Find the people your current process misses

CHROs are accountable for talent quality and the rules that shape who gets seen. Nodes is the People Decision Engine. It starts with one bounded hiring decision and replays it against your own post-hire outcomes. It returns the evidence, limitations, and group-level results for review. Recruiters and managers make the final decision.

Start with the decision that carries the most talent risk

The first program should be narrow enough to test. A practical starting point is who advances from screening for one role family, measured against a post-hire outcome the business already trusts. The Decision Blueprint defines the population, permitted evidence, prohibited evidence, outcome, human owner, proof threshold, and abstention rules before the model evaluates a case.

This gives the talent team a clear test:

• Which current signals are associated with the outcome?
• Which filters reject people who later succeed?
• Where is the evidence too weak to support a recommendation?
• Which differences between groups need review?
• Who has authority to approve, change, delay, or decline?

A filter can hide measurable value

The public evidence comes from enterprise talent at one Fortune 500 insurance carrier. The study covered four years of production data and 10,765 agents. One insurance-experience filter would have rejected 2,863 producing agents, putting $17.7M in annual production at risk. After correction, none of 3,597 testable resume keywords predicted the production milestone, while 30 were anti-predictive.

This evidence is specific to one carrier, one role context, and defined production outcomes. It cannot establish a universal hiring rule. Every new company and role requires historical validation against its own outcomes. Nodes never automatically rejects a candidate.

Bias monitoring needs evidence and limits

A group-outcome review can surface differences that require investigation. It cannot prove that a process is unbiased. Your people, legal, and data-governance teams define the protected-data handling, comparison groups, thresholds, review path, and corrective steps.

Nodes records the evidence available, the recommendation, model version, policy checks, human input, approved action, and later outcome in a signed Decision Trace. That gives reviewers a reproducible record of what happened without replacing job-related validation, adverse-impact analysis, or human judgment.

Frequently asked questions

Does this replace the ATS?

No. The ATS remains the system of record. Nodes connects candidate records to HRIS performance data and other approved evidence, then returns a recommendation and its evidence to the workflow your team already uses.

What if historical decisions contain bias?

Human decisions are context. They are never treated as ground truth. Measured outcomes remain governed evidence with uncertainty, and every calibration candidate must pass validation before promotion. Monitoring continues after launch because a pre-launch check cannot guarantee future behavior.

Who makes the employment decision?

A named recruiter or manager does. Nodes can recommend, explain, abstain, and prepare the next step. The designated human approves, edits, delays, or declines before anything executes.

Can the same engine support mobility, succession, and compensation?

The architecture can support those decisions, but the hiring evidence does not transfer automatically. Each program needs its own outcome, permitted evidence, historical validation, policy constraints, and human approval path. Insurance talent is the only current production-proven domain.

Put one hiring rule to a 72-hour test

Bring one repeated hiring decision and at least two years of linked candidate and outcome history. Before the clock starts, Nodes checks that the volume, join keys, outcome coverage, and data quality can support the test. Your finance team defines the $1M evidence bar and calculation in writing.

The 72 hours start only after Nodes accepts the agreed deidentified dataset delivered under a mutual NDA. The Replay is read-only. No live candidate or workflow changes. The result is historical evidence, not realized savings. If the historical record falls short of the written bar, there is no production fee and you keep the analysis.