AI For Everyday Work

This one was fun. An evening with AI Builders and Qoder in Singapore, where I represented Notion alongside the Notion SG Community and shared how Notion AI works best inside a team operating system, not as a bolt-on. The theme I kept coming back to all night: the best AI setups cut busywork, respect permissions, and keep humans accountable for the final call.

Photos

What I covered

• Why Notion AI gets more valuable in a team workspace, where docs, projects, and decisions stay connected instead of scattered across tools
• On-demand agents vs autonomous agents, and when each one earns its keep
• My core design rule: build specific, scoped, permission-aware agents, never generic do-everything bots
• Multi-agent orchestration in practice, like one agent summarising a meeting while another assigns the follow-ups
• Keeping humans in the loop, because AI handles the repetitive work but judgement stays with people
• The crowd favourite: a live demo using Coda with Notion MCP to search my workspace and spin one event page into an HTML presentation in a single shot

Key takeaways

• Embed AI where work already happens. A separate AI layer just becomes another tab nobody opens
• Your context, permissions, and business logic are the moat, not a custom model
• Automate one painful, repetitive workflow first and do it well before scaling up
• Least privilege for agents: only the access they need, with a clear end-to-end scope
• Fundamentals still win. Critical thinking, communication, and accountability do not get automated

Outcome

• The questions afterwards were the best part, all about practical team use cases rather than abstract AI talk
• I had great conversations on Notion agents, secure AI adoption, and how lean teams can start small without overcomplicating their stack
• My takeaway from the night: calmer systems win, especially when AI quietly absorbs the admin and context-switching