CRMs were always graph databases that refused to admit it. Contact records are nodes. Deals are edges. Email threads are weighted relationships between nodes. The data model was there; the interface pretended it wasn’t.
This week, three companies — Salesforce, HubSpot, and a quieter entrant we’ll get to — each shipped features that treat the CRM as what it actually is: a graph of relationships, not a spreadsheet of fields. The timing is not a coincidence.
What Salesforce shipped
Salesforce’s Einstein relationship graph update connects deal records to people records to company records in a way that surfaces “degree of separation” information natively. If your CFO went to school with the CFO at your target account, the CRM now surfaces that. If your VP of Sales worked with the target account’s new head of procurement three jobs ago, that connection is visible before the first call.
This sounds simple. It isn’t. The data science required to identify and rank relationship paths across the noise of LinkedIn connections, email signatures, and CRM history without drowning salespeople in false positives took significant engineering. Salesforce has been quietly building it for 18 months.
What HubSpot shipped
HubSpot’s version is less sophisticated but more immediately usable for the SMB segment. “Connection Paths” shows sales reps the shortest route from their current network to a target contact, pulling from email, calendar, and imported LinkedIn data. The UX is clean. The accuracy on smaller networks (fewer than 500 contacts) is genuinely impressive.
The caveat: HubSpot’s relationship graph degrades significantly at enterprise scale. Large networks produce too many low-quality connections to be useful without better signal ranking. This is the mid-market tool. Salesforce is gunning for enterprise.
The third entrant
The quieter announcement came from a series A company we’re watching closely. They’re not competing with Salesforce or HubSpot — they’re building a relationship-graph layer that sits on top of both. The pitch is simple: you already have the data in your existing CRM; we read it, build the graph, and surface the paths your reps aren’t seeing.
The “above the CRM” positioning is interesting because it doesn’t require enterprises to switch systems, which is the first thing that kills any CRM-adjacent deal. It integrates. It reads. It layers intelligence on top without threatening the system of record. That framing is increasingly common in enterprise AI — and it’s the right framing.
What this week means
When three companies ship the same category of feature in the same week without coordinating, it’s not coincidence. It’s a signal that the underlying capability — LLM-powered relationship extraction from messy contact data — has crossed a quality threshold that makes it demo-able and, more importantly, usable in production. The race to own the relationship graph layer of enterprise sales is now live. The window for new entrants is measured in months, not years.