Notion – Outside-In Analysis

2-Minute Brief

Three arguments run through this piece. Each is a conclusion, not a fact – I show the evidence, and you can disagree.

Argument 01

Revenue roughly doubled. The valuation didn't.

Estimated ARR went from about $300M to about $600M while the headline mark moved from $10B (2021) to $11B (2026) – and that $11B is a secondary employee-share sale, not a fresh round. A great product, growing fast, still marked flat. That gap is the story.

Argument 02

The most important number about Notion is one it has never disclosed.

Revenue, 100M+ registered users, half of paying customers using AI, and 90% of the business coming from teams – all public. Daily or monthly active users, and a paid-customer count, are not. That gap is what this piece sits on.

Argument 03

Notion is racing to become the system agents write to.

The May 2026 Developer Platform is a bet that the workspace becomes agent infrastructure. It's the right bet, and named customers are already running agents in production. What's unproven is the economics: Workers don't start billing until October 15, 2026, and there's still no paid third-party marketplace.

⚑ Primary Observation

A beloved product, a flat mark

Here is the tension in one question. If Notion is as loved and as widely used as its story suggests, why hasn't the valuation moved? Part of the answer is that no one has set a fresh mark – there has been no primary round since 2021, and the only recent print is a secondary sale of employee shares. So the flatness is an open question, not a verdict: either the number would jump if Notion raised, or the paid-usage the headline hides would hold it back. I can't resolve it from the outside. But I can show both sides honestly.

✓ Evidence it compounds

The case that Notion is winning

Notion told CNBC it crossed $500M in annualized revenue in September 2025 (the confirmed figure); third-party trackers estimate it near $600M by late 2025, up from roughly $300M in 2024.
The share of paying customers using AI went from 10–20% in 2024 to over 50% by September 2025 – fast adoption of a paid capability.
90% of the business comes from multiplayer use – teams, not individuals. That's a strong enterprise signal.
More than half of the Fortune 500 have a presence on Notion; profitable since around 2023; roughly 80% of users outside the US.
Product velocity is real: 3.0 Agents (Sep 2025), Custom Agents (Feb 2026), the Developer Platform (May 2026) – all shipped, not promised.

△ What stays undisclosed

The case for waiting

No daily or monthly active-user figure has ever been published. Not in press, not on the site, not in four founder interviews.
No absolute paid-customer count has ever been disclosed. We know 90% of the business is teams and 50%+ pay for AI – but not how many customers that is.
The headline valuation is essentially flat over four-plus years – though the only recent print is a secondary sale of employee shares, not a fresh round, so it's a soft signal rather than a clean market verdict. The whole category also repriced below its 2021 peaks.
A generous free tier likely inflates the registered-user number relative to people who actually pay or depend on it.
Notion's own "State of AI at Work" survey concedes the depth gap: 88% of organizations still use AI mainly as an assistant, and only 2% operate at the most advanced level. Its president, Akshay Kothari, put it plainly this year – most people are "still using it primarily as search."

Context: Private companies disclose selectively, and Notion's discipline here is deliberate, not evasive – it publishes the numbers that flatter the enterprise story and withholds the ones that would let outsiders model retention. That's normal. It just means the adoption question stays open – and with no fresh primary round since 2021, the flat mark is more an absence of a new data point than a verdict either way.

By the Numbers

The numbers Notion shares

Everything below is public and sourced. Notion is private, so most financials are third-party estimates; I've labelled what's estimated and led with what Notion itself has stated. I also flag, plainly, the numbers that don't exist.

~$600M Estimated ARR, late 2025

Up from ~$300M in 2024. About half from AI-enabled customers – see the caveat below.

100M+ Registered users

Notion-stated, crossed ~Sep 2024. Registrations, not active users.

50%+ Paying customers using AI

Notion-stated (CNBC, Sep 2025); 10–20% back in 2024.

$10B → $11B Valuation, 2021 to 2026

Series C ~$10B (2021); secondary tender ~$11B (Jan 2026).

Funding

Roughly $343M in primary equity through Series C: Series A $18.2M (2019), Series B $50M at $2B (2020), Series C $275M at ~$10B (2021, Coatue and Sequoia). No primary capital has been raised since.

The January 2026 tender

A ~$270M secondary sale at an $11B valuation (GIC, Sequoia, Index). This is not new money for the company – investors bought existing employee shares. It sets a price; it doesn't fund operations.

Reach

More than half of the Fortune 500 have a presence on Notion, and around 80% of the Fortune 100; roughly 80% of users are outside the US. Footprint is not the same as a signed enterprise contract – a handful of seats counts.

One number worth reading slowly: Notion says about half its recurring revenue now comes from "AI-enabled customers." That sounds like AI drives half the business. It means half the revenue sits on the Business and Enterprise plans, which bundle AI in by default – which plan people are on, not whether AI is why they pay. The genuinely AI-driven revenue isn't broken out.

Who uses it – from Notion's customer pages

OpenAI

Figma

Ramp

Nvidia

Volvo

Vercel

Amazon

Toyota

How OpenAI actually uses it

Research, engineering, product, data science, and go-to-market all run on Notion – experiment tracking, technical docs, automated reporting (the data-science team saves over an hour a week), and a GTM hub that organizes hundreds of pieces of content. It connects to GitHub, Linear, Slack, and Drive.

In a customer's words

Notion has become an integral part of our entire research flow and product development process.

Karina Nguyen, Research, OpenAI

This is the product working exactly as pitched – deep, load-bearing, hard to rip out – at a flagship customer. The honest question the brief keeps returning to is whether that depth reaches the median team, or stops with the power users. The numbers that would answer it are the ones Notion doesn't publish.

The agents are already running somewhere

Read the customer stories and the agent bet stops looking theoretical. These are Notion's own case studies – hand-picked, self-reported, and no substitute for a disclosed retention number – but they are named companies putting specific numbers to production agent use, which is more than most of the category can show.

Ramp

300+ active custom agents and a 70% cut in productivity-tool spend, with Notion used as the system of record. Built inside a year.

Rakuten France

80+ custom agents and 58,000+ workflows automated – built by non-technical staff rather than the four-person AI team that used to be the bottleneck.

Vercel

Launches run on Notion end to end: 35% faster shipping, 89% higher shipping confidence, and up to nine hours a week reclaimed per person.

The pattern in every one of these is the same, and it's the interesting part: the agents get built by people who can't code. Clay rolled out 80+ agents in a week, and its head of education describes connecting Slack, writing a prompt, and testing results in five minutes. That is Notion doing what it has always done best – handing ordinary people a builder's tool – applied to automation instead of databases. If it generalises beyond the design-literate startups that already love Notion, it's the strongest answer the company has to the dependency question. Whether it generalises is precisely what isn't published.

My Read · What Draws Me to Notion

Why I keep coming back to Notion

I write about Notion because it's genuinely interesting to me, not because I've judged it. Here is what draws me in, where I stay skeptical, and what would actually change my mind.

✓ What draws me in

The comeback and the craft

The story is a genuine comeback: deleted three years of work, rebuilt in Kyoto, survived a near-death moment when a single database almost filled up, and came back to 100M+ users.
The way Ivan runs it is unusual and I like it. "Jazz band, not a marching band." No product plan, only a financial one. Hire for taste and will over experience. 50–60 founders on the staff to keep the company from calcifying.
The craft is real. Notion is one of the few tools people bend to fit exactly how they think, and defend like a personal space.
The motion to serve more use cases – docs to databases to projects to agents – is the right direction if the pieces ever compound.

△ Where I stay skeptical

The dependency I can't confirm

Notion pitches itself as the place your work lives – the single source of truth. I've never been sure it becomes that for most people. I build a system, admire it, then drift back to other tools for the daily truth. The question isn't frequency – a system of record shouldn't ping you – it's whether work genuinely depends on it. That's the number Notion won't show.
The missing engagement number bothers me precisely because everything else is disclosed. When a company shares revenue and attach rates but not active usage, I assume the usage number is the less flattering one.
The agent-infrastructure pivot is the right idea, but positioning is not proof, and I've seen "platform" narratives used to justify a flat valuation before.

What would change my mind: one disclosed active-usage or retention number that's strong, and evidence the Developer Platform earns real third-party revenue once Workers start billing on October 15, 2026. Show me people live inside it, and show me agents pay to operate on it, and the flat mark starts to look like the cheap moment.

Product Quality Audit · Public Surface Only

Loved for craft. Ambiguous about what it replaces.

This is a public-surface read – the app stores, the review platforms, and my own use of Notion over the years. The pattern is consistent. People love the craft and the flexibility, and they struggle with the blank page and, increasingly, with predicting what AI will cost them.

Radar – 9 Dimensions (scored /10)

Scorecard

Craft & polishi

9

Flexibility & poweri

9

Collaborationi

8

AI capabilityi

7

Onboardingi

6

Mobile experiencei

6

Scope clarity / positioningi

5

Pricing clarity ⚑i

4

Cost predictability ⚑i

4

Scores are editorial. Notion earns its reputation on craft, flexibility, and collaboration. It loses points where it always has – a blank canvas that overwhelms new users – and on a newer problem: AI billed by metered credits, which makes the monthly cost genuinely hard to predict from the pricing page alone.

How Notion Charges for AI

The two red scores, in one picture

Notion's AI is sold two different ways, and the line between them is where a buyer loses the ability to forecast the bill. Everything here is from Notion's own pricing and help pages.

Free / Plus$0 / $10 per seat

Trial AI capabilities only – enough to see what it does, not enough to run on it.

Predictable

Business / Enterprise$20 per seat / custom

Notion AI included by default – the agent, meeting notes, enterprise search. This is the tier the "about half of revenue comes from AI-enabled customers" figure actually describes: it counts who is on the plan, not who uses the AI.

Predictable

Below this line the bill stops being a seat count

Custom Agentsbilling since 4 May 2026

$10 per 1,000 monthly credits, sold only as an add-on to Business and Enterprise. Credits reset monthly and do not roll over; if you run out, agents pause at the next service date.

Variable

Workerscredits from 15 Oct 2026

Free through the beta, then drawing on the same credits – roughly $0.0023 a run, about 4,348 runs per 1,000 credits.

Variable

Third-party marketplace–

Doesn't exist yet. There is no paid ecosystem where outside developers sell agents that run on Notion, which is the piece the platform bet ultimately needs.

Not built

The whole pricing risk in this piece sits at that middle line. Above it, a finance team can forecast Notion to the dollar. Below it, the cost of the exact behaviour Notion most wants to encourage – handing agents real work – is the part nobody can predict in advance.

The 10-Star Experience Test · The core workspace journey

Notion sits at roughly 4 stars on the journey that matters most.

Design the impossible experience first, then see where the product actually lands. The journey: you arrive with a messy problem and want a workspace that just works.

10 ★

The workspace builds itself. You describe your team and your work in a sentence, and Notion assembles the docs, databases, dashboards, and agents – already populated, already connected, kept current by agents without you touching it.

7 ★

It meets you where you are. It reads what you already use, proposes the right structure, and an agent keeps the wiki and the project tracker updated on its own. You approve; you don't build.

5 ★

A setup that fits on day one. You pick a use-case and it feels made for you – routed by role, connected across docs and databases, useful in minutes rather than an afternoon of configuring.

~4 ★
Notion now

Powerful, but you assemble it. Today Notion hands you the best building blocks in the category and, mostly, a blank canvas. Power users build something extraordinary. New users face an empty page and a hundred templates. The craft is a 9; the "what do I do first" is a 4.

3 ★

A document with a database bolted on. The category baseline: notes, tables, some sharing, no memory of who you are or what you're trying to do.

1 ★

Ten tools that don't talk. The founding problem: knowledge scattered across a wiki, a tracker, a doc, a chat, and a spreadsheet, none aware of each other. Notion clearly cleared this bar.

Product × Persona Coverage

Deep for teams and builders. Thin for the solo newcomer.

Six product surfaces against four personas. The strong column is teams; the thin one is the solo newcomer who has to build everything before the product pays off.

Persona

Docs & Wiki

Databases

Projects

AI & Agents

Calendar

Dev Platform

Solo creator
Student, individual

●

△

○

Startup team
Ops, all-in-one

●

△

Enterprise worker
Fortune 500 seat

●

△

●

△

Developer / builder
Agent, integration

△

●

○

●

○

●

● Strong fit △ Partial / not designed for ○ Not addressed

Rating Benchmark · G2, the common surface (August 2026)

Rated with the best in its category

Notion's rivals are all web-first, and G2 is where they meet on equal terms. Notion holds 4.6 on the second-largest review base in the group – essentially tied with Airtable, ClickUp and Coda, and a hair behind Monday's 4.7. This is a category of genuinely well-liked products, and Notion is firmly in the top tier.

G2 review counts: monday.com ~14,800, Notion 11,962, ClickUp ~12,200, Airtable 3,179, Coda 490.

App Store (iOS)

4.8★

~90,000 ratings

Google Play (Android)

4.6★

381,000 ratings on 10M+ downloads – its largest sample, still strong

G2 / Capterra

4.6★ / 4.7★

11,962 G2 reviews

What users actually say · themes, not verbatim

Praise – craft and flexibility

Beautiful, endlessly flexible, and the one tool people bend to fit exactly how they work – reviewers routinely say it replaced several other apps.

Complaint – the blank-page wall

New users describe being overwhelmed: powerful, but you have to build everything yourself before it pays off, and the learning curve is steep.

Praise – it becomes the team's brain

Teams describe Notion as the shared source of truth – wiki, projects, and docs in one place – and say the collaboration is what keeps them from leaving.

Complaint – performance and AI cost

Recurring gripes: it can feel slow on large workspaces and on mobile, and the newer AI credits make the monthly bill hard to predict.

The interpretation: The reviews split cleanly. Everything people love is about craft, flexibility, and being the team's shared brain. Everything they struggle with is the cost of that flexibility – a blank page for newcomers – and, newly, the unpredictability of metered AI. Value is delivered through the product, not through people. That's a healthier signal than the opposite. It still doesn't tell me whether work truly depends on it.

First-hand observation: Using Notion over the years, the pattern holds for me too. When I invest a weekend building a system, it's the best in the category. When I want something useful in five minutes, I still stall on the empty page. And I've never been sure my work truly lives here – I build systems I admire, then reach for other tools when I need the fast, reliable source of truth. Whether Notion becomes the thing you depend on, or the thing you set up and drift from, is the question its numbers won't answer.

Journey Map

Two versions of Notion

The Could-Be is an integrated, agent-operated workspace that maintains itself. Today's Notion is powerful but human-maintained – and it leaks exactly where that maintenance burden falls on the user.

Solo newcomer

Sign up free

→

Beautiful blank page

→

Hunt for a template

✗

Overwhelmed, drifts off

Startup team

One shared doc

→

Whole team joins

→

Sprawl grows

✗

No one maintains it

Enterprise worker

Search for an answer

→

Finds the doc

→

Doc is stale

✗

Trust erodes

Developer / builder

Wire up MCP

→

Agent reads Notion

→

Hits alpha API limits

✗

Waits for GA + pricing

The problem: Every lane breaks at the same place – maintenance. Notion gives you the best blocks in the category and then asks you to keep the whole thing current by hand. The agent bet is precisely an attempt to remove that burden, which is why it matters so much.

Describe your work

→

Agents build the workspace

→

Agents keep it current

→

You supervise, not maintain

↓ Every doc an agent writes becomes context the next agent uses

↓ The system of record improves as work happens, not as you tidy it

↓ Third-party agents plug in and extend it without leaving Notion

Reliable orchestration

Agents that run long, multi-step work across hundreds of pages without supervision are still early – the External Agents API is in preview, rolling out over the coming months.

A monetized ecosystem

The Developer Platform has no paid third-party marketplace yet. Custom Agents began drawing paid credits on May 4, 2026; Workers start on October 15, 2026. Positioning is not yet proof.

Predictable economics

Agent work runs on metered credits. Until cost is predictable, teams won't hand agents the run of their workspace.

The Unlock – the whole bull case in one paragraph

If the agent orchestration becomes reliable and the ecosystem actually monetizes, Notion stops being a place you maintain and becomes the system agents operate on. That is the version that re-rates the company – from a workspace priced flat to infrastructure priced on a multiple. Everything else in this piece is a question about whether it gets there.

The Lifecycle · Strong and Weak Spots

Strong at the edges, uncertain in the middle

Trace a user from sign-up to dependence and Notion is strongest at both ends, weakest exactly where the money is. Signing up is frictionless and the free tier is generous. Team adoption is the real strength – 90% of the business is multiplayer, and once a team is in, it stays. Those two facts are why 100 million people have registered.

The middle is where it strains. Activation runs into the blank page: many sign up, fewer build something that survives the first session. The same generous free tier that wins adoption delays the paid moment. And the newest step is the most fragile – flat per-seat pricing gives way to metered AI credits right at the point where a customer is deciding how deeply to commit, which is the least predictable and most churn-prone place to introduce a variable bill.

Between adoption and payment sits the step nothing is published about: whether work genuinely depends on Notion. No retention, no active-use, no paid-seat count. That single gap is what makes the rest of this lifecycle a hypothesis rather than a measurement – mine, from using the product and reading the review themes, not from internal data.

Next 24 Months

What could break the story

Six risks, each with the consequence that follows. I mark whether each is mostly internal – something Notion controls – or external, something the market does to it.

01 · Internal

The engagement no one can see

High

Notion has never disclosed daily/monthly active users, retention, or a paid-customer count. Consequence: no outsider – and maybe no acquirer – can verify how sticky it really is, which caps the multiple until the number is shown or inferred from a data room.

02 · External

A flat mark, and what it costs internally

Medium

Revenue roughly doubled while the headline valuation went $10B to $11B – and that $11B is a secondary employee-share sale, not a fresh round. Whatever the cause, the practical consequence is the same: employees hold options priced against a mark that isn't rising, and each year of it makes retention and hiring harder against rivals whose paper is still climbing.

03 · Internal

Seat pricing turning into a meter

High

Below the line in the pricing ladder above, buyers who could forecast a flat per-seat bill no longer can – and unpredictable cost is one of the most reliable churn triggers in software. It is already surfacing in Notion's own reply threads: in late July 2026 an annual Business customer called mid-term credit limits "a breach of trust," and another said that after repeated testing he "still cannot determine the actual pricing." Several add they would rather use the ChatGPT or Claude subscription they already have.

04 · Internal

Margins by choice

Medium

Ivan has said he's willing to let gross margins get "radically worse" to win the AI bet. Consequence: if AI costs outrun the credit revenue, a profitable company becomes a less profitable one right as it's trying to justify a growth multiple.

05 · External

The bundlers are already inside the building

High

Microsoft bundles Loop and Copilot Notebooks free inside Microsoft 365; Google put Gemini into Workspace by default and raised prices 16–22% to pay for it. The point: for the average enterprise buyer, a workspace tool that's already included in a suite they pay for beats a better standalone product that costs extra. Those are exactly the seats Notion is trying to win.

06 · External

Chat surfaces eating from above

High

ChatGPT and Claude keep accreting workspace primitives – projects, memory, canvas, connectors. Consequence: if the place people start their work becomes the chatbox, the "one tool for everything" premise erodes from the top, and Notion becomes a back-end others front.

Competitive Intelligence

Everyone refounded around agents

Dozens of tools compete with Notion; four threads matter. The whole category repriced hard below its 2021 peaks – Airtable's $11B-to-$1.3B sale in August 2026 is the sharpest example, and the context for Notion's own flat valuation. ChatGPT and Claude loom over all of it, but they compete for where work starts rather than head-to-head, so I treat them as a risk, not a rival.

The Bundlers

Microsoft + Google · distribution beats features

BundledFree inside M365 & Workspace

DefaultGemini in Workspace

+16–22%Workspace price rise 2025

LiveLoop shipping in M365

Microsoft Loop plus Copilot Notebooks is a direct Notion-doc analog, bundled free into suites enterprises already buy. Google bundled Gemini into Workspace by default and raised prices. Neither has to be better than Notion to win the seat.

They don't win on craft. They win on the invoice the company already signed.

Granola

Not a head-to-head rival · a claim on the same context layer

$192MTotal raised

$125MSeries C, Mar 2026

$1.5BValuation

Notes→ enterprise AI app

Be precise about this one: Granola's homepage still sells "the AI notepad for back-to-back meetings," and nobody is choosing between it and a workspace. The threat isn't the product, it's the strategy behind the March 2026 round – TechCrunch framed it as expanding from notetaker to enterprise AI app, and Granola now pitches itself as the context layer other AI tools read from, with an enterprise API and MCP piping meeting data into Claude and ChatGPT.

If agents end up reading a company's context from the meeting record rather than the workspace, Notion's "system agents write to" thesis loses its foundation – to a product that never had to win the seat.

Monday · ClickUp

The independent work-OS rivals · refounded around agents

$1.47BMonday FY26 revenue guide

PublicMonday on Nasdaq (MNDY)

Brain2ClickUp's agent layer

4.6–4.7Rated with Notion on G2

The nearest still-independent head-to-head rivals both rebuilt around AI: Monday – the profitable public benchmark – repositioned as an AI work platform, and ClickUp shipped its Brain2 agent layer. They review as well as Notion and chase the same teams.

Notion's edge here is craft and coherence, not features – each of these can tick the same boxes on a procurement sheet.

Coda + Airtable

The cautionary tale · rivals absorbed at a discount

~$1.4BCoda's peak → Grammarly, undisclosed

Airtable→ Bending Spoons, Aug 2026

$11B → ~$1.3BAirtable's fall from its 2021 peak

AbsorbedBoth no longer independent

Notion's two closest philosophical rivals both fell from grace. Grammarly bought Coda – last valued around $1.4B – for undisclosed terms (now "Superhuman Docs"); and on August 4, 2026, Airtable, worth more than $11B in 2021, agreed to sell to Bending Spoons for roughly $1.3B, a firm known for buying mature software and running it for cash. Two "refoundings" that ended in a sale.

This is the exit path made concrete – twice – and the reason Notion's own exit scenario is not far-fetched.

The Moat, Honestly

Loyalty and craft on one side, low switching cost on the other

Notion's strength is the template and creator community around it, and the fact that it's becoming the shared data layer agents read and write to – plus a level of craft and brand loyalty rivals can't fake. Its weakness is switching cost. For the median enterprise seat, an AI assistant bundled free into a suite they already pay for is "good enough," and the "one tool for everything" promise is exactly what chat surfaces and AI-native tools attack. The Developer Platform is the attempt to turn that loyalty into lock-in by making Notion the place agents operate. It's the right move. It isn't proven yet.

Product Org & Culture

Founder-led by conviction – and the scaling question that comes with it

Notion is one of the most distinctly founder-led products in software. That's its strength and, at around 1,000 people racing into an AI transition, its open question. This is a read on the company's product org from the outside – not a comment on any role.

Framing note: This is external analysis of how Notion appears to be run, drawn from Ivan Zhao's interviews and public signals. Leadership has context no outside read can replicate.

The strength: coherence

Ivan Zhao has "refounded" Notion twice – restarted it outright, not adjusted it – and kept the company from settling in between by absorbing other founders' teams – 50 to 60 founders and ex-founders now sit on the staff. He runs it like a "jazz band, not a marching band": no fixed product plan, only a financial one, because the technology changes week to week, and he hires for taste and will over experience.

This is why the product feels like one person's taste at scale. AI is driven by co-founder Simon Last; there's no separate "Head of AI" figurehead – the founders own it. And the newest acquisition is the revealing one: Notion is buying the retrieval layer rather than renting it, which is the unglamorous thing that has to work before agents can be trusted on a messy workspace.

The question: does taste scale?

A jazz band works because everyone can still hear the leader. At around 1,000 people, with an enterprise sales motion to build and internal agents now doing a growing share of the work, the same intuition-led, plan-light style that made Notion great is harder to hold together. Enterprises signing six-figure contracts want roadmaps, not improvisation.

The next phase has to reconcile two things: keep the founder-taste that is the whole brand, and add the predictability – of roadmap, of pricing, of the numbers it discloses – that enterprise buyers and public markets expect. That reconciliation, not any single competitor, is the hardest internal problem.

Eleven Years of Not Settling

Two refoundings, five teams absorbed

"Refounding" is Ivan's own word for it: deliberately restarting a company that has drifted or been overtaken, and accepting the damage to the near-term numbers. He borrows the idea from artists who reinvent themselves rather than repeat themselves. It is a stronger claim than a pivot – a pivot changes direction, a refounding starts over with the same people.

2015 · Kyoto

Refounding one

Deletes three years of work, lays off the small team, and rebuilds from scratch with co-founder Simon Last.

2021

Automate.io

Integrations.

2022

Cron, then Flowdash

Cron becomes Notion Calendar.

Nov 2022

Ships AI writing – weeks before ChatGPT

Built on early GPT-4 access, prototyped at a company retreat. The whole AI era arrives days later.

2023 · Cancún

Refounding two

At an offsite, turns a SaaS company into an AI company – the decision the rest of this piece is about.

Feb 2024

Skiff

Becomes Notion Mail – which Notion then shut down in 2026 to go all in on agents.

Jul 2026

ZeroEntropy

A Y Combinator retrieval startup, $4.2M raised, whole team in. Its reranker had already been running Notion AI's search for two months, cutting latency 85%. Its co-founder now leads Model Research.

● Refoundings● Teams absorbed

Glassdoor – culture read (directional)

Talent signal

Strong for its size

Glassdoor sits at 4.1 out of 5 across 330 reviews, with 83% recommending the company and 90% approving of the CEO. Compensation scores highest at 4.4. Strong numbers on a modest sample.

Pace signal

Fast, and it shows

The recurring theme is speed. People describe a fast-paced culture with real burnout risk and reorganizations that felt chaotic as the company scaled. That's consistent with a plan-light, refound-often operating style. Directional.

Product signal

Intensity for impact

The weakest scores are senior management (3.7) and work-life balance (4.0) – the classic signature of fast scaling. The trade people describe is intensity in exchange for impact. For a company betting everything on a fast AI transition, that's the culture you'd expect – for better and worse.

The read: The culture and the strategy are the same thing here – reinvent constantly, hire for taste, keep the company from calcifying. It has produced a genuinely great product. The bet is that it also survives becoming a large enterprise vendor without losing what made it good.

Forward View

Three ways the next two years go

The variable isn't capital – Notion is profitable and hasn't needed primary money since 2021. It's whether the agent bet turns Notion into infrastructure before the bundlers and chat surfaces make it optional. Running underneath all three scenarios is a second question: independence or an exit.

Governance moves – they buy optionality

Notion formed its first board in 2026 and added Jonathan Chadwick, VMware's former CFO and COO – an unmistakable public-markets hire. Add the CFO appointment (Rama Katkar) and the employee tender, and it reads as IPO-prep. It reads equally as exit-prep: Coda's absorption, and Airtable's, show that path is real.

One operating move – it buys years

The ZeroEntropy acquisition cuts the other way. Buying the retrieval layer outright and standing up a model-research group is a multi-year commitment, and it takes cost out of AI at scale – what public-market investors scrutinise hardest in software. A company being dressed for sale rents that layer; it doesn't buy it.

Neither set of signals settles it. What settles it is what the platform earns.

Bear

Optional, then acquired on someone else's terms

Low probability, high consequence

The bundlers and chat surfaces make Notion a nice-to-have. Growth slows, the flat mark stops being an open question and hardens into a ceiling, and Notion sells into a bigger platform from a position of weakness – the way Coda did, and the way Airtable just did at a fraction of its old value. Signal to watch: revenue growth decelerating while AI adoption plateaus.

Base

A strong, profitable independent that stays flat-ish

Most likely over a 2-year horizon

Notion keeps growing revenue, teams stay sticky, the Developer Platform reaches general availability and monetizes modestly. It remains one of the best products in software and a very good business – but the market keeps pricing the dependency question conservatively until Notion discloses more, or an IPO filing forces the numbers into the open. Signal: what the platform earns once Workers start billing in October 2026, and whether any retention or active-usage number is ever shown.

Bull

The system agents write to – and a re-rating

Achievable, but needs two things to land together

Two things land together: the External Agents API reaches general availability and actually works, and third-party agents monetize on the platform. Notion becomes the durable system of record agents operate on, engagement turns out to be strong, and the framework flips from "productivity app priced flat" to "agent infrastructure" – so the mark finally moves. Or a strategic acquirer (Microsoft, Google, Salesforce, or an AI lab) pays up for exactly that position. Either way the flat era ends.

Closing Thought

Notion is one of the best-loved products of its generation and a genuinely good business – profitable, growing, built with more craft than almost anything in its category. That makes it comfortably a Tier-1 company. Whether it's a generational one turns on the question it has chosen not to answer: do teams actually depend on it as their source of truth, or do they admire it and drift back to other tools for the daily work? The agent bet is the attempt to make that dependency structural rather than optional. If it lands, a valuation that hasn't moved since 2021 will read as the last cheap moment to own this. If it doesn't, that flat price was the market seeing what the fan base couldn't. I don't know which yet. That's what makes it worth watching.

Public sources only: Ivan Zhao's interviews and Notion's developer keynote; the business and trade press; the app stores and review platforms; Notion's own product, pricing, and release pages; Glassdoor; and the competitor set. No internal data, financials, or proprietary metrics were accessed. Review patterns are thematic, not verbatim. Everything here is open to challenge.