Stop fighting Notion AI’s generic responses. Learn to give it the context and persona it needs to become a genuinely powerful assistant for technical documentation, summaries, and more.

I See You Fighting Notion AI. Here’s How to Win.

I was walking past the junior engineers’ pod last week and saw something painfully familiar. One of our sharpest new hires, Maya, was staring daggers at her screen, muttering. She’d just spent an hour writing a detailed incident post-mortem for the `prod-auth-service-01` outage and asked Notion AI to “summarize this for the leadership team.” What she got back was a generic, fifth-grade book report that completely missed the point. “This thing is useless,” she said. I just smiled and said, “You’re holding it wrong. Let me show you.”

That frustration is real. You dump a ton of valuable, context-rich information into Notion, ask its shiny new AI feature for help, and it gives you back the most sterile, unhelpful drivel imaginable. It feels like a gimmick. But it’s not the AI’s fault; it’s a context problem. We, as engineers, need to treat it like any other tool: give it clear parameters and a well-defined operational scope.

The “Why”: Your AI is an Amnesiac Intern

Think of the default Notion AI as a brand-new intern with no memory and zero context about your company, your team, or your tech stack. It sees only the text you’ve highlighted or the words on the current page. It doesn’t know that “Project Chimera” is your high-priority Q3 initiative, or that a “P0” means “drop everything and fix it now.”

When you ask it to “summarize this,” it does exactly that, with the knowledge of the general internet. It doesn’t know you need the summary to be for a non-technical stakeholder or that it should be formatted as a bulleted list for a Slack update. You have to tell it. Every. Single. Time. Or do you?

Here are the three levels of telling Notion AI what to do, from a quick fix to a proper, scalable solution.

Level 1: The Quick & Dirty Fix (In-line Priming)

This is the fastest way to get a better result, and it’s perfect for one-off tasks. You simply prepend your instructions—the context and the persona—to your request in the AI prompt.

Instead of just highlighting your text and clicking “Summarize,” you invoke the AI with spacebar and write the whole instruction out.

The Bad Prompt (what Maya did):

Summarize the following text: [...pasted post-mortem notes...]

The Good Prompt (In-line Priming):

You are a Senior DevOps Engineer writing a summary for a non-technical executive leadership team. Your tone is professional, concise, and focused on business impact (customer-facing downtime, financial loss, and remediation steps). Do not use technical jargon like 'pods', 'load balancers', or 'database replicas'.

Summarize the following text: [...pasted post-mortem notes...]

It’s verbose, but it works. The difference in output quality is night and day. But copying and pasting that monster prompt every time is a massive pain.

Level 2: The “Golden Page” Fix (Centralized Context)

This is my preferred method and what I showed Maya. We create a single, dedicated page in our team’s Notion workspace that acts as a “brain” or “persona hub” for the AI. This is a game-changer.

Step 1: Create a new page. Call it something clear like AI Persona & Context Hub.

Step 2: On this page, create sections for different personas or tasks you do frequently. Here’s a stripped-down version of what our team’s page looks like:

Step 3: Now, when you want to use the AI, you just @-mention that page in your prompt. This tells the AI to read that page first for its instructions.

Using the context from @AI Persona & Context Hub, act as the Executive Summary persona and summarize my notes below.

This is powerful. You’ve now created a reusable, single source of truth for how the AI should behave for your most common tasks. When a new project or acronym comes up, you just add it to the hub page, and everyone on the team gets the benefit.

Pro Tip: Lock the `AI Persona & Context Hub` page in Notion. You don’t want someone accidentally deleting the AI’s brain during a routine documentation cleanup.

Level 3: The Architect’s Fix (Automated Context with Templates)

This is for when you want to truly scale this process. If you’re constantly creating the same *type* of document (e.g., post-mortems, new service design docs, meeting notes), you can embed the AI context directly into your Notion templates. This is the “don’t make me think” solution.

Step 1: Create a new template for, say, an “Incident Post-Mortem”.

Step 2: At the very top of the template body, create a collapsed “toggle” block or a simple text block. I call mine “AI Instructions.”

Step 3: Inside that block, put the specific persona and instructions for this document type.

--- FOR NOTION AI ---
INSTRUCTIONS: When asked to work with this document, assume the persona of a Senior Site Reliability Engineer (SRE). Your goal is to be precise, factual, and blameless. When summarizing, focus on the timeline of events, the root cause analysis (RCA), and the action items for remediation. The audience is technical. Use markdown for code snippets and tables for timelines.
---

Now, whenever someone on the team creates a new “Incident Post-Mortem” from the template, this context block is already there. All they have to do is write their content, then highlight everything (including the instruction block) and ask the AI to “Summarize” or “Find action items.” The instructions are baked right in. No @-mentioning, no copy-pasting.

It takes a bit more setup, but it ensures consistency and makes life easier for the whole team. It’s the difference between giving someone a fish and teaching them how to build an automated, AI-powered fishing trawler. And that’s what we do here.

Darian Vance

Lead Cloud Architect & DevOps Strategist

With over 12 years in system architecture and automation, Darian specializes in simplifying complex cloud infrastructures. An advocate for open-source solutions, he founded TechResolve to provide engineers with actionable, battle-tested troubleshooting guides and robust software alternatives.

🤖 Frequently Asked Questions