I Replaced a CRM With a Notion Database and Claude
For years the sales side of running Dolphin Infotek lived in a mix of email threads, WhatsApp messages, and whatever I remembered from the last call. It worked in the sense that deals still closed, but "worked" is doing a lot of lifting in that sentence. At some point I sat down and built a proper pipeline — not in Salesforce or HubSpot, but in a single Notion database, with Claude doing the part a CRM usually charges you extra for.
Why not just buy a CRM
The honest answer is that most CRMs are built for sales teams, and I don't have one. I run business development for a twelve-person company largely on my own, alongside actually building the product. A tool designed around territories, quotas, and multi-rep handoffs solves problems I don't have, and charges per seat for the privilege. What I actually needed was much smaller: a place to see every open deal at a glance, know what happens next on each one, and not lose the thread of a conversation that started three weeks ago.
The schema is deliberately small
The pipeline is one Notion database with a handful of properties: Stage (Lead, Qualified, Demo, Proposal Sent, Negotiation, Closed-Won, Closed-Lost, On Hold), Priority, Product, Type — because not every row is a deal, some are marketing threads or internal strategy notes — and a Next Action plus Next Action Date pair that drives a calendar view. That's it. No custom fields for lead scoring, no automation rules, no pipeline stages nobody ever moves things out of.
Every opportunity gets its own page underneath the database row, and that page has one section that matters more than everything else: a running Conversation Log, newest entry on top, dated, where I paste or summarize what happened in every call, email, or meeting. It's unglamorous and it's the entire system.
Where Claude actually fits in
A conversation log is only useful if you can quickly get an answer out of it, and that's the part a plain database doesn't help with. This is where the setup stops looking like a CRM and starts looking like something else: I can point Claude at a deal's page and ask "what's the actual status here" or "did I already tell them about the pricing change," and get a synthesized answer pulled from months of loosely-formatted notes instead of scrolling myself. Before a call, I'll ask it to summarize the last three log entries so I'm not walking in cold. After a call, I paste the raw notes in and let Claude help tighten them into something my future self can actually use.
The part I didn't expect going in was how well this handles the thing every informal pipeline eventually breaks on: deals going quiet. A CRM would need a rule engine to flag a stale deal. Here, I can just ask which opportunities haven't had a Next Action Date update in a while, and get a real answer instead of a dashboard I'd have to remember to check.
What this setup doesn't do
It's worth being honest about the limits. There's no forecasting model, no automated email sequences, no reporting dashboard for a board deck. If I had a sales team of ten reps, this would fall apart fast — the whole thing works because there's one person filling it in and one person reading it back. It's a system sized correctly for a company where the person doing sales also writes code most days, not a scaled-down enterprise CRM.
Why this is worth writing about at all
I don't think the interesting story is "founder uses Notion," which is not exactly news. It's that the actual replacement for CRM software, for a business this size, wasn't more structure — it was a place to write things down honestly and a way to ask questions of that writing without doing the reading myself. That's a different shape of tool than what most sales software is optimized to sell you, and it's a pattern I think applies well beyond sales: any process that's mostly "remember what happened and decide what's next" is a good candidate for a plain database plus an AI layer, before it's a good candidate for dedicated software.