Project Deck · Workshop Recap
Enterprise Singapore
× Notion AI Workshop
From Notion AI fundamentals to autonomous agents and AI computer use. A hands-on evening for a mixed audience, with live demos throughout.
The Brief
An evening built around live building
- A mixed audience, mostly Notion beginners with a handful of intermediate users, taken end to end from fundamentals to autonomous agents.
- Three presenters split the arc: Ernest on Notion and Notion AI basics, Nasri on AI skills and computer use, Dennis on custom agents and Notion Workers.
- Every attendee got a demo workspace to keep for two weeks after the session to experiment on their own.
- The whole event ran with AI meeting notes transcribing live in the background, which became a demo in itself.
The Arc
Five acts, one continuous demo
01
Notion AI Basics
Databases, views, and Notion AI as your context layer across connected tools.
02
AI Skills
Reusable instruction pages that replace prompt engineering and stay consistent across a team.
03
Custom Agents
Autonomous agents that run on a schedule or trigger, with fine-grained permissions.
04
Computer Use
A sandboxed environment that turns documents and data into HTML artefacts.
05
Notion Workers
Cheap scheduled programs that sync external data and power a lightweight BI stack.
Q&A
Live throughout
Open questions after every section, answered on the spot.
Ernest
Notion AI as your context layer
- Notion is a workspace tool built around databases, structured pages with consistent properties across table, board, calendar, and list views.
- Notion AI is Notion's branded AI interface, not a separate model. It supports 10+ leading models (Claude, GPT, Gemini, GLM and more), with usage unlimited on the business plan.
- Native AI connectors (Gmail, Google Calendar, Google Drive, GitHub, Slack) let you query across tools from inside Notion.
- Live demo: building a task-tracking database by prompt, then pulling outstanding tasks for an upcoming event from pages and connected tools.
Nasri
Skills beat prompt engineering
- A skill is just a Notion page of structured instructions the AI reads before acting. Instead of retyping long prompts, you tell the AI to "save this as a skill".
- Skills are shared across the team for consistent proposals, email tone, and blog format. Write the workflow once, everyone benefits.
- They live in a skills database with owner and usage metadata, embedded in each person's instruction page so the AI scans short descriptions on demand.
- Real example: the Setup Call Discovery Prep skill researches the contact, drafts discovery questions, and prepares the pre-call email in one prompt.
Dennis
Custom agents work while you sleep
Instructions
The system prompt
A page defining the agent's role and behaviour.
Connections
Your tool stack
Gmail, Slack, Drive, Figma, Attio, or any custom MCP server.
Trigger
When it runs
On a schedule or an event like a new email or tagged message.
Permissions
Least access
Fine-grained control, for example draft only, never send.
Live build: a daily 9am triage agent that drafts replies to emails needing follow-up and ignores newsletters, receipts, and FYIs. Audit every run in the Insights tab.
Nasri · Section 04
Computer use and HTML generation
- A sandboxed computer lets Notion AI generate interactive HTML artefacts, dashboards, prototypes, and PowerPoint files.
- Built live: an HTML presentation from a mock proposal, embedded straight into a page with the new /html block.
- NasriFlix and Nas Bobble: a Netflix-style portfolio and a browser game shipped from a Notion PRD.
/html
From proposal to shareable deck, embedded where the work already lives.
Dennis
Notion Workers complete the stack
- Small programs on Notion's cloud with two main uses: syncing external data into databases, and giving custom tools to agents that lack a native MCP.
- They pair with Notion's charts and dashboards for a lightweight BI stack, right where teams already work.
- They run on cheap compute credits, not AI. Syncing Google Analytics every 15 minutes costs roughly $1 a month.
- Coding agents make writing them accessible to non-engineers.
What to remember
Five takeaways
- Skills beat prompt engineering. Write once, share with the team, get consistent outputs.
- Custom agents run in the background. Trigger them, give minimum access, audit via Insights.
- Notion AI is strongest when Notion is your context layer. It is post-trained on Notion search and relationships.
- The PRD workflow avoids context degradation. Decide in one chat, distil to a PRD, build in a fresh session.
- Workers complete the stack. Cheap syncs plus charts give a BI setup inside Notion.
The Result
What the room walked away with
2 wks
Hands-on demo workspace access for every attendee
1
Live email triage agent built end to end
100%
HTML deck generated live from a proposal doc
- The audience followed along building their own agents in real time.
- Strong Q&A on custom writing styles, agent safety, second brains, and Notion AI versus Claude, all answered on the spot.