐、best-interest style controls、幻觉控制、投诉监督和模型风险连接成一套可审计产品架构。
Source Anchors
以下官方来源作为学习锚点。本文只把公开材料转成产品、流程、架构和证据设计语言, 不得被解读为法律意见、合规结论或适用性判断。正式项目的准确适用范围由 Legal / Compliance / Risk / Supervision 结合机构注册类型、产品类型、渠道、司法辖区和内部政策确认。
| Source | Link | 本文使用方式 |
|---|---|---|
| SEC Regulation Best Interest | https://www.sec.gov/regulation-best-interest | 作为 broker-dealer 客户推荐场景的 best-interest 风格控制锚点, 用于抽象 disclosure、care、conflict、compliance/supervision 语言 |
| SEC Robo-Advisers Investor Bulletin | https://www.sec.gov/oiea/investor-alerts-bulletins/ib_robo-advisers | 作为 robo-adviser 客户体验、算法提问、费用、限制、人工参与和客户理解风险的投资者教育锚点 |
| SEC Investment Adviser Fiduciary Interpretation | https://www.sec.gov/rules-regulations/2019/06/ia-5248 | 作为 investment adviser fiduciary duty 学习锚点, 用于提醒 AI advice 系统必须区分业务角色和法律身份 |
| FINRA AI Key Topic | https://www.finra.org/rules-guidance/key-topics/artificial-intelligence-ai | 作为成员机构使用 AI 时治理、监督、客户沟通、模型与数据风险的官方主题锚点 |
| Investor.gov Automated Investment Tools | https://www.investor.gov/introduction-investing/getting-started/working-investment-professional/automated-investment | 作为客户理解 automated investment tool、问题清单、费用、限制和人工服务边界的教育锚点 |
| NIST AI RMF | https://www.nist.gov/itl/ai-risk-management-framework | 用 Govern / Map / Measure / Manage 组织 AI wealth 系统的风险管理闭环 |
| ISO/IEC 42001 | https://www.iso.org/standard/42001 | 用 AI management system 视角组织责任、运行控制、持续改进和管理评审 |
一句话:
AI wealth advice is an advice-boundary control architecture, not an investment chatbot with nicer language.
1. 为什么财富 AI 是边界架构问题
财富管理场景天然包含四类张力:
客户想要明确答案
+ 投资结果不确定
+ 机构有产品、费用、激励和监督义务
+ AI 擅长生成流畅语言但不天然理解法律/适当性边界
= 必须把建议边界产品化、策略化、证据化
普通金融客服 AI 的主要风险是回答错误、阻断人工或泄露信息。财富 AI 额外面临:
- 客户可能把一般教育理解成个人建议。
- AI 可能把客户一句话里的目标、资金、期限和风险偏好拼成具体推荐。
- 推荐排序可能受费用、campaign、proprietary product 或模型偏差影响。
- 投资组合建议需要同时考虑目标、期限、风险承受能力、风险承受能力外的风险承受能力、流动性、经验、税务敏感度、集中度和现有持仓。
- 交易执行把 conversational output 变成真实资金动作, 需要比普通问答更强的确认、披露、记录和监督。
- 出错后不能只说“AI 回答不准确”, 必须能重建当时的客户画像、产品宇宙、算法版本、规则评估、披露和人工参与。
PM / 架构师的高级表达:
我不会把 robo-advisor 设计成一个“推荐基金的聊天机器人”。我会把它拆成 investor profile、approved investment universe、portfolio engine、advice boundary policy、disclosure library、execution gateway、human escalation、evidence graph 和 supervision loop。LLM 只负责解释、澄清和生成受控文本, 不拥有适当性判断或交易权限。
2. Advice Boundary Taxonomy
AI 财富产品必须先定义“系统此刻在做什么”, 再决定能说什么、能展示什么、能不能进入交易。
| Layer | 客户意图 | AI 可做 | AI 不应做 | 核心控制 |
|---|---|---|---|---|
| L0 Public education | “ETF 和 mutual fund 有什么区别?” | 解释概念、费用、风险、流动性、税务一般影响 | 暗示某产品适合客户 | Approved educational content + source grounding |
| L1 General product information | “你们有哪些 IRA / managed portfolio?” | 展示公开产品类别、费用区间、最低投资额、风险说明 | 根据客户资料排序为“最适合你” | Product master + approved claims |
| L2 Neutral comparison | “这两个组合有什么差异?” | 比较公开属性、费用、资产配置、历史波动、限制 | 下结论“你应该选 A” | Comparison method + disclosure |
| L3 Needs-based guidance | “我 3 年后买房, 该怎么放钱?” | 识别目标、期限、流动性、风险承受能力缺口, 引导正式评估或人工 | 推荐具体证券或模型组合 | Profile completeness gate |
| L4 Personalized recommendation/advice | “根据我的情况推荐组合” | 在正式画像、approved universe、策略规则、披露和监督下生成推荐 | 跳过画像、忽略限制、隐藏冲突 | Suitability / best-interest style policy engine |
| L5 Portfolio management | “帮我自动再平衡” | 在客户授权、投资策略、阈值和执行控制内管理组合 | 自行扩大授权或改变策略 | Investment policy contract + monitoring |
| L6 Trade execution | “买入这个基金” | 跳转正式交易、展示风险费用、获取确认、留证据 | 用生成式文本直接替代 order ticket | Execution gateway + multi-step confirmation |
边界判断不是一次性。一个对话可能从 L0 进入 L4, 再进入 L6。每次边界提升都要重新触发:
- identity / entitlement。
- customer profile completeness。
- product universe eligibility。
- role / channel permission。
- disclosure。
- human escalation requirement。
- evidence capture。
3. Reference Architecture
Client channel
-> identity / consent / entitlement
-> intent and advice-boundary classifier
-> profile completeness service
-> approved investment universe service
-> portfolio recommendation engine
-> suitability / best-interest style policy engine
-> LLM response composer with approved claims and citations
-> disclosure and consent workflow
-> human escalation / advisor desktop
-> execution gateway
-> evidence graph, supervision, complaints, model risk monitoring
3.1 关键组件
| Component | 职责 | 不应承担的职责 |
|---|---|---|
| Intent classifier | 判断 education、comparison、recommendation、execution、complaint、human request | 不能自己决定可交易或可建议 |
| Advice boundary policy | 把客户意图映射到允许动作、需要披露、升级和拒答 | 不能依赖 prompt 里的软约束 |
| Investor profile service | 管理风险承受能力、目标、期限、财务状况、经验、流动性、限制和 freshness | 不能让 LLM 从聊天中随意推断正式画像 |
| Approved investment universe | 管理可推荐产品、模型组合、风险标签、费用、限制、冲突和有效期 | 不能把未批准证券通过检索漏进推荐 |
| Portfolio engine | 做资产配置、约束优化、模型组合匹配、再平衡、税务和集中度检查 | 不负责写自然语言披露 |
| Policy engine | 执行 eligibility、suitability、best-interest style、conflict、disclosure、human gate | 不由 LLM 解释性回答替代 |
| LLM composer | 解释概念、澄清问题、生成受控摘要和 advisor draft | 不做最终建议、产品准入、交易决策 |
| Execution gateway | 处理 order preview、确认、签名、交易限制、撤销和审计 | 不接受未结构化自然语言直接下单 |
| Evidence graph | 记录 profile snapshot、universe version、规则结果、输出、披露、确认、人工动作 | 不能只保存对话文本 |
| Supervision console | 抽样复核、例外处理、投诉联动、趋势监控和 remediation | 不能只在事故后手工查日志 |
3.2 LLM 的正确位置
在财富建议架构中, LLM 应被限制为:
- education narrator。
- clarification assistant。
- approved content composer。
- rationale explainer。
- advisor meeting note drafter。
- customer question triage。
- complaint and escalation detector。
LLM 不应作为:
- risk tolerance scoring authority。
- product approval authority。
- portfolio optimizer。
- suitability engine。
- conflict evaluator。
- order execution authority。
- legal/compliance applicability decider。
这不是低估 LLM, 而是承认财富建议的关键风险来自可审计决策边界, 不是语言能力。
4. Investor Profile Architecture
财富 AI 的核心输入不是“客户问了什么”, 而是“系统是否拥有足够、有效、可追溯的客户画像来支持当前动作”。
4.1 Profile Data Contract
| Field | Architecture role | Freshness / trigger |
|---|---|---|
| Investment objective | 判断增长、收入、保本、退休、教育、购房等目标 | 新目标、新账户或重大生命周期事件时确认 |
| Time horizon | 判断资产配置、锁定期、波动承受周期 | 目标变化、临近用款时更新 |
| Risk tolerance | 客户心理和偏好层面的波动承受 | 定期复核, 市场压力或异常行为时重新确认 |
| Risk capacity | 财务能力层面的亏损承受 | 收入、资产、负债、年龄、现金流变化时更新 |
| Liquidity need | 判断短期现金需求和退出限制 | 大额支出、退休、失业、疾病、hardship signal 时更新 |
| Financial situation | 收入、资产、负债、税务敏感度 | 账户开立、建议前、重大变更时更新 |
| Investment knowledge | 判断复杂产品理解能力 | 复杂产品或自助交易开启前检查 |
| Existing holdings | 避免集中、重复风险和不必要替换 | 每次推荐前读取或声明外部资产限制 |
| Constraints | ESG、宗教、行业排除、雇主股票限制、税务偏好 | 每次 portfolio proposal 前应用 |
| Vulnerability signals | 识别认知、语言、年龄、诈骗、压力销售风险 | 实时对话和 case-level 信号 |
| Complaint / dispute history | 识别未解决伤害、重复问题和销售后悔 | 互动前和推荐前检查 |
| Consent and channel preference | 控制个性化、营销、通信和数据使用 | 每次触达和渠道切换时检查 |
4.2 风险承受能力不是一个分数
低成熟度做法:
5 个问卷问题 -> risk score = 7 -> 推荐 aggressive portfolio
高成熟度做法:
risk tolerance
+ risk capacity
+ investment horizon
+ liquidity need
+ objective priority
+ experience and knowledge
+ concentration and external holdings
+ vulnerability and behavioral signals
= profile suitability envelope
| 维度 | 问题 | 失败模式 |
|---|---|---|
| Tolerance | 客户主观能接受多大波动 | 牛市中高估、熊市中后悔 |
| Capacity | 客户财务上能承受多大损失 | 高收入但短期现金需求被忽略 |
| Need | 目标是否需要承担风险 | 为了不必要收益承担过高风险 |
| Horizon | 资金何时使用 | 短期目标被分配到高波动资产 |
| Behavior | 客户压力下会如何行动 | 市场下跌时 panic sell |
| Knowledge | 客户是否理解复杂性 | 把结构性产品当普通存款 |
架构上应输出 profile envelope, 而不是单一 risk score:
profile_envelope:
objective: retirement_income
horizon_bucket: long_term
risk_tolerance_band: moderate
risk_capacity_band: medium_high
liquidity_constraint: medium
knowledge_level: intermediate
concentration_limit:
single_security: 10_percent
illiquid_assets: 15_percent
allowed_complexity: medium
advice_status: complete_for_model_portfolio
profile_snapshot_id: ips_2026_06_30_001
5. Approved Investment Universe
AI wealth advice 不能从“全市场知识”里自由挑产品。它必须使用 approved investment universe。
5.1 Universe Metadata
| Metadata | 用途 |
|---|---|
| product_id / cusip / ticker / model_portfolio_id | 稳定标识和证据重放 |
| product type | ETF、mutual fund、model portfolio、cash sweep、bond ladder、managed account |
| risk rating | 与 profile envelope 匹配 |
| complexity | 限制复杂产品进入自助渠道 |
| liquidity | 判断短期目标和紧急资金需求 |
| costs and fees | 费用披露、替代方案比较、conflict 分析 |
| minimum investment | 资格和组合可执行性 |
| target investor / constraints | 目标客户、限制、地域、账户类型 |
| eligible channels | digital self-service、advisor-assisted、branch、call center |
| required role/license | human advisor、registered representative、RIA channel |
| conflict metadata | proprietary product、revenue share、campaign、commission |
| approved claims | 允许描述的收益、风险、费用和限制文本 |
| prohibited claims | 禁止暗示保证、最优、无风险、适合所有人 |
| effective date / expiry | 防止过期产品和过期话术 |
| performance data source | 历史表现来源和更新时间 |
| disclosure package | prospectus、fee schedule、risk disclosure、conflict disclosure |
5.2 Product Universe Gate
candidate product
-> active and approved?
-> allowed for customer jurisdiction/account/channel?
-> allowed for customer profile envelope?
-> within complexity and liquidity limits?
-> conflict disclosure available?
-> approved claim available?
-> supervision policy satisfied?
= eligible_for_recommendation | compare_only | education_only | blocked
5.3 Universe Failure Modes
| Failure mode | Example | Control |
|---|---|---|
| Open-market hallucination | AI 推荐机构未批准基金 | Universe allowlist only |
| Stale product | 已关闭或费用变更的产品仍被推荐 | Effective-date filter |
| Channel leakage | 只允许 advisor channel 的产品出现在 self-service chatbot | Channel gate |
| Conflict blindness | Proprietary product 被排第一但无披露 | Conflict-aware ranking |
| Unsupported performance claim | AI 说“过去十年稳定跑赢” | Performance source contract |
| Missing cost comparison | 推荐高费产品但不呈现费用影响 | Fee and alternative disclosure |
6. Portfolio Recommendation Pipeline
组合推荐不应由 LLM 在对话里“想出来”。推荐流程应是结构化管道。
goal intake
-> profile completeness check
-> account and restriction check
-> approved universe filter
-> strategic asset allocation / model portfolio matching
-> constraints and concentration check
-> suitability / best-interest style policy check
-> rationale and alternatives
-> disclosure and customer confirmation
-> execution workflow
-> ongoing monitoring and rebalancing
6.1 Recommendation Objects
| Object | 示例 |
|---|---|
| goal | retirement income, emergency reserve, college savings |
| recommended allocation | 60% equity ETF, 35% bond ETF, 5% cash |
| instrument mapping | approved ETFs / model portfolios |
| constraints applied | no sector funds, liquidity reserve, tax-sensitive account |
| excluded alternatives | higher risk model excluded due to liquidity need |
| rationale | objective / horizon / risk / cost / diversification match |
| disclosures | market risk, fees, conflicts, no guarantee, rebalancing rules |
| action path | save plan, discuss with advisor, open account, trade preview |
| evidence id | recommendation_trace_id |
6.2 Decision Table
| Customer / profile signal | Candidate | AI action | Policy decision |
|---|---|---|---|
| Emergency fund goal, 6-month horizon | Aggressive equity model | Recommend | Block, explain horizon/liquidity mismatch |
| Retirement, 20-year horizon, moderate profile | Balanced model portfolio | Recommend | Allow with standard risk/fee disclosure |
| Profile incomplete | Any personalized portfolio | Recommend | Ask structured questions or advisor handoff |
| High concentration in employer stock | Equity-heavy model | Recommend | Require concentration warning and advisor review |
| Customer asks “highest return” | Leveraged ETF | Recommend | Education only, no specific product recommendation |
| Customer mentions complaint about prior advice | Any product | Continue sales flow | Pause recommendation, complaint workflow |
| Vulnerability signal + complex product | Structured note | Compare | Licensed / trained human escalation |
| Outside approved universe request | Meme stock | Buy | Education or execution-only boundary, no recommendation |
6.3 推荐理由要有证据, 不是营销文案
弱理由:
这个组合很适合你, 因为它收益潜力高, 风险也比较平衡。
强理由:
系统选择 balanced model 的原因是: 你的目标期限为 10 年以上, 风险档位为 moderate, 已预留现金储备, 且组合费用低于同类上限。higher-growth model 被排除, 因为它超过了当前 risk envelope 的权益暴露上限。该解释不保证收益, 也不替代正式投资顾问审查。
架构要保存:
- 匹配规则。
- 被排除方案。
- profile snapshot。
- product universe version。
- fee/cost basis。
- disclosure set。
- AI output hash。
7. Best-Interest Style Control Model
本文不判断某个产品或渠道是否适用 SEC Reg BI、investment adviser fiduciary duty、FINRA 规则或其他义务。准确适用性由 Legal / Compliance 拥有。架构师要做的是把 best-interest 风格问题转成控制对象。
7.1 Control Translation
| Best-interest style question | Architecture translation |
|---|---|
| 客户是否获得足够重要信息 | Disclosure library, timing, comprehension, evidence |
| 推荐是否基于客户具体情况 | Profile completeness and suitability envelope |
| 产品风险、费用、替代方案是否考虑 | Product metadata, cost comparison, alternative rationale |
| 机构或员工冲突是否被识别和披露 | Conflict metadata, ranking constraints, approved disclosure |
| 推荐是否被监督 | Supervision sampling, exception queue, advisor review |
| 机构能否证明当时怎么做的 | Evidence graph with versions and decision trace |
7.2 Care / Conflict / Disclosure / Supervision 四层
| Layer | 控制目标 | 设计实现 |
|---|---|---|
| Care-style | 推荐要和客户画像、目标、期限、风险、成本、限制匹配 | Suitability policy engine + portfolio rationale + alternatives |
| Conflict-style | 排序和话术不能隐藏机构激励 | Conflict-aware ranking + disclosure + compensation metadata |
| Disclosure-style | 重要信息要在正确时点出现 | Contextual disclosure service + click/acknowledgement evidence |
| Supervision-style | 推荐、例外、投诉、员工采纳和 AI 输出要可监督 | Surveillance dashboard + sampled review + issue remediation |
7.3 Ranking Policy
财富 AI 的排序不是单纯 recommender system。
ranking_score =
customer_goal_fit
+ risk_profile_fit
+ cost_reasonableness
+ diversification_benefit
+ liquidity_fit
- conflict_penalty
- complexity_penalty
- concentration_penalty
PM 要求不能只写“推荐最匹配产品”。要写:
- 允许哪些特征进入排序。
- 禁止哪些特征进入排序。
- 如何处理 proprietary product。
- 如何展示 alternatives。
- 如何记录被排除原因。
- 如何在投诉、监管问询或客户争议时重放。
8. Education vs Advice vs Execution: Runtime Boundary
8.1 Boundary Decision Tree
客户是否只问一般概念?
是 -> education answer with source and no product push
否 -> 是否问具体产品公开信息?
是 -> product info / neutral comparison
否 -> 是否提供个人资金、目标、期限、风险或账户信息并要求建议?
是 -> profile and advice-boundary gate
否 -> 是否要求下单、调仓、卖出或再平衡?
是 -> execution gateway, not free-form chat execution
否 -> clarify intent
8.2 对话中常见越界信号
| Customer utterance | Boundary risk | Default handling |
|---|---|---|
| “我该买哪个?” | Personalized recommendation | Ask profile / route to controlled advice |
| “哪个收益最高?” | Return-chasing and risk understatement | Explain risk-return tradeoff, no product push |
| “帮我把退休金都换成股票” | Execution + suitability risk | Strong warning, advisor/human review |
| “这个是不是稳赚?” | Guarantee risk | Reject guarantee, explain market risk |
| “我看不懂, 你替我决定” | Delegation / authority risk | Clarify AI cannot make unsupported decision |
| “上次你们建议我亏了, 我要投诉” | Complaint | Pause sales, create complaint workflow |
| “我急需用钱还债” | Vulnerability / liquidity | Stop investment recommendation, route support |
8.3 Channel and Role Boundaries
| Channel / role | Allowed baseline | Restricted baseline |
|---|---|---|
| Public web AI | Education and product information | Personalized recommendation, account-specific execution |
| Logged-in self-service | Profile-driven planning and model proposal if permitted | Complex product advice without required workflow |
| Advisor copilot | Draft notes, compare products, prepare rationale | Send recommendation without advisor approval |
| Call center agent assist | Education scripts and escalation cues | Investment advice scripts beyond role |
| Robo-advisor engine | Model portfolio recommendation within program | Open-ended securities picking outside program |
| Execution-only flow | Customer-directed trading with required controls | AI saying a trade is best for the customer |
9. Human Escalation Architecture
财富 AI 的人工升级不是客服兜底, 而是 boundary control。
9.1 Escalation Triggers
| Trigger | Why it matters | Action |
|---|---|---|
| Profile incomplete or stale | 个性化建议依据不足 | Ask structured questions or advisor review |
| Complex product request | 理解、费用、流动性和风险更高 | Licensed / trained human handoff |
| Vulnerability signal | 更高 conduct risk | Slow down, simplify, human support |
| Complaint or regret | 销售监督和补救入口 | Pause recommendation, create case |
| Large trade / liquidation | 客户伤害和后悔风险高 | Additional confirmation and review |
| Concentrated position | 单一证券或行业风险 | Advisor review and concentration disclosure |
| Low confidence / missing source | 幻觉风险 | Refuse or handoff |
| Prompt-injection or adversarial instruction | 安全和越权风险 | Refuse, log, security review |
| Market stress event | 客户情绪驱动交易 | Cooling language, education, advisor option |
| Tax / legal / estate planning | 跨专业边界 | Route to appropriate professional process |
9.2 Warm Handoff Package
人工接手时不能让客户重复全部信息。handoff package 应包含:
- customer intent summary。
- boundary tier。
- profile completeness status。
- known constraints。
- products discussed。
- recommendation candidates and excluded alternatives。
- disclosures shown。
- AI confidence and refusal reasons。
- complaint / vulnerability flags。
- full trace id and conversation excerpt。
10. Hallucination Control
在财富 AI 中, 幻觉不是“回答不准”这么简单。它可能变成虚假收益、错误费用、未批准产品、错误税务暗示、交易承诺或不当建议。
10.1 Control Stack
source authority
-> retrieval permission filter
-> approved claims library
-> structured product/profile APIs
-> constrained generation
-> post-generation policy scanner
-> source support check
-> refusal / escalation
-> evidence logging
10.2 Forbidden Wealth Claims
| Category | Block pattern |
|---|---|
| Guarantee | 保证收益、不会亏、一定跑赢、无风险 |
| Suitability overclaim | “最适合你”, “唯一正确选择”, “不用再考虑” |
| Performance overclaim | 选择性引用历史收益, 暗示未来结果 |
| Fee omission | 推荐高费或复杂产品但不披露费用 |
| Liquidity omission | 忽略锁定期、赎回限制、罚金 |
| Conflict hiding | 不披露 proprietary、commission、revenue share |
| Tax/legal overreach | 给出个性化税务或法律结论 |
| Execution ambiguity | 把聊天里的“好”当成交易确认 |
| Human impersonation | 让客户以为 AI 是持牌顾问本人 |
10.3 Response Contract
高风险回答应遵守结构化输出:
response_type: education | compare | profile_required | recommendation_summary | execution_redirect | escalation
boundary_tier: L0-L6
source_ids: []
profile_snapshot_id: optional
universe_version: optional
policy_decision_id: optional
disclosures: []
human_escalation: true_or_false
customer_visible_text: string
blocked_claims_detected: []
这样做的目的不是让客户看到 YAML, 而是让系统在生成前后都能测试、拦截、记录和重放。
11. Evidence Graph
财富 AI 的证据不应只保存 chat transcript。真正有用的是 decision-time evidence graph。
11.1 Evidence Events
| Event | Evidence |
|---|---|
| Identity / session | customer id, channel, auth level, consent |
| Intent classification | utterance, predicted intent, boundary tier, confidence |
| Profile check | profile snapshot, freshness, missing fields |
| Universe filter | universe version, included/excluded products, reason codes |
| Portfolio generation | engine version, constraints, allocation, alternatives |
| Policy decision | rule ids, pass/fail, disclosure and escalation requirements |
| LLM generation | prompt version, model route, source ids, output hash |
| Disclosure | disclosure ids, timing, customer acknowledgement |
| Human review | reviewer id, decision, changes, rationale |
| Execution | order preview, confirmations, trade id, cancellation path |
| Complaint / incident | case id, linkage to trace, remediation |
| Monitoring | KRI breach, alert, review outcome |
11.2 重放能力
监管、审计、投诉和模型风险复盘会问:
- 当时客户画像是什么?
- 哪些产品被允许进入候选集?
- 哪些产品被排除, 原因是什么?
- 推荐算法和规则版本是什么?
- 客户看到哪些披露?
- AI 是否使用了 approved claims?
- 人类是否参与, 做了什么修改?
- 交易是否由客户明确确认?
- 后续是否有投诉、撤销、补救或控制更新?
如果系统只能回答“我们有聊天记录”, 证据能力不足。
12. Complaint, Supervision and Surveillance
财富 AI 需要把投诉、监督和模型风险接到同一条闭环。
12.1 Complaint Detection
AI 应识别:
- “你们之前建议错了”。
- “我亏钱了, 要投诉”。
- “我没有理解费用”。
- “这个不是我想买的”。
- “顾问/系统误导了我”。
- “我以为这是保本的”。
默认处理:
complaint signal
-> pause recommendation and sales language
-> create or route complaint case
-> preserve trace and evidence
-> provide customer confirmation
-> notify supervision / conduct risk workflow
12.2 Supervision Signals
| Signal | Interpretation |
|---|---|
| High override rate | AI recommendation or policy rules not aligned with advisor judgment |
| High customer abandonment after disclosure | Disclosure may be confusing or recommendations too aggressive |
| Complaint cluster by product | Product explanation, fees or suitability controls may be weak |
| Concentrated recommendations | Ranking model or campaign conflict may be oversteering |
| Frequent boundary refusals | UX may invite advice requests before profile completion |
| High manual correction of AI drafts | LLM generation or approved claims retrieval weak |
| Segment-specific failure | Risk of unequal customer experience or accessibility issue |
| Market-stress sell prompts | Behavioral risk and escalation policy need tightening |
12.3 Supervision Operating Loop
sample interactions
-> review recommendation traces
-> classify finding
-> determine customer impact
-> update policy / prompt / universe / disclosure
-> retrain reviewers and advisors
-> test control effectiveness
-> document closure evidence
13. Model Risk Management
AI wealth advice 是 AI system risk, 不是单一模型风险。
13.1 Validation Scope
| Area | Validation question |
|---|---|
| Use case boundary | 是否清楚区分 education、recommendation、execution |
| Profile data | 画像是否完整、有效、可追溯 |
| Universe data | 产品、费用、风险、限制、披露是否准确 |
| Portfolio algorithm | 资产配置、约束和优化是否符合设计 |
| Policy engine | 是否拦截 profile 缺口、冲突、复杂产品和高风险场景 |
| LLM output | 是否 grounded, 是否避免 forbidden claims |
| Human oversight | 升级是否真实、及时、可用 |
| Execution workflow | 交易确认、撤销、记录是否完整 |
| Monitoring | 投诉、override、drift、segment failure 是否可见 |
13.2 Eval Set Design
评估集应覆盖:
- 普通教育问答。
- 产品比较。
- 画像不足仍要求推荐。
- 高风险偏好但低风险承受能力。
- 短期目标却要求高收益。
- vulnerable customer。
- 投诉和后悔。
- 市场剧烈下跌。
- 外部证券请求。
- 复杂产品请求。
- prompt injection。
- 多语言和低金融素养表达。
- advisor copilot copy-paste 风险。
13.3 Revalidation Triggers
- base model 或 model route 变化。
- prompt / system instruction 变化。
- RAG source 或 product universe 变化。
- risk questionnaire 或 profile scoring 变化。
- portfolio optimizer 或 model portfolio 变化。
- disclosure library 变化。
- 新产品、新渠道、新客户群。
- 投诉趋势、supervision finding、incident。
- 法规、监管 guidance 或内部 policy 更新。
14. PM / Architect Implications
| PM / architecture decision | 高级判断 |
|---|---|
| 是否做 customer-facing AI wealth assistant | 先定义允许边界和人类升级, 再设计对话体验 |
| 是否允许个性化推荐 | 只有在 profile、universe、policy、disclosure、evidence 和 supervision 成熟后进入 |
| 是否用 LLM 做 portfolio recommendation | LLM 不应做优化和适当性判断, 只做解释和受控生成 |
| 是否开放自然语言交易 | 应通过 execution gateway 和多步确认, 不让聊天文本成为订单 |
| 如何衡量成功 | 不能只看 AUM conversion; 要看 recommendation quality、complaints、override、evidence completeness、boundary accuracy |
| 如何处理客户体验 | 边界说明要清楚但不恐吓; 客户应知道何时需要画像、何时转人工、何时进入正式交易 |
| 如何处理业务增长压力 | 用 conflict-aware ranking 和 supervision 防止 AI 变成隐性销售推动器 |
15. Interview-Ready Language
15.1 30 秒版本
我会把 AI wealth advice 设计成 advice-boundary control architecture。教育、产品信息、个性化推荐、组合管理和交易执行要分层。风险承受能力、目标、期限、流动性、现有持仓和 approved investment universe 由结构化服务管理, 组合推荐由 portfolio engine 和 policy engine 控制, LLM 只做解释、澄清和受控文本生成。上线必须有披露、人工升级、证据图、投诉监督和模型风险验证。
15.2 2 分钟版本
Robo-advisor 的关键不是回答“买什么”, 而是证明推荐是在当时客户画像、产品宇宙和控制规则下生成的。我会先做 advice boundary taxonomy: education、neutral comparison、personalized recommendation、portfolio management、execution。然后建立 investor profile service, 包括 risk tolerance、risk capacity、horizon、liquidity、knowledge、constraints 和 freshness。产品侧建立 approved investment universe, 带 risk、fee、liquidity、complexity、conflict、channel 和 approved claims metadata。推荐由组合引擎生成候选, 再经过 suitability / best-interest style policy engine, 输出 rationale、alternatives、disclosures 和 evidence。LLM 不能绕过这些服务, 只负责解释和摘要。高风险、画像不足、复杂产品、投诉、vulnerable signal 和低置信都升级人工。最后用 evidence graph 保存 profile snapshot、universe version、rule results、disclosure、human review 和 execution confirmations, 并通过 complaints、supervision、override、drift 和 model validation 持续监控。
15.3 面试追问
| 追问 | 答法要点 |
|---|---|
| AI 能不能直接推荐基金? | 不用一句话回答能或不能。先问机构角色、渠道、客户画像完整度、产品宇宙、监督和法律适用。架构上只有在正式受控 advice workflow 内才允许个性化推荐。 |
| 怎么防止 hallucination? | 把投资事实、产品、费用、风险和披露从权威服务取出; LLM 只用 approved claims 和 citations 生成; post-generation scanner 拦截保证收益、最适合、费用遗漏等 forbidden claims。 |
| 怎么证明推荐合理? | 保存 profile snapshot、universe version、algorithm version、rules fired、excluded alternatives、fee/risk rationale、disclosures、customer confirmation 和 human review。 |
| robo-advisor 和 advisor copilot 的边界差异? | Robo-advisor 是客户程序化建议和组合执行流程; advisor copilot 是员工辅助。copilot 输出仍需 advisor 复核, 不能让员工把 AI draft 当最终建议。 |
| 业务想提高 conversion 怎么办? | 优化 profile completion、explanation clarity 和 handoff, 不能用隐藏冲突、弱化风险或 push high-fee product 提高转化。用 conduct KRIs 平衡 AUM growth。 |
16. 与已有学习资产的连接
| 已有文档 | 连接方式 |
|---|---|
docs/AI_CONDUCT_RISK_SUITABILITY_SALES_GUARDRAILS_PLAYBOOK.md | 本文把 suitability / sales guardrails 聚焦到财富建议、组合推荐和交易边界 |
docs/AI_CUSTOMER_FACING_REGULATED_PRODUCT_PLAYBOOK.md | 本文扩展其中的 customer-facing AI 边界到 wealth-specific advice workflow |
docs/AI_MODEL_RISK_MANAGEMENT_PLAYBOOK.md | 本文把 MRM 具体化为 robo-advisor 的 profile、universe、portfolio engine、LLM 和 supervision validation |
docs/AI_MODEL_VALIDATION_INDEPENDENT_CHALLENGE_PLAYBOOK.md | 本文提供财富 advice 系统的 independent challenge 问题清单 |
docs/AI_CONTROL_LIBRARY_ASSURANCE_EVIDENCE_GRAPH_PLAYBOOK.md | 本文的 evidence graph 可作为财富建议场景的 control-evidence slice |
17. 核心结论
AI 财富建议的高级架构判断是:
- 先定义 advice boundary, 再做对话体验。
- 客户画像是正式数据资产, 不是聊天中随手推断。
- 推荐只能来自 approved investment universe。
- 组合推荐由 portfolio engine 和 policy engine 生成, LLM 负责解释。
- Best-interest 风格控制要落成 disclosure、care、conflict、supervision 和 evidence。
- 人工升级是产品能力, 不是异常客服。
- 投诉和监督要能反向更新 prompt、policy、universe、questionnaire 和模型评估。
- 证据必须能重放每一次客户影响型建议。