Spaces:
Running
Running
| title: Agent Ken — Data-Informed PM Copilot | |
| emoji: 🤖 | |
| colorFrom: blue | |
| colorTo: purple | |
| sdk: gradio | |
| sdk_version: "6.5.1" | |
| app_file: app.py | |
| pinned: true | |
| license: mit | |
| short_description: Data-informed PM Copilot | |
| # 🤖 Agent Ken — Your AI Companion for Product Management & Data | |
| An AI assistant that helps you **learn and practice product management** — from strategy, frameworks, and execution to real-world data analysis. | |
| Agent Ken is also integrated by **Microsoft Fabric** with real product analytics: **5,000 users**, **204K events**, **90 days of metrics**, and **3 trained ML models** (churn, LTV, anomaly detection). But it's not just about data — you can also ask about **PM frameworks, AI & tech, business strategy, and more**. | |
| > **This isn't a generic ChatGPT wrapper.** Ask about retention, and you'll get actual numbers. Ask about PRDs, and you'll get a structured template. Ask about AI, and you'll get practical insights. | |
| ## ⚡ Powered By | |
| | Component | Technology | | |
| | :--- | :--- | | |
| | 🧠 **AI Agent** | Azure AI Foundry Agent Service (GPT-5) | | |
| | 🏭 **Data Platform** | Microsoft Fabric (OneLake + MLflow) | | |
| | 🤖 **ML Models** | Churn Prediction · LTV Estimation · Anomaly Detection | | |
| | 🖥️ **Frontend** | Gradio (Hugging Face Spaces) | | |
| ## 🚀 What Can Agent Ken Help You With? | |
| ### 📊 Data-Informed (Powered by Fabric ML Pipeline) | |
| Ask about real product metrics — Agent Ken answers with **specific numbers**, not vague advice: | |
| * "What's our D7 retention?" → **25.6%**, with segment breakdown by channel. | |
| * "Which channel has highest LTV?" → **Organic at $25.65** (1.9x vs Paid Social). | |
| * "Any anomalies in our metrics?" → **Nov 24-25 churn spike** (2.8x average). | |
| * "How did checkout experiment do?" → **+11.9% conversion**, p=0.027, ship it. | |
| ### 🧠 General PM (Powered by GPT-5.1 Knowledge) | |
| All the PM capabilities you'd expect from a senior product manager: | |
| * 🔍 **Product Discovery** — Jobs-to-be-Done, personas, problem statements, user journeys. | |
| * 📊 **Prioritization** — RICE scoring, MoSCoW, trade-off analysis. | |
| * 📝 **PRDs & User Stories** — Product requirements, acceptance criteria, MVP definition. | |
| * 🧪 **Experiment Design** — A/B tests, hypothesis templates, success metrics. | |
| * 📈 **Metrics & OKRs** — North Star Metrics, funnel metrics, guardrail metrics. | |
| * 🤝 **Stakeholder Communication** — Decision memos, release notes, progress updates. | |
| * ♿ **Inclusive Design** — Accessibility, ethical considerations, vulnerable user safety. | |
| --- | |
| ## 💾 What Data Powers Agent Ken? | |
| A synthetic but realistic product analytics dataset simulating a mobile app over 3 months: | |
| | Dataset | Records | What's Inside | | |
| | :--- | :--- | :--- | | |
| | **Users** | 5,000 | Channel, country, plan, age group | | |
| | **Events** | 204,356 | App open, purchase, support ticket, etc. | | |
| | **Daily Metrics** | 90 days | DAU, revenue, retention, churn, NPS | | |
| | **A/B Tests** | 5 experiments | Checkout, onboarding, push timing, pricing, dark mode | | |
| ### 🔑 Key Numbers | |
| | Metric | Value | | |
| | :--- | :--- | | |
| | Avg DAU | **980** | | |
| | D7 Retention | **25.6%** | | |
| | Avg LTV | **$19.87** | | |
| | Churn Rate | **34.2%** | | |
| | 90-day Revenue | **$99,241** | | |
| --- | |
| ## 🤖 ML Models Trained in Fabric | |
| | Model | Algorithm | Score | Key Insight | | |
| | :--- | :--- | :--- | :--- | | |
| | **Churn Prediction** | Gradient Boosting | AUC: 1.00 | Users inactive 3+ days = high risk | | |
| | **LTV Estimation** | Linear Regression | R²: 0.77 | Organic users worth 1.9x more than paid | | |
| | **Anomaly Detection** | Isolation Forest | 5 anomalies | Nov 24-25 churn spike = possible incident | | |
| ## 🧪 A/B Test Results | |
| | Experiment | Lift | Significant? | Action | | |
| | :--- | :--- | :--- | :--- | | |
| | **Checkout Redesign** | +11.9% | ✅ p=0.027 | **Ship** ✅ | | |
| | **Onboarding Simplification** | +8.0% | ✅ p=0.021 | **Ship** ✅ | | |
| | **Push Notification Timing** | +22.0% | ✅ p=0.035 | **Ship** ✅ | | |
| | **Premium Pricing Test** | -5.0% | ❌ p=0.289 | **Rollback** ❌ | | |
| | **Dark Mode Default** | +3.0% | ❌ p=0.212 | **Extend test** ⏳ | | |
| --- | |
| ## 💡 How to Use | |
| 1. **Type your question** in the chat box. | |
| 2. **Upload files** — drop your CSV, PDF, Excel, or other files. I'll analyze them and give you actionable insights. | |
| 3. **Ask about data** — Agent Ken will cite specific numbers from the Fabric pipeline. | |
| 4. **Ask about PM** — Agent Ken will use product management expertise. | |
| 5. **Try the examples** — click "Try an example" for inspiration. | |
| 6. **Start fresh** anytime with "New Conversation". | |
| ## 🎯 Tips for Best Results | |
| * ✅ *"What's our retention by acquisition channel?"* | |
| * ✅ *"Help me write a PRD for a referral program based on our LTV data"* | |
| * ✅ *"Score these features using RICE: push notifications, dark mode, onboarding revamp"* | |
| * ✅ *"Design an A/B test for our new premium pricing"* | |
| * ❌ Off-topic questions (biology, cooking, etc.) — Agent Ken is scoped to PM & tech. | |
| ## 👤 About | |
| Built by **Kendrick Filbert** — AI + PM + Social Impact Practitioner | |
| *Powered by **Azure AI Foundry** (GPT-5.1) · **Microsoft Fabric** (OneLake + MLflow) · **3 ML Models** (Churn · LTV · Anomaly)* |