1.4Notion AI & relations between databases
The essentials in 30 seconds
- Relation = the link: a Relation property links two databases (Tasks ↔ Projects) — without duplicating information.
- Rollup = the calculation: it aggregates the linked info (counting tasks, summing amounts, min/max date) — no relation, no rollup.
- Notion AI = the assistant: summarises, generates, rephrases, fills properties — a paid option beyond the trial, to evaluate.
Running mini-case — Lina links everything in Oran
Lina now manages two databases: «Projects» and «Tasks». She adds a Relation property to Tasks pointing to Projects and links each task to its project; each project then displays its tasks. She adds two rollups on Projects — «Number of tasks» (count) and «Completed tasks» (filtered count) — and tests Notion AI to summarise her long meeting notes and automatically categorise her tasks. Her isolated tables have become an information system.
Final episode of cases 1.1-1.3 — Lina assembles her relational system.
1. Notion AI: the built-in assistant
Notion AI is a built-in AI: it summarises a page, generates a draft, rephrases, translates, extracts actions from meeting notes, or automatically fills properties (summary, categorisation). It is invoked while writing, without leaving the tool — a real time-saver for content production and processing.
Lina summarises her three-page meeting notes in one sentence and lets the AI categorise her tasks: she rereads and corrects, the AI suggests, she decides.
2. Relations: link instead of duplicating
Above all, you link two databases like in a real relational database: Projects ↔ Tasks, Clients ↔ Orders. A task «belongs» to a project, and the project «sees» all its tasks. You add a Relation property on one side, and the other side shows the reciprocal.
Lina links each task to its project: opening a project immediately lists all its tasks — no information copied, no possible divergence.
3. Rollups: aggregating the linked information
You add rollups to aggregate the linked information (number of tasks in a project, sum of amounts, nearest due date). The relation is the pipe, the rollup is what flows through it and gets calculated: no relation, no rollup. This is what turns isolated tables into a coherent information system — the gateway to Airtable and automation.
Lina reads on each project its self-calculated «Number of tasks» and «Completed tasks»: her steering is up to date without her touching a figure.
🔎 Concrete example — 🔎 Concrete example — Projects ↔ Tasks
A mini relational system.
Tasks.Project → Relation to Projects
Projects.«Task count» → Rollup (count)
Projects.«Completed tasks» → Rollup (filtered count)
→ Notion AI: automatically categorise each task.Reading: each of Lina's projects shows the aggregated state of its tasks.
Lesson: link, never duplicate information.
Section vocabulary
- Notion AI
- AI built into Notion: summarises, generates, rephrases, translates, extracts actions, or fills database properties.
Simple example: The page summary and auto-categorisation of Lina's tasks.
What it is not: external automation via Make/n8n (more flexible, level 4).
Reflex: the AI suggests, I reread and approve.
- Relation
- A link between two databases (Projects ↔ Tasks): a record on one side points to one or more on the other.
Simple example: Each of Lina's tasks attached to its project.
What it is not: the project name copied as text in each task.
Reflex: link, never copy.
- Rollup
- A property aggregating information from linked records (count, sum, min/max date…).
Simple example: The «Number of tasks» calculated on each of Lina's projects.
What it is not: a calculation set with no relation (impossible).
Reflex: relation first, rollup second.
- Relational database
- A set of databases linked together forming a coherent system, with no duplicated information.
Simple example: Lina's Projects ↔ Tasks, with progress rollups.
What it is not: isolated tables copied into one another.
Reflex: one place per fact, links everywhere.
- AI-generated property
- A database field filled automatically by Notion AI (summary, category, key points).
Simple example: The auto category of each of Lina's tasks.
What it is not: systematic manual entry when the AI does just as well.
Reflex: test on the trial, cost it before paying.
In practice — In practice — Linking databases and using Notion AI
- Create two databases: Projects and Tasks. Add a Relation property to Tasks pointing to Projects.
- Link each task to its project; open a project and notice that it lists its tasks.
- Add a rollup on Projects: «Number of tasks» (count) and «Completed tasks» (filtered count).
- Test Notion AI: ask for a summary of a notes page, or an AI property that automatically categorises your tasks.
Ready-to-say script — explaining relation vs rollup (30 s)
«The relation is the pipe: it says this task belongs to this project. The rollup is what flows through it: it counts the project's tasks or sums the amounts. So we create the link first, the calculation second — no relation, no rollup.»
To say to a colleague who wants to «add a rollup» with no relation.
Copy-paste template — relational plan
DATABASE A: _______________ (e.g. Projects)
DATABASE B: _______________ (e.g. Tasks)
RELATION: B._______________ → to A [ ]
RECIPROCAL CHECKED: A lists its B [ ]
ROLLUP 1: A.« _______________ » → _______________ (count / sum / min-max)
ROLLUP 2: A.« _______________ » → _______________ (filtered count)
NOTION AI TESTED: summary [ ] | categorisation [ ] | cost evaluated [ ]Two lists to connect
Identify what «belongs» to what (task → project, order → client), create the relation in that direction, check the reciprocal — like Lina with Projects ↔ Tasks.
Steering in Oran
Lina reads progress on the rollups («Number of tasks», «Completed tasks») in the weekly meeting: figures always up to date, zero manual updates.
Tight AI budget
Stay on the trial quota to learn; for intensive use, cost the time saved vs the add-on — or prepare the AI call via Make/n8n (level 4).
Towards Airtable
This pattern (relation = link, rollup = aggregation) is exactly that of Airtable linked records + lookups/rollups: mastered here, transferred at level 2.
✓ Best practices
- Link rather than repeat.
- Rollups for auto aggregation.
- Notion AI: test, then cost.
✗ Common pitfalls
- Duplicating instead of linking.
- Rollup with no relation (impossible).
- Paying for AI without measuring the gain.
Key takeaways
- Notion AI summarises, generates, rephrases, translates and can fill properties automatically.
- Relations link two databases (Projects ↔ Tasks) without duplicating information.
- Rollups aggregate the linked information (count, sum, min/max date).
- Relations + rollups turn isolated tables into a coherent information system.
48h challenge — apply it tomorrow
Create the Tasks → Projects relation (template above), add 1 rollup («Number of tasks») and test 1 Notion AI summary. 30 minutes is enough.
- I can explain relation (pipe) vs rollup (calculation).
- I know Notion AI is paid beyond the trial.
- I linked 2 databases + 1 rollup + tested the AI.
Frequently asked questions
What is the difference between a relation and a rollup?
The relation creates the link between two databases (it says «this task belongs to this project»). The rollup exploits this link to bring back and aggregate information from the other side («how many tasks does this project have?», «what is the sum of the amounts of these orders?»). In other words: the relation is the pipe, the rollup is what flows through it and gets calculated. No relation, no rollup possible. It is exactly the same logic as Airtable linked records + lookups/rollups (level 2): mastering it here prepares what comes next.
Is Notion AI free?
Notion AI is generally a paid option (add-on) beyond a trial quota, billed on top of the subscription. For learning, the trial is enough to understand the value; for intensive use (daily content processing), you must assess whether the time saved justifies the cost. Good to know: in an advanced no-code workflow (level 4), you can also call AIs (ChatGPT, Claude, Gemini) from Make/n8n to process Notion data, sometimes more flexibly and with cost control. Notion AI is the simplest way; external automation is the most flexible.
What is a relation between two databases for?