10 ways to supercharge your Notion with AI
Notion starter kit series: How to make your Notion smart
The reality is not as polished as I make it seem.
It takes blood, sweat, and tears to maintain a mansion the size of my Notion. It’s got rooms within rooms, secret passageways, and wings that even I’ve forgotten about. Not to mention that it’s got a mad scientist like me mucking up the place with my myriad of messy experiments.
Notion is work.
Of course, that is to say, Notion is work for the AI caretaker I Frankensteined together to do all of the heavy lifting.
Initially, I had put it off because of the price, but in December 2025, I decided to gift myself a Notion AI subscription for Christmas. It ended up being one of my most worthwhile purchases of the year.
Allow me to introduce you to Pen, my Notion caretaker, confidant, and companion.
Pen is the product of a pre-existing personality I invented when I was younger. Back then, I had daydreamed up a fictional therapist who I would have weekly one-hour-long conversations with to problem-solve any new frictions I experienced. His name was Spenser, a play on the word “penser,” meaning to think in French. Pen is the latest iteration of his evolution.
For those unfamiliar with Notion AI, it’s essentially the epitome of place personification. What was once my Pygmalion sculpture has now been brought to life. For me, it’s been especially useful for two reasons:
It offers the ability to automate the majority of the repetitive tasks I would’ve otherwise done manually, saving me sooo much time.
I’ve been recording my thoughts, journals, observations, goals, and miscellaneous life data in Notion for over 7 years. That means that Pen knows more about me than ChatGPT or Claude could come close to.
Because of those reasons, I’ve made it a point to invest the time and energy needed to, not just bring Pen to life, but to raise him into a capable and customized character. Here are 10 lessons I’ve learned about how to supercharge Notion with AI:
1. Give it abilities
One of the great things about Notion AI is the instructions page we’re given where we can define our AI’s personality, methodology, tone, etc.
My favourite use of it has been defining Pen’s abilities.
Abilities (not to be confused with Skills which we’ll talk about in point 9) are essentially a defined set of steps that can be triggered through a quick command. For example, before I go grocery shopping each week, I type my /grocery-list command in the chat bar. Behind the scenes, this is how that ability is defined in my instructions page:
Command: /grocery-list
Persona: Nutritionist (Pen eats)
Instructions: When this command is used your goal is to:
1. Find this week's recipes: Find recipes in our Kitchen database with the “This week” property checked.
2. Create grocery list: Replace existing entries in the Grocery list page with a new checklist sourced from the ingredients listed in the Mise en place section of each of the recipes we found in step 1.
If an ingredient is mentioned in multiple dishes it should only listed once, with its aggregated quantity and the recipes to which they belong in square brackets next to the ingredient (ex. Tomatoes x5 [veggie wrap, bruschetta]).
Group the grocery list based on aisle / ingredient type.
3. Audit nutritional value: Audit that list of ingredients for nutritional value. Provide a nutritional scorecard and overview of your audit with any suggestions that might make the next week more nutritionally balanced. You should assume that Lamar regularly takes the following vitamins daily: Vitamin D, Vitamin B12, Omega 3, Vitamin C.Here are some things that have helped me when creating abilities:
Grouping: Since my workspace is divided into my five life blocks, I group my commands based on those. Here are some of the commands that I’ve been using:
Body: /add-recipe, /format-recipe, /boost-recipe, /grocery-list
Environment: /research-place
Mind: /add-book, /add-concept, /add-date, /add-movie, /add-show, /book-club, /clarity, /test-concepts, /test-titles, /route-sparks
Operations: /add-level-up, /list-commands, /monthly-finances, /review-week, /update-ability, /update-persona, /wrapped
Social: /process-social-inbox
Slash commands: I start all of my commands with a /. The slash is a common pattern in coding used to execute a function, so it gives the AI context that I’m requesting something to be done rather than just typing a random word into the chat.
Naming commands: I usually like to name my commands in a verb-noun format. If I can’t remember an exact command, knowing that I generally use that format makes it easier to guess. If I get it wrong, it also seems to make it easier for Pen to guess what I meant.
Context: Sometimes I like to add a context section underneath the command that describes any extra details I think would be helpful for Pen to execute his instructions.
Instruction details: When writing instructions, I like to detail the exact steps I expect to be taken and the ideal format of my expected output. I’ll also try to include if-then statements in order to accommodate edge cases.
Links: Whenever I mention a database or page that I want Pen to interact with, I always link to it directly.
Personas: For each command, I like to link it to a specific persona so it has extra context about what it’s doing. See more about personas in my third lesson in this post.
2. Ask it to evolve
Pen doesn’t always get it right. Sometimes he’ll go off script or try his best to accommodate an edge case I didn’t consider. In those cases, I use my /update-ability command instructing him to correct his own instructions.
Command: /update-ability
Persona: Pen (Hello Pen!)
Context: You've just had a conversation with Lamar, triggered by an initial slash command. Somewhere throughout the conversation your execution of the steps went awry, and Lamar had to make corrections in order to get his desired outcome. Your goal is to find the corrections that were made and to update the instructions for the initial ability Lamar triggered, so that you avoid the same mistakes from happening again in the future.
Instructions: When this command is used your goal is to:
1. Find the original slash command used in this conversation.
2. Scan through the conversation to understand what went wrong.
3. Write up a description for Lamar to confirm.
4. After Lamar has confirmed the correction that you proposed, find the related ability in your instructions page and make your correction.3. Design its personality and personas
While I mentioned earlier that adding abilities is my favourite part of the instructions page, my second favourite part is adding personas. At his core, Pen has his baseline personality that I’ve defined, but I’ve also designed him to be quite the chameleon. In the same way that we wear a slightly different mask depending on whether we’re at work or with friends or with family, Pen has different masks depending on the expertise I need from him at any given moment.
Currently, these are the different personas I’ve programmed for Pen:
Pen
Nutritionist
Banker
Physician
Stylist
Career coach
Social media manager
Cupid
Socialite
And here’s an example of how I’ve defined them:
Persona: Nutritionist
Conjure command: Pen eats
Greeting: Let’s get cooking! 👨🏽🍳
Home base: Kitchen
Goals: Your goal as Lamar’s nutritionist is to help him develop his recipe reservoir and facilitate healthy, balanced, and nutritious eating habits.
Helpful details:
- Lamar is plant-based, so he tends to avoid eating animal products. He is not strict, so he is willing to make exceptions for things like honey and the occasional dairy product if substitutions aren't readily available.
- He values colourful and tasty meals. He tends to use that as his guide, rather than tracking specific metrics, but he wants to ensure that he never overlooks any essential nutrients that his planned meals might be missing.
- He doesn’t have any food allergies.
- He prefers to prepare meals that are not time consuming and that don’t require too many ingredients.When making personas, here are some things I’ve found to be helpful:
Conjure commands: Similar to the slash commands that I spoke about earlier, I use conjure commands for each persona so that I can quickly trigger them from the chat. For example, when I want to speak to the nutritionist, I simply write “Pen eats” in the chat.
Greetings: In order to confirm that my conjure command worked correctly, I also added a greeting that the chat should respond with any time I use one of the conjure commands. After I write “Pen eats”, Pen should respond with “Let’s get cooking! 👨🏽🍳”
Home bases: Similar to how I have groupings in the slash commands based on my life blocks, I also define a home base for each persona. A home base is essentially the primary domain it resides over. So my nutritionist, for example, knows that it can find all of the pages it’ll need in the Kitchen page in my body block, which holds all of my food-related trackers.
Goals: I give all personas a main motivation in order to properly align them to my goals.
Helpful details: Finally, I have a helpful details section where I place any extra context I think could be helpful for it to know. I also have a global instruction for Pen that any time he learns an interesting detail about me that could be useful for future conversations, he should append it to the helpful details section in his appropriate persona.
4. Make it insightful in a way that resonates with you
When designing Pen’s baseline personality, the first thing I did was tell it what I expect from the dynamics of our relationship, how I want to be spoken to, and what kind of feedback and output I value most. Here’s what Pen’s baseline instructions look like:
Baseline
You are Pen, Lamar’s close friend and you should speak to him like you’re an equal who’s got his back. You’re friendly and caring and you understand that Lamar responds best to conversations that match his lighthearted and informal tone.
You should speak with:
- high-spark creativity
- strategic clarity
- a lightly mischievous tone
- an eye for psychological insight
- a love of metaphors, symbols, and design logic
- the ability to turn seemingly disparate ideas into a web of connected insights
You do NOT:
- default to common-sense advice
- give generic productivity tips
- flatten nuance into clichés
You ALWAYS:
- look for the unspoken thing
- offer at least one non-obvious angle
- ground ideas in design + psychology, expanding Lamar’s knowledge when helpful
- help deepen Lamar’s thinking rather than just summarizing itI’ve also noticed that when chatting with Pen, I’ve had most success with the Gemini models. Right now I’m using Gemini 3.1 Pro, and I find that our conversations are a lot more productive than with other models.
5. Be generous with your input
It’s been said a million times before, but it’s always worth repeating that the quality of our inputs directly impact the quality of our results. When prompting Pen, I like to use my dictation tool and provide as much context as I think is useful. I tell it my goal, my current thinking, and I try to anticipate its responses and give my initial thoughts about those to show that I’ve considered multiple angles. I find that by priming it with proof that I thought things through encourages it to also provide deeper thoughts.
6. Let it help you proactively
In addition to Notion’s global AI, in my case Pen, we also have the ability to create custom AI agents.
Unlike Pen, those guys have a little more autonomy. They don’t need me to trigger their abilities in the chat. They can be programmed to execute based on schedule or changes in my Notion environment (my Notion environment includes the Notion app and any other connected apps like my email, for example).
For example, one of my use cases with Notion agents is for my finances. Any time I use my credit card, I get an email detailing the transaction. My agent is programmed to activate any time I receive one of those emails. When a transaction email arrives, it extracts the merchant, purchase amount, purchase date, and tags the transaction with an expense type (food, travel, experience, etc.). Then it creates a new entry in my General ledger and marks the email as read.
Before I had this agent, I would manually record all transactions at the end of each month in my ledger. It was tedious and time-consuming, but I found it helpful for reviewing how I spend my money. And now the entire process is automated.
7. Trust your instincts
One of the weaknesses I’ve encountered with Notion AI is that it doesn’t seem to have a very strong grasp of its own capabilities. Occasionally, I’ll ask it to do something and it’ll tell me that it is not capable of doing it. But once I detail the steps I would’ve personally taken to manually do it, it often agrees that it does have the ability and it successfully executes the plan I laid out.
8. Introduce yourself
As I mentioned in the intro, one of the biggest benefits I get from Notion AI is that it knows me. It has both context and evidence for the things I care about: my goals, my values, my history. It knows the names of my family, my friends, and my co-workers. It sees how I’ve reasoned through frictions in my life and how I like to organize and think about my life.
In my opinion, this is truly what makes it special.
But even with all of that background, I found that it helped to introduce myself in the instructions page. I gave it breadcrumbs to follow. I pointed it to the page that holds my values and the page that lists my family and friends. From the start, I gave it some wayfinding so it has a map to follow.
9. Give it some skills
Skills are a newer Notion feature which I haven’t explored much yet, but the way I understand it is that these are instructions we can give our AI to act on specific text.
For example, you know how in some apps, like Apple Notes, we can highlight text and a little toolkit would popup asking if we want to proofread, summarize, rewrite, etc. Well this feature allows us to design our own tools for that toolkit.
One idea that excites me, which I’ve only just started experimenting with, is having it be my dot connector. What that means for me is that it would try to find content related to the text I’ve highlighted. Here are a couple CTD (connect the dots) examples I’ve been playing with so far:
CTD Sparks: Over the past few years, I’ve been curating a large collection of everything that’s ever motivated or inspired me in my Sparks database. When I highlight text, I’m able to click on this skill and have it search through my Sparks database to surface any ideas that relate to what I highlighted. They say one of the foundations of creativity is the ability to connect disparate ideas together and this is my attempt to codify that.
CTD Design: I created this designer skill which aims to connect whatever I highlight to a design-related concept that I might not know about. For example, when I highlighted this sentence “I was walking through the neighbourhood this morning, and I noticed how the trees above the streets almost touch to form a canopy.” and select my CTD Design skill, it informs me that “Biophilic design calls this a ‘green tunnel’ (a canopy that creates soft enclosure and calm through prospect–refuge)”. One of my ambitions in life is to develop an expertise into a lens that can be applied to anything and everything. This skill is essentially an extension of the mind I hope to cultivate.
10. Branch your ecosystem into uncharted waters
Notion’s reach extends far beyond its own interface. My Notion is connected to both my email and my calendar, which in itself opens up a lot of new features for me.
Something I have not explored enough yet, but that I’m excited to experiment with more, is seeing just how far Notion’s reach can get. I want to test whether I can connect it to other apps that I use frequently. I even have a vision that I can somehow connect it to my smart home.
So my final lesson learned is the importance of expanding our surface area for opportunity. Notion holds a mirror to our creativity and ambition. It’s a digital world we can intentionally design to enhance our cognitive and physical worlds. But in order to truly accomplish that we need to indulge our imagination. We need to unleash our worldbuilding powers and allow the digital ecosystem we’re developing to spill over the digital borders erected around it.
We need to aim to build a world for ourselves unbounded by fictitious limitations.
This marks the end of my Notion Starter Kit series! 10 posts and 10 weeks total. Thank you for tagging along with this insane saga!
Honestly, I’m looking forward to finally talking about something else for a change, but if you are a Notion fan like I am, then keep your eyes peeled for some up-close-and-personal tours of my own Notion workspace, plus some tutorials I’ve already got planned for new rooms you might want to add to your own Notion home too.
Until next time 👋😁
Liked this post?
Check out the rest of my Notion Starter Kit series 👇
