By Shankar Ganapathy, Co-founder, Boomerang AI. Published November 2, 2026.
A CRM is built around accounts, contacts and opportunities. A relationship is something else: a tie between two people, with evidence, a strength and a date, and very often between people who will never be in your CRM. That is a graph, not a set of tables. The answer is not to replace the CRM. It is to keep the relationship graph next to it, and write the outcomes back.
I have watched a lot of teams try to force relationships into CRM fields. A "Relationship owner" picklist. A "Board connection" checkbox. A notes field that says "ask Sam, he knows their CFO". None of it survives a quarter.
What is a relationship graph?
A relationship graph has people as nodes and ties as edges. The people include your team, your investors and board, your customers and your partners, plus the buyers you want to reach. Each edge carries what a CRM field cannot: the type of tie, the evidence behind it, when it was last active, and how strong it is for this particular buyer.
Why can't the CRM hold relationships?
What breaks when you force relationships into the CRM?
Connectors are not contacts. Your board member's former colleague, now CFO at a target account, is the most valuable person in the chain. She is not your contact, and adding her as one muddies every report you run.
Ties decay. A relationship that was strong in 2022 may be cold now. A field does not know that. A graph that refreshes from calendars, email and public signals does.
The ask needs rules, not fields. Customer asks should go through the CSM. Board asks should come from the CEO, and only for large deals. A connector may prefer Slack to email, or one ask a quarter. Those are routing rules and preferences, and they belong with the relationship, not in a picklist.
What should live where?
Keep the CRM as the system of record for the deal: the account, the opportunity, the stage, the amount, and the fact that an intro was requested and what came of it. Keep the graph for the ties and their evidence. Then connect them, so a rep never has to leave the CRM to find a path, and the CRM always knows what the relationship produced.
That is how Boomerang AI works. The graph is built from four connector types (your team, investors and board, customers and partners) and from signals that already exist, so no connector has to install or upload anything. Rudy reads your CRM to know which accounts and deals matter, finds the paths, and logs every intro request and outcome back to the opportunity. It works natively with Salesforce, HubSpot and Attio. For your own systems, relationship lookups in the public API return each connector's relevance, the reasoning behind a path and its evidence links.
Does this matter more now that agents do the work?
Yes. An agent working only from CRM tables can tell you who you have met at an account. It cannot tell you who your investors, customers and partners know there, because that context was never in the tables. People still lean on people: LinkedIn's Dan Shapero says buyers "rely on their colleagues or friends over AI or search engines" when choosing, in The Trust Advantage research. Give your agents the graph, or they will keep recommending cold outreach to accounts you could have walked into.
Questions to ask about your own stack
- Can we see who our board, investors and customers know at a target account, without asking them?
- Does every relationship show its evidence and when it was last active?
- Do our routing rules live anywhere a system can enforce them?
- When an intro happens, does the CRM know, and does it show up in pipeline reports?
If the answer to most of these is no, start with what a relationship CRM is, then read how Rudy scores a warm path.
Frequently asked questions
What is a relationship graph?
A relationship graph stores people as nodes and the ties between them as edges. Each tie carries its type, the evidence behind it, when it was last active and how strong it is for a given buyer. It covers people who are not in your CRM, such as your investors' contacts.
Can a CRM be used as a relationship graph?
Not well. A CRM models accounts, contacts and opportunities. Relationships are person to person, decay over time and need routing rules. Forcing them into fields or notes goes stale fast. Keep the CRM as the system of record and connect a relationship graph to it.
What should stay in the CRM?
The account, the opportunity, its stage and amount, and the record that an intro was requested and what it produced. Boomerang AI logs every intro request and outcome to the opportunity, so pipeline reports show what relationships produced.
How does Boomerang AI build its relationship graph?
From four connector types, your team, investors and board, customers and partners, using signals that already exist such as CRM history, calendars, work overlaps and shared boards. Connectors do not have to install or upload anything.
.png)
