Building a Micro AI company in 2017
summary
Overview of building a tiny AI product focused on narrow domains like chat automation and email scheduling, with practical steps: pick a specific customer, collect and label data, iterate fast, and monetise by tying models to business outcomes. Examples include a production KeyReply chatbot for media and government workflows, using open corpora (Gigaword, Enron) and generating missing examples with GTA synthetic data, automating parking-ticket appeals and asylum forms, and launching on platforms where 40,000 bots prove distribution.
AI-written summary · about
chapters (19)
- 0:00Intro
- 2:22Challenges of AI
- 2:50Heads Up Poker Chips
- 3:36Big Companies Buying Smaller Companies
- 4:20Chatbots
- 5:10Singapore Voice
- 5:32Do we just sit back
- 6:22The Numbers
- 6:57The Problem
- 9:55The Good
- 12:20Data
- 13:37Job Scope
- 14:30Image Recognition
- 15:29Open Data Movement
- 16:39Email Scheduler
- 17:57Generating Data Automatically
- 19:08Fast iterations
- 19:25Experiment and deploy
- 19:43Is AI bad
related talks
27mBuilding AI applications: Successes and Lessons
15mIt doesn’t need to be a chatbot: Unlocking the product value of smaller AI models
49mHow to be an AI First company
27mBuilding Generative AI features
26mFour Steps to Embed AI Agents into Workflow
28mThe AI-First Playbook: Building and Running a Business with AI
26mFour Steps to Embed AI Agents into Workflow
25mConversational AI: Sales and Customer Service Chatbot Use Cases for Small and Medium Sized Business