By Shankar Ganapathy, Co-founder, Boomerang AI. Published November 2, 2026.
Rudy, our AI referral agent, scores every route into a target account as capability times willingness: can this person actually get the meeting, and will they ask for it. To answer that, Rudy reads 80+ relationship signals, weights each one by the persona, seniority and department of the people involved, and ignores signals that only look like relationships, such as a shared former employer on its own. The score then decides two things: which path is worth using, and who inside your company should make the ask.
This post explains how that works, using only what we already show in the product and on our site. If you score warm paths yourself, in a spreadsheet or your own agent, the same rules apply.
What is relationship scoring?
Relationship scoring is putting a number on how likely one person is to help another reach a buyer. A connector score rates a person; a path score rates a specific route, from your rep to a connector to a named buyer. Path scores are the useful ones, because the same connector can be a strong route to one buyer and useless for another.
Why does a shared employer not count as a relationship?
Because most overlaps are crowds, not ties. If 212 people in your network once worked at the same large company as your buyer, you do not have 212 relationships. You have one data point that says almost nothing on its own.
So Rudy treats a shared employer alone as noise. It counts when something narrows it down: the same team, the same years, the same function, or a second signal such as a meeting or a co-authored paper. A hundred people who passed through one fund or one employer are not a hundred relationships.
Which signals does Rudy read?
The signals fall into a few families. These are the same evidence types listed on our How it works page.
How are the signals weighted?
The same signal means different things for different people, so every signal is weighted by persona, seniority and department.
- Function matters. Some functions stay close at any company size. In-house legal teams tend to know each other even at a 10,000-person company, where two reps may never have met.
- Seniority matters. A CISO and a CSO are peers. A path between peers is usually stronger than one that asks a junior person to reach up three levels.
- Recency matters. A comment on the buyer's post last week is a live connection. A meeting four years ago is history.
What does capability times willingness mean?
Capability asks whether this connector can get the buyer's attention: how close the tie is, and how senior and relevant the connector is to the buyer. Willingness asks whether they will actually ask: how close the connector is to your company, and whether they have helped before.
Multiplying the two keeps a strong tie through a reluctant connector from looking like a sure thing. In an illustrative example on our comparison pages, a customer CISO with a four-year work overlap, a shared conference panel and a recent comment from the buyer scores 0.93 on capability and 0.95 on willingness, for a path score of 88. The numbers are illustrative; the shape of the calculation is the point.
What happens after the score?
A score without a sender is just a list. Rudy picks the internal hop, the person inside your company with the strongest tie to the connector, and applies the routing rules you set:
- Customer asks route through the CSM who owns the account, never directly.
- Board asks route through the CEO, and only above the deal size you set.
- Partner asks route through your alliances lead.
- Every ask respects the connector's preferences on deal size, channel and cadence.
Rudy then drafts the ask in that person's voice. The relationship owner approves it and sends it from their own account. Rudy never sends on anyone's behalf.
Can I see why a path scored the way it did?
Yes, and you should insist on it from any tool. Since the week of October 26, every path row shows its evidence: the post a connector commented on, the patent or paper they share, the event page, each with its link and date. If a score cannot show its evidence, treat it as a guess.
How to score warm paths yourself
- Start from the buyer, not the connector. List the people you need at the account.
- For each possible path, write down the evidence, not the overlap. Who, what, when, with a link.
- Drop any path whose only evidence is a shared employer or school with no matching team or years.
- Rate capability and willingness separately, then multiply.
- Decide who asks before anyone drafts anything.
Rudy does this across every target account in your CRM and keeps it current as signals change. See it on the Path to Power use case.
Frequently asked questions
How does Boomerang AI score a warm path?
Rudy scores each path as capability times willingness: can the connector get the meeting, and will they ask. It reads 80+ relationship signals and weights each by persona, seniority and department, then picks who inside your company should make the ask.
Why does a shared employer not count as a relationship?
Most shared employers are crowds. Hundreds of people can share one former company without knowing each other. Rudy ignores a shared employer on its own and counts it only when the team, years or function match, or when a second signal backs it up.
What signals are used in relationship scoring?
Conversation history from CRM, calendars, call transcripts and Slack; work overlaps; public collaboration such as co-panelists, co-inventors and co-authors; classmates and alumni; shared boards and investor portfolios; and social engagement such as comments and recommendations.
Can I see the evidence behind a path score?
Yes. Every path row in Boomerang shows the evidence behind the relationship, such as a post the connector commented on, a shared patent or paper, or an event page, each with a link and date. A score without evidence should be treated as a guess.
.png)
