ZOLA ✦ Work 01 — Insight Hub All work →
AI Strategy Product · 0→1 · Product Owner

Insight Hub

A subscription intelligence layer that clients' own AI assistants query directly.

Insight Hub 是从 0 搭建的洞察资产中台,由实时洞察、创意拆解、月度深研三条生产线持续供给,已沉淀 376 条结构化可检索资产。对内成为销售提案的素材库,对外已作为 MCP 服务上线,客户的 Claude / ChatGPT 可在对话中直接调用。

产品实录 · 3 分 17 秒完整走查

What I owned

作为项目总负责人,从 0→1 搭建并运营这套商业化机制,四类实际交付

Owned end to end: converged 20+ requests into five core features, wrote the PRD, defined the product logic and every front-end screen (R&D built the backend), designed the collection → vectorisation → retrieval pipeline, and led the positioning research that benchmarked 10+ commercial MCP services to set pricing, tool scope and phased rollout.

Definition
需求定义与规格
  • 把销售、创研、市场的分散诉求逐条翻译为功能语言,20+ 需求收敛为五大核心功能
  • 撰写 PRD v1.1:信息架构、核心流程、字段规范、验收标准(DoD)
  • 定义北极星与次级指标口径,并写清 Phase 2 启动前必须回答的开放问题
Design
产品逻辑与界面
  • 设计全部前端界面与交互:三类洞察库、收藏/反馈、需求社区、关系图谱、管理后台
  • 定义四类角色权限(超管 / 管理员 / AE·AM / 全员),按职责而非职级划分
  • 关键决策:录入层与存储层分离,两条录入路径落成同一套 schema
Delivery
跨职能交付
  • 向创研讲清业务场景、对齐数据规范,推动采集 → 向量化 → 检索链路落地
  • 在技术求稳与快速验证之间做 MVP 取舍:排优先级、保核心体验、快速演示收反馈
  • 平台、实时洞察、月度 PPT、社媒、客户提案多线并行调度
Commercial
商业化与采用
  • 主导商业化定位调研,对标 10+ 家商业 MCP 服务,确定收费模型、工具范围与分阶段路线
  • 输出对外定位声明与 MCP 工具设计,推动 Phase 2a 落地
  • 搭建采用闭环:使用追踪、反馈表单、培训会、需求社区,北极星=洞察有效使用率
376
Insight assets live
结构化洞察资产上线
−75%
Time to insight
2–3 小时 → 0.5 小时
7
High-ticket engagements
支撑的高客单价商机
0→1
To a client-facing service
从内部资产到对客服务

Inside the product

GTM mechanism → commercial pipeline → productized asset. A platform was the means, not the objective; the business needed a repeatable commercial mechanism: signals in, retrievable assets banked, sales materials out, proven components packaged into products we can sell. GTM 市场机制 → 商业线索 → 可产品化资产。平台是手段,不是目的;业务真正需要的是一套可复用的商业机制:市场信号进来,沉淀为可检索资产,输出为销售可用素材,验证过的部分再包装成可售产品。

01

The Growth Flywheel

增长飞轮策略

It began with strategy, not tooling. Market analysis exposed two gaps: lead generation had no systematic mechanism, and market-sensing sat with individuals. The growth-flywheel strategy I proposed in Feb–Mar answered both — a business flywheel for acquisition, an organisational one for asset accumulation. It was the first time both shared a single framework, and everything built afterwards hangs off this line. 起点是战略判断,不是工具。公司战略规划与市场分析暴露出两个缺口:销售线索缺少系统化机制、市场感知能力依附于个人。2–3 月独立提出的双飞轮增长策略同时回答这两点:左轮负责获客与转化,右轮负责资产沉淀与提效。获客与组织提效第一次有了统一框架,后续所有建设都挂在这条线上。

Growth flywheel strategy
02

Real-time Market Sensing

实时洞察 · JA-Research

Daily market signals across 40+ verified sources, each carrying the summary, the strategic read, three recommendations and full source attribution. Content is produced by JA-Research, the company's first AI skill, architected end to end — prompt system, source network and template bases. Content throughput rose 3–4×, and time to a client-ready insight dropped from 2–3 hours to 30 minutes. 覆盖 40+ 验证信源的每日市场信号,每条包含摘要、战略解读、三条可执行建议与完整信源标注。内容由 JA-Research 生产——公司首个 AI Skill,提示词体系、信源网络与模板底座均为独立搭建——以结构化 markdown 上传后自动解析成卡。内容产能 3–4×,洞察就绪从 2–3 小时降到 30 分钟。

Real-time market sensing feed
03

Monthly Deep-dives

月度深度洞察

Research packaged as ready-to-use client material — each card carries the talking points, the pitch language and the recommended service line, so an account manager lifts it straight into a proposal without rewriting. 20 topic streams fed 7 client engagements, from tenders to high-ticket service packaging. 深度研究被包装成可直接对客的物料:每张卡都带沟通要点、对客话术与推荐服务线,客户经理无需改写即可搬进提案。20 个主题方向支撑 7 个对客项目,覆盖投标到高客单价服务包装。

Monthly insights library
04

Creative Teardowns

创意洞察库

What's working across Xiaohongshu, TikTok, Instagram and YouTube — each entry tagged by method rather than topic, which makes it a reusable technique bank for campaign planning instead of a reference folder. 37 teardowns became source material for product narratives and client-facing creative proposals. 覆盖小红书 / TikTok / Instagram / YouTube 的有效创意:每条按手法而非主题打标,成为可复用于策划的技法库,而不是参考文件夹。37 条创意拆解成为产品叙事与对客创意提案的素材来源。

Creative insight library
05

Client Interest Graph

客户兴趣关系图谱

49 clients · 209 tags · 58 dimensions mapped as a live graph, with each client's decision-makers and interest dimensions attached. The team runs the question in reverse: not "what does this client care about", but "which clients should receive this insight". That inversion turned a content library into a business-development tool, feeding 7 MQLs and a 4+10 GTM opportunity pool. 49 家客户 · 209 个标签 · 58 个维度连成动态图谱,每家客户的决策人与兴趣维度都挂在节点上。团队因此能反向提问:不是"这个客户关心什么",而是"这条洞察该推给谁"。这一反转把内容库变成商机工具,支撑 7 个 MQL(市场合格线索)与 4+10 GTM 机会池。

Client interest graph
06

Admin & Measurement

后台与量化机制

Publishing trends, tagging heat and feedback boards, with one-click exports and daily pushes. The North Star is measured here: insights tagged "used" or "closed a deal", not page views. Product logic and every screen were designed in-house; R&D built only the backend. The company's first in-house platform now backs partnerships with NUS, HKU and other leading universities. 发布趋势、标注热度、团队反馈看板,加上一键导出与每日定时推送。北极星指标就在这里被度量:被标注「已使用 / 促成销售」的洞察占比,而不是浏览量。产品逻辑与全部界面均为独立设计,创研仅负责后端。作为公司首个自建平台,已进入公司可商业化资产清单,目前支撑与新加坡国立大学、香港大学等多所高等院校的合作。

Admin console and measurement