RESOURCES · EXPERTINI ATS

Responsible Hiring Maturity

A four-level model for locating where your actual hiring process sits, honestly.

3 min read · Updated July 2026 · Expertini Editorial

Most hiring teams can't answer a simple, uncomfortable question: if a rejected candidate — or a regulator, under an increasing number of automated-hiring disclosure requirements — asked exactly why they were screened out, could the team reconstruct and defend that decision? The Responsible Hiring Maturity model below gives a four-level way to locate a real process honestly, developed from operating recruitment infrastructure since 2008 and watching where accountability typically breaks down.

▶ Watch the 30-second overview — no sign-up needed

4maturity levels, self-assessable honestly

1level where scores are both accurate and fully auditable

18years of hiring infrastructure informing this model

0compliance guarantee from tooling alone — process still matters

01Level 1 — Ad hoc

Screening happens by inbox order, gut feel, and whatever criteria happen to be in a hiring manager's head that week, with no consistent documentation. This is fast and low-overhead for five applicants and completely indefensible — and, in aggregate, discriminatory in effect even without discriminatory intent — at five hundred, because there's no record of what criteria were applied to whom, and the criteria themselves likely shifted candidate to candidate without anyone noticing.

02Level 2 — Keyword-automated

An ATS filters resumes by string matching against a list of required keywords. This is consistent in the narrow sense that the same filter applies to every resume — but consistently wrong, filtering out qualified candidates who describe real experience in different vocabulary while letting through candidates who padded a resume with the right nouns without the underlying substance. Nobody at this level can explain an individual rejection beyond "the filter didn't match," which is not an explanation a regulator or a rejected candidate will accept.

03Level 3 — AI-assisted, unaudited

Modern AI screening tools produce genuinely impressive-looking rankings — and, if built on an ungoverned model asked to score directly, those rankings are frequently stochastic (see AI & Automation for why) and the criteria behind them are opaque even to the team using the tool. This level represents better raw matching quality than Level 2 with meaningfully worse accountability, since a keyword filter's logic is at least inspectable in principle, while a black-box AI score often isn't even in principle.

04Level 4 — Deterministic and auditable

Semantic understanding, bounded by transparent, published mathematics; personal identifiers stripped before any AI analysis; every score decomposable into weighted, evidence-cited dimensions a human can check by hand; and every decision recorded with human interview judgement layered explicitly on top rather than hidden inside an opaque final number. This is the level Expertini ATS is built to make achievable by default rather than requiring a custom-built compliance program — though reaching it also depends on how a hiring team writes job descriptions and conducts interviews, not on tooling alone.

05Locating your own process honestly

A useful self-assessment: can you name the exact criteria used to screen your last hundred rejected candidates? Could you reproduce a specific candidate's rejection today, showing the evidence and reasoning, if asked six months from now? Does your interview process use the same structured criteria for every candidate for a role, or does it vary by interviewer and by how the day is going? Teams honestly answering these questions usually locate themselves lower than they initially expected — which is the point of the exercise, not a criticism of any individual team.

Frequently asked questions

What's the difference between Level 2 and Level 3?⌄

Level 2 (keyword-automated) is consistent but inaccurate — the same crude filter applies to everyone. Level 3 (unaudited AI) is often more accurate in raw matching quality but frequently less accountable, since the scoring logic is opaque even to the team using it.

Can a team reach Level 4 using tools other than Expertini?⌄

In principle, yes — Level 4 describes a set of properties (deterministic scoring, PII stripping, auditable evidence, documented human judgement), not a specific product. Few tools on the market currently implement all of them together, which is the gap this model was designed to describe.

Is Level 4 the same as being legally compliant with automated-hiring regulations?⌄

Not automatically — Level 4 gives you the auditability and reproducibility that regulations like NYC Local Law 144 and the EU AI Act's high-risk AI provisions generally require, but actual compliance depends on your specific jurisdiction, process, and legal review. We're not lawyers and this isn't legal advice.

Does tooling alone determine your maturity level?⌄

No — job description quality and interview process discipline matter as much as the scoring tool itself. A Level-4-capable tool used with vague job descriptions and unstructured interviews doesn't fully deliver Level 4 outcomes.

At a glance

Four-level self-assessment, not a certification
Distinguishes accuracy from accountability explicitly
Grounded in 18 years of hiring infrastructure operation
Level 4 achievable by default, not custom-built
Tooling plus process — neither alone is sufficient
No compliance guarantee claimed from tooling alone

See responsible hiring maturity on your own hiring.

Bring a real job description to a 30-minute demo — free trial included.

Book a demo