Alrighty, today we’re in for a treat from my pal Isaac.

You might remember him from our piece on Buffer’s radical build-in-public playbook.

Well since then we’ve both moved from Sydney 🏝️ to London 💂, still co-work all the time, and still furiously debate whether brands should have a personified logo or not.

Over to Isaac.

— Tom

Today we’re unpacking Notion’s AI agent.

Every software company has now rolled out some kind of AI or agent features, but there’s been public push back on how they’ve been rolled out - even Duolingo’s brand was torched last year.

Yet, Notion have walked away with nothing worse than a few bug complaints, all because of some very considered and clever design choices.

The follow-up nobody sent, the signup nobody researched, the buyer signal nobody saw. Attio’s agentic CRM closes those gaps: revenue agents keep every deal moving between the moments your team is in the room.

Every customer signal becomes your next move - that's how pipelines stop leaking.

You know the drill: ideas in Slack, context buried in docs, and Jira tickets you swear you saw somewhere. That was Maple too.

See how they rolled out Jira Product Discovery to get ideas and prioritisation into one place, plus the month-by-month playbook they used to go from one team’s pilot to an org-wide standard.

_{Thank you for supporting our sponsors, who keep this newsletter free.}

The what: A TLDR explanation of the strategy

When Notion’s AI is thinking, its eyebrows do a wave. Hit an error, and the whole face comes apart, then puts itself back together.

The team nicknamed it Nosy.

Underneath that face is an agent that works across hundreds of your documents on its own, about as close as software gets to the “AI employee” everyone’s either excited about or terrified of.

In 2025, Duolingo’s CEO said the company would phase out contractors for work AI could handle, on top of cutting 10% of staff. Users who’d held 100-day streaks ended them in protest. Duolingo lost 400,000 TikTok followers in a few weeks and wiped its own accounts.

Notion released something with a similar risk profile: an agent working autonomously across your work. It got criticised for being slow, for hallucinating, and for chasing AI while the basics were still broken.

Software complaints. Not a revolt.

The difference wasn’t the technology. It was a year Notion spent designing a face before anyone saw the agent behind it.

The how: The strategic playbook boiled down to 3x key takeaways

Open any AI feature in any product and you’ll see a sparkle icon. It’s shorthand for magic. That’s exactly the problem. Magic means you don’t need to understand it. And when every competitor reaches for the same icon, the icon stops telling you anything about who made it.

Notion’s team named the problem directly. Design lead Ryo Lu said they needed “to do something that differentiate ourselves from the sparkles in the world.” Colleague Linus Lee went further:

Early concepts leant into the magic anyway: wizards in different hats for different tasks and a wand that turned into a pencil or a brush depending on the job. All of it got thrown out, because introducing magic didn’t make the product any easier to understand.

Everyone else designs for how advanced the technology is. Notion designed for how much of it you can understand.

In September 2024, Notion’s Head of Creative Rob Giampietro said the quiet part out loud: an agentic future was coming, but “right now, I think we have to make people comfortable with being helped [by AI]… that’s the first step.”

The face designs shipped that month. The agents themselves didn’t arrive for another year.

The character is deliberately minimal: two eyes, a pair of eyebrows, and a nose. “The nose is almost like a baton that conducts your eye,” Giampietro said. Processing shows up as a wave of the eyebrows. An error makes the whole face fall apart, then reassemble.

None of this is decoration. Notion asked Steven Sinofsky, who ran the Office team that shipped Clippy, whether he’d give an assistant a face again. He pointed to a Stanford study, The Media Equation, which found people generally treat technology as “right” by default. Give them something softer, and “they stop blaming themselves for mistakes and are more open to questioning back and forth.”

Notion pushed the idea further than a static drawing: you can rename your agent and dress it in accessories. A character you customised yourself is hard to be afraid of.

Notion never says “AI employee”. You won’t find the word “headcount” anywhere near the feature. They won’t even call it an “assistant”, the same instinct that has Apple refusing to call the Vision Pro a “VR headset”. Name the category, and you import every fear that comes with it.

Cofounder Ivan Zhao said he wanted the product to feel less like productivity software and more like playing Tears of the Kingdom: easy to build with, endless complexity hidden underneath. Hide the difficulty. Keep the user inside the work.

Compare that to Duolingo, which said the quiet part out loud: AI would replace contractors. It paid for that in cancelled streaks and 400,000 lost followers.

Same underlying technology. Opposite framing. Opposite outcome.

Every AI agent asks the same question of the person using it: what is this doing, and what does it mean for me? A sparkle answers with mystery. A pair of eyebrows answers with a colleague, thinking it through alongside you.

That’s all from me!

If you want more of this, brand strategy explained through the actual decisions companies made, subscribe to Brand Chemistry here and follow me on Linkedin and X.

And if you’re a B2B brand trying to earn the kind of trust Notion just did, that’s the work we do at Pistachio. You can book time to chat with me here.

The where: 3x high-signal resources to learn more

The raw, behind-the-scenes scrapbook behind the Strategy Breakdowns rebrand.

→ The initial moodboard and vibe sliders

→ The concepts that got thrown away

→ The iteration loop with Jade (product designer @ Atlassian) who brought the brand to life

Notion’s own AI face interviews Steven Sinofsky - the exec who shipped Clippy - on the Stanford research behind friendly interfaces, and why Clippy was “early and wrong”.

The most interesting part: Why softer interfaces make users question the interaction instead of blaming themselves

The full cultural history:

Also contains illustrator Kevan Atteberry’s original designs - super cool!

💬 Do you like reading guest posts? Why or why not? Reply to this email and let us know!

Our flagship course on how to use free internet data to make better strategic decisions. Contains 5 years of strategy expertise, proven methods, and actionable tactics to accelerate your career with modern-day strategy skills.

We have a growing audience of 150,000+ strategists from top companies like Google, Meta, Atlassian, Stripe, and Netflix. Apply to feature your business in front of Strategy Breakdowns readers.

One of the most common questions we get asked is: “What tools do you use to run Strategy Breakdowns?” So, we’ve open-sourced our tech stack to give you an inside-look at exactly what tools we’re using to power each corner of this operation.