Notion sells you a canvas, and the canvas becomes the work

The last company I advised kept everything in Notion. I mean everything: the product roadmap, the interview loop, the comp bands, the onboarding checklist, a database of every enterprise account related to a database of every support ticket. It was genuinely impressive. Then the operations lead who had built most of it gave notice, and the founder's reaction was not grief. It was relief. Four months later I sat in a conference room while a director scrolled through the workspace hunting for the definition of an internal metric, and two people in that meeting had different numbers in their heads, and both numbers were defensible, because one was reading a page last edited in 2023 and the other was reading a duplicate somebody had forked and never reconciled. The workspace had survived three reorgs and each one had left sediment: pages called Untitled (7), a second roadmap with no first, a graveyard of automations still sending a Tuesday summary nobody had read in a year. Nobody in the room called this a Notion problem. It took me another year to call it one.

That room is the argument. Notion did not fail this company; the tool worked exactly as designed, which is to say it let six smart people build whatever they wanted without asking anyone's permission. The freedom is real. So is the bill, and the bill is denominated in hours that do not appear on any invoice.

The blank page is the product

Most software encodes opinions. A mature tool tells you what a project is, what a done ticket looks like, where the fields go. Notion's opinion is that you should have your own. What it sells is primitives: blocks, databases, relations, rollups, properties, permissions, the Lego kit rather than the model you snap together on the box. Every team that adopts it compiles its own small ERP, and the compilation is done by whoever has the most energy and the least seniority, and the schema decisions are never reviewed because there is no schema to review, only conventions, and conventions spread like dialects. The Wikipedia entry describes it as a workspace. The honest description is a factory for workspaces.

This is why the demo is so good and day four hundred is so bad. A capable person can assemble something beautiful in an afternoon. What they have actually assembled is a load-bearing artifact with no documentation, and the cost of that does not show up while the builder is still in the building.

Lock-in, but the polite kind

Consider the incentive. Notion Labs priced above ten billion dollars in its last reported round, which was 2021, and it has not gone public, so the return logic is seat growth, and seat growth is paid for by how much of your company's working life happens inside the product. Current list prices: Free, Plus at ten dollars a seat, Business at twenty with the AI layer bundled in. Offline use, which people begged for over years of forum threads, arrived properly only recently, and even now it is choose-pages-to-download, which tells you something about the priority queue at Notion: the features that make leaving harder rank ahead of the features that make staying less painful. The data technically leaves whenever you want; the workspace exports to HTML, Markdown, and CSV, and PDF export is saved for Enterprise. But what you export is content. The relations, the filters, the automations, the private teamspaces, the four years of undocumented convention that tells a new hire which page is the real one: that stays, because it was never in the files. Notion's lock is not the format. It is the fact that leaving means re-teaching your organization how to think, which is not the same as migrating.

The pricing tells you where the company thinks it lives. Notion spent years selling itself as the thing that replaced five tools, and the current Business plan bundles a sixth: Notion Agent, AI meeting notes, enterprise search across your connected apps, all included at twenty dollars, with a separate meter for "Custom Agents" that bills ten dollars per thousand credits. The workspace you and your colleagues built, the messy one with the duplicated pages, is now the training context for a service that acts on your behalf. I will be unsympathetic here: a human director in that conference room could not tell which metric page was current. An agent will not solve that. An agent will just guess faster, with a receipt.

The case for the mess

The strongest counterargument, and I mean to give it its full weight, is that the mess is the org chart at its actual resolution. For a company under fifty people, the alternative to Notion is not a better-organized workspace. It is three subscriptions, an integration consultant, and a context-switching tax that every study of attention says is worse than the duplication. A twelve-person company is improvising anyway. Its conventions change weekly because the org does, and a tool that hard-codes last year's org will fight them weekly. Some of the most functional teams I have seen ran on Notion with one rule, an owner per database, and it was fine. The failure mode I sat through was a governance choice. Notion's actual offense is narrower and more interesting: it makes the absence of governance cheap, and cheap things get bought, and nobody notices they are buying anything because the purchase arrives as a feature.

It depends, then, on one question: is somebody in your organization explicitly paid to say no inside the workspace. If the answer is yes, Notion is a good canvas and the flexibility is compounding. If the answer is no, the flexibility is compounding too, just in the other direction.

The variable worth watching is that credits meter. A tool that billed you for the mess it made would have an interest in cleaning it up. A tool that bills you per thousand agent actions is paid by volume of guesses. The director is still scrolling.