Building an A.I. Assistant that Speaks Human
summary
A scheduling agent built at mimetic.ai reads and writes email and autonomously finds slots by extracting signatures, tokenizing and POS tagging, applying an ontology, a semantic chunker and a graph to resolve phrases like “lunch” or “Holland Village,” handle time zones and meeting rooms, and propose/confirm times. It learns per-organization terminology, supports ambiguous commands like “sometime next week,” targets busy executives and startups (not mass market), and claims productivity gains close to six weeks per year while people only spend ~45% of their day on core work.
AI-written summary · about
chapters (11)
- 0:00AI assistants vs chatbots
- 1:52Human-like AI assistants
- 3:58Natural language challenges
- 9:13NLP pipeline for scheduling
- 11:41Scheduling logic beyond language
- 13:30Human-like AI responses
- 16:50Open-source tools and knowledge
- 21:24Handling unknown scenarios
- 23:13AI assistant market fit
- 26:46Learning across users
- 27:51Future of AI assistants
related talks
16mDon't just talk to AI, do more with AI: how to improve productivity with AI agents
16mAI is your personal co-founder
49mHow to be an AI First company
22mThe AI-augmented leader: Architecting your own virtual chief of staff
29mAgentic AI: What It Is and Why It Matters
28mBuilding a Micro AI company in 2017
26mFour Steps to Embed AI Agents into Workflow
40mAI Agents in Practice