Mosaic AI White Paper · Part 2 of 3

The Talent Graph: Recruiting on Relationships, Not Keywords

A resume isn't a bag of words. It's a web of facts — who has a skill, who worked where, who built what, alongside whom. Mosaic turns that web into a map your team can navigate.

What a Resume Actually Says

Read any resume closely and you'll notice it's made of small, connected statements. She has a PMP certification. He worked for a regional hospital network. She led a migration to a new billing platform. He worked with a data science team. Each statement links a person to a skill, a company, a tool, a responsibility.

Keyword systems throw all of that structure away. They see that “billing platform” and “data science” appear somewhere in the text — but not who did what, in what role, or how deeply. The connections are the meaning, and the connections are exactly what gets lost.

From Statements to a Graph

Mosaic keeps the connections. As it reads the resumes and jobs in your ATS, it links every person to the entities in their story through the relationship that actually holds between them: has a skill, worked for a company, certified in a credential, responsible for a function, worked with a technology or a kind of team.

Connect millions of those links together and you get a talent graph: a living map of your entire candidate database, where every person, skill, company, and credential is a point, and every relationship is a path between them.

When a recruiter asks “who could do this job?”, Mosaic doesn't re-read a million documents. It walks a map that already knows the answer.

Why the Distinction Between “Has” and “Worked With” Matters

Here's where the graph earns its keep. Two candidates both mention Salesforce. One administered it for five years. The other worked with a team that used it. A keyword match treats them identically. The graph doesn't — because the relationship is different, and the relationship was preserved.

That lets your recruiters do something keyword search never could: dial requirements up and down with precision. Must directly have the skill? Nice to have exposure through adjacent work? The graph can answer both — and show you exactly which candidates sit on which side of the line, and why.

A Map That Compounds in Value

It's always current. New resumes and reqs from your ATS flow into the graph continuously. Nothing to rebuild, nothing to re-index by hand.
It's consistent. The same question gets the same answer, because the map doesn't change between Monday and Tuesday unless the facts do. Shortlists are reproducible, not roulette.
It explains itself. Every match traces back through the graph to specific statements in real resumes. When a hiring manager asks “why this person?”, the answer is right there.
It rediscovers. Past candidates stay connected to future jobs. The silver medalist from last year's req surfaces on next year's — automatically, even if the job titles share no words at all.

The Industry Is Catching On

Research across the AI industry points the same direction: systems that reason over raw, unstructured documents get slower, costlier, and less consistent as data grows — while systems that reason over a pre-built graph of connected facts stay fast, affordable, and reproducible at any scale. The graph isn't a feature. It's the architecture that makes everything else trustworthy.

The Bottom Line

Keyword search sees words. Mosaic sees the web of relationships behind them — and that web, built from your own data, is what turns a pile of resumes into a talent map your whole team can navigate.