There is a default move every founder between 30 and 250 employees makes around the same revenue mark. The team needs a CRM. Marketing wants automation. Support wants tickets. Someone Googles "best CRM for scale-ups," the answer is HubSpot, three contracts get signed, an onboarding bill gets paid, and a year later the company is paying six figures a year to keep its own data tidy inside a tool nobody really likes.
The instinct made sense in 2018. In 2026 it doesn't. The system of record for a company under 250 people is no longer a vertical CRM stack. It is a workspace where the data, the documents, the workflows and the agents all live in the same place. For most of these companies that workspace is Notion.
This is not a soft take on tooling preference. It is a structural argument about where the data lives, who is allowed to change it, and what gets to run on top of it.
Why HubSpot stops working
HubSpot is a good product that ages badly inside a growing company. Three things compound and none of them are visible at signing.
First, the pricing. The March 2024 restructure moved HubSpot to seats-based pricing with separate hubs and contact-tier surcharges layered on top. Existing customers documented 5x to 20x cost increases in the months that followed. By the time you are at 100 seats with Sales Hub Enterprise, Marketing Hub Enterprise and Service Hub Pro, you are looking at well into six figures a year before you have paid for onboarding, training or a single integration.
Second, the data model. Contacts, companies, deals, tickets. Four object types, hard-coded, with custom properties bolted on. Anything that does not fit, partner accounts, multi-entity billing, project delivery, supplier onboarding, ends up either as a spreadsheet on someone's desktop or as a $40K Operations Hub workflow that breaks on the next release. The schema is the vendor's, not yours.
Third, the change cycle. Every meaningful workflow change inside HubSpot is an admin task. Sometimes it is a vendor task. The people who actually run the workflow, the rep, the CSM, the ops lead, cannot redesign it themselves. They file a request, wait two weeks, get something close to what they asked for, and adjust.
“We were spending more time configuring HubSpot than we were spending with customers. And every quarter the bill went up while the product stayed the same shape it had always been.”
- Head of RevOps, 80-person SaaS
What changed about Notion
Notion in 2020 was a wiki with databases. Notion in 2026 is a different category of product. Three releases moved it from "docs tool" to "system of record." Notion 3.0 in September 2025 rebuilt Notion AI as Agents. Notion 3.3 in February 2026 shipped Custom Agents, autonomous, scheduled, team-shared. Notion 3.4 in April 2026 added AI Autofill, putting agents directly inside database rows, and Skills, so the agent can learn a workflow once and run it forever.
In the two months after Custom Agents launched in beta, teams built over a million of them. Ramp, Vercel and other operationally serious companies are now running internal workflows that used to live across half a dozen vendors inside one Notion workspace. The agent reads from the same databases the humans read from, writes back to the same databases the humans write to, and the audit trail is the page history.
Notion is your CRM, because the schema is yours
The single most underrated thing about Notion is that the database is the primitive. Not pages. Not docs. Databases. A deal pipeline is a database. A customer account is a database. A renewal cohort is a database. The properties are yours, the views are yours, the relations between them are yours, and they take an afternoon to build, not a quarter.
Building a CRM inside Notion is no longer a hack. It is a pattern. Companies are doing it in three layers.
- 01A Contacts database, related to a Companies database, related to a Deals database, related to an Interactions log that ingests email and meeting data via the API or a connector like Pipedream.
- 02Views per persona. Sales sees a kanban of deals. CS sees a list of accounts by health. Finance sees deals by ARR. Same data, different lens, no migration.
- 03Custom Agents on top. One drafts follow-up emails from the Interactions log. One triages inbound leads. One generates the Monday pipeline review and posts it in the team channel.
What you get is a CRM you actually control. When sales decides next quarter that the pipeline needs a new stage, that's a column. When finance needs a partner-channel breakdown, that's a relation. When a new motion gets piloted, the database supporting it exists by lunchtime. The reason this matters is not speed. It is that the operating model is no longer a permission you have to ask the vendor for.
The agent layer is where it stops being a comparison
HubSpot has Breeze. Notion has Custom Agents. The two are not the same shape. Breeze is AI added to a closed product surface. Custom Agents are autonomous workers that live on top of your data with your permissions, your schemas, your skills, and they treat the workspace as their long-term memory.
A Custom Agent in Notion is set up once, given a trigger or a schedule, and runs in the background. The Notion team's own benchmark is over 20 minutes of multi-step actions per run, with state held in pages and databases. AI Autofill puts the same agents directly into database rows: a new contact is enriched, a new ticket is categorised, a new account is researched, automatically, with the agent writing the result back into the same record everyone else is reading.
The integration story is the other piece. Notion's API plus native MCP support means the agent can act outside Notion without leaving its context. Slack, Gmail, Stripe, Linear, Intercom, Snowflake, internal APIs, all reachable through the same agent that just read the deal record. The workflow used to be: human reads CRM, human switches tab, human pastes data into another tool, human pastes result back. The workflow now is: agent reads database, agent calls external tool, agent writes result back. The human reviews. That is the entire change.
“Anything you can do in Notion, your agent can do for you.”
- Notion 3.0 release notes
The pricing argument is not subtle
A 150-seat company on HubSpot Sales Hub Enterprise plus Marketing Hub Enterprise plus Service Hub Pro is paying, conservatively, $200K to $400K a year once contact tiers and required add-ons are layered in. Onboarding fees are separate. Operations Hub is separate. Any meaningful integration work is separate.
The same 150-seat company on Notion Business at roughly $20 per user per month, plus the AI add-on at roughly $10 per user per month, is paying around $54K a year. All-in. No contact tiers. No hub bundling. No onboarding fee. The CRM is custom, but you built it in a week and you own the schema.
The honest counter-argument is that HubSpot ships things Notion does not ship out of the box. A native dialler. Sequences with reply detection. A built-in marketing automation runtime. If your business genuinely depends on those, HubSpot is doing real work for the money. For most companies under 250 people, those features are either unused, replaced by a $30-per-seat point tool, or replaced by a Custom Agent inside Notion that does the specific version of the job the team actually needs.
When HubSpot is still the right answer
Three cases. Be honest about whether you are in them.
- 01You are a B2C or freemium business with hundreds of thousands of contacts where Marketing Hub's automation runtime is the product, not the wrapper. Notion is not a marketing automation engine.
- 02You have a 50-rep outbound sales team where the dialler, sequencing and forecasting features are doing real, daily work and switching them off would visibly hurt revenue.
- 03You have an enterprise procurement requirement, often public sector or regulated, that lists HubSpot as a preferred vendor and lists no flexible substitute.
If you are not in one of those three, the default has flipped. The cost of running on HubSpot is no longer just the bill. It is the operating model you can't change because the vendor owns the schema.
How the migration actually goes
Not a big bang. Nobody who has done this successfully migrated all of HubSpot to Notion on a Friday. The pattern is the same every time.
- 01Stand up the Notion data model. Contacts, Companies, Deals, Interactions, Tasks. One week with someone who has built a CRM before.
- 02Mirror, do not migrate. Pipe HubSpot data into Notion via the API. Both run in parallel for a quarter. Sales lives in Notion. HubSpot becomes read-only.
- 03Move workflows one motion at a time. Inbound triage first. Then renewals. Then pipeline review. Each motion gets a Custom Agent.
- 04Replace marketing automation last, with a focused tool plus a Notion agent, not with another all-in-one bundle.
- 05Decommission HubSpot at the next renewal. Export the historical record. Keep it cold. Stop paying.
“By the third month nobody opened HubSpot. By the sixth month finance asked why we were still paying for it. We weren't, the next quarter.”
- Founder, 120-person services business
The unifying frame
The system of record question used to be: which vendor's database has the cleanest customer data? In 2026 it is: which workspace lets the humans, the data and the agents share one operating model? For companies under 250 people that workspace is Notion. The CRM is something you build inside it in a week. The agents are something you ship every Friday. The bill is a fraction of what it was. And the shape of the company finally matches the way the founder actually thinks about it.
HubSpot is not going away. It is going upmarket. For everyone else, the centre of gravity has already moved. The only question is when the renewal lands.
