How does relationship intelligence work?
Relationship intelligence works in three steps. It ingests the connection signals your team already generates, like email headers and calendar history. It builds a graph of who knows whom, scoring each tie by how recently and often people interact. Then it ranks the warmest path to any target person, refreshing the map automatically as new interactions happen.
Each step solves a problem the previous one creates. Ingestion gathers raw signals that would otherwise stay locked in individual inboxes. Scoring turns that raw activity into a measure of relationship strength, so you can tell a real connection from a one-time email. Ranking turns the whole graph into a usable answer to a single question: who here can reach this person, and how strongly? The result is a map of your firm's actual reach, built on data nobody had to type.
What signals does it read, and how does it score them?
It reads metadata first: who emailed whom, how often, who sat in which meetings, and which contacts are shared. Each relationship gets a strength score based on recency and frequency. Recent, regular contact scores high. The score decays over time, so a tie you haven't touched in a year ranks below one you nurtured last week.
Metadata does most of the work, and on purpose. You don't need to read the body of an email to know that two people exchange messages weekly and meet monthly. That pattern alone is a strong signal of a real relationship. Here's how the main inputs map to what they tell the system:
| Signal | What it indicates |
|---|---|
| Email frequency | How active a relationship is right now |
| Recency of contact | Whether a tie is warm or cooling off |
| Meeting history | Depth and seriousness of the relationship |
| Shared contacts | Overlap and reach into a wider network |
The decay rule is the part that keeps the map honest. Relationships fade if you don't tend them, and the score reflects that. To understand the structure these scores live inside, read what a relationship graph is.
How does it find the warmest path?
Once the graph is scored, finding the warmest path is a search problem. You name a target person, and the system traces every chain of connections from your team to that person, weighs each chain by tie strength, and returns the shortest, strongest route. Instead of guessing who might know someone, you get a ranked, specific answer in seconds.
Say a partner wants to reach a CFO at a target account. The system checks who on the team has a tie to that CFO, how strong each tie is, and whether a two-step path through a trusted intermediary beats a weak direct one. It returns the best route, names the colleague to ask, and shows why that path is warmest. That's the difference between a hopeful guess and a deliberate move, and it's the heart of using relationship intelligence in your sales process.
This is also where relationship intelligence pulls away from a plain CRM. A CRM can store a contact, but it can't rank paths to a person it never recorded a relationship for. As covered in what relationship intelligence software is, the warm-path ranking is the payoff the whole pipeline builds toward, and it's a question a traditional CRM simply can't answer.
The ranking logic handles a subtlety people miss: a warm path isn't always the shortest one. A direct connection that's gone cold can be a worse route than a two-step path through someone whose tie to the target is strong and recent. The system weighs total path strength, not just hop count, so it might recommend you go through a trusted colleague rather than reach out yourself. That mirrors how introductions actually work in the real world, where the person who vouches for you matters more than the number of degrees between you and your target. The strength of the voucher carries the introduction, so a warm two-step beats a cold direct line almost every time.
How do you put it to work, and how is privacy handled?
Put it to work before outreach, not after. Search a target account, see who holds the strongest tie, and ask that colleague for a specific introduction. On privacy, AVNIR uses a graduated trust model: metadata analysis is the baseline, and reading email bodies stays opt-in and off by default, so each person controls the access they grant.
The graduated model is what makes a firm-wide rollout realistic. People are wary of tools that vacuum up their inboxes, and rightly so. By defaulting to metadata only, AVNIR lets teams get real value while keeping message content private until someone deliberately opts in. The map still works on metadata alone, so no one has to trade privacy for usefulness.
To apply it this week, learn one play and repeat it: pick a single account, find the warmest path, and make one warm introduction instead of one cold email. Do that across a handful of accounts and the compounding effect becomes obvious. When you're ready to try it on your own data, request early access and start mapping the relationships your team already has.
A quick note on what good looks like as the graph matures. In the first week, the value is discovery: you see ties you'd forgotten and paths you didn't know existed. By the second month, the value shifts to habit, checking the warmest path before outreach becomes automatic, and reps stop sending cold emails when a warm route is one ask away. The decay scoring keeps doing quiet work in the background, flagging key relationships that are cooling so someone can reach out before the tie goes dormant. That's the difference between owning relationship data and actually using it.