Is My Hiring Tool an AEDT? A Tool-by-Tool Scope Test

You Have Five Hiring Tools and No Idea Which Ones Count

Your ATS flags keywords. A separate vendor scores video interviews. A third tool ranks applicants before a recruiter ever opens a resume. Somewhere in a Slack thread from eighteen months ago, someone mentioned a chatbot that pre-screens candidates for scheduling. None of this was procured as a single system — it accumulated, tool by tool, contract by contract, and nobody sat down to ask which pieces are regulated under NYC's Local Law 144 and which aren't.

This is the most common starting point for HR Directors at New York employers right now. Not "we're out of compliance" — "we don't actually know what we're compliant about." And that uncertainty is expensive to leave unresolved, because the clock on notice requirements and audit obligations only starts running once you've correctly identified that a tool is in scope.

Before you build an inventory, run a scope test. This article walks through the actual statutory test, applies it to the hiring tools employers use most, and shows you how to record a dated verdict for each one — the kind of record that holds up if DCWP or a candidate ever asks why a given tool was, or wasn't, treated as an AEDT.

Is My Hiring Tool an AEDT? The Two Questions the Statute Actually Asks

Local Law 144 doesn't regulate "AI" broadly, and it doesn't regulate every automated feature in your hiring stack. It regulates a specific, narrowly defined category: the automated employment decision tool, or AEDT.

The statutory definition, as summarized by employment counsel, covers any computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that issues a simplified output — a score, classification, or recommendation — used to substantially assist or replace discretionary decision-making in hiring or promotion (Perkins Coie, 2023).

Strip that down and you get two questions worth asking about every tool in your stack:

1. Does it produce a simplified output? A score, a rank, a pass/fail flag, a "top candidate" label — something that compresses a candidate down to a single signal.
2. Is that output used to substantially assist or replace a human decision? If a recruiter still reviews every resume line by line and the tool's output is decorative, that's a different situation than a tool whose score determines who gets a callback.

If the answer to both is yes, you're very likely looking at an AEDT. If either answer is a clean no, you're likely outside scope — but "likely" is doing real work in that sentence, which is why a documented verdict matters more than a gut feeling.

The Statutory Test, Broken Into Plain English

It helps to separate the test into its component parts rather than eyeballing the whole tool at once.

Part one: the technology. Is there machine learning, statistical modeling, data analytics, or AI actually doing computation on candidate data — not just a static rules list a human wrote once and never touches? A tool that learns from historical hiring outcomes or continuously reweights criteria based on data patterns is squarely inside this part of the test.

Part two: the output. Does the tool reduce a candidate to something simplified — a numeric score, a percentile, a tier, a recommended/not-recommended flag? This is where a lot of borderline tools live. A resume parser that extracts years of experience and displays it as a field is not the same as a parser that converts that field into a fit score.

Part three: the role in the decision. This is the part employers most often skip, and it's the one that determines whether a tool that passes parts one and two actually functions as an AEDT in your workflow. A scoring tool whose output a recruiter genuinely overrides as a matter of course is a different case, on paper, than one whose score functions as a de facto gate. DCWP guidance and legal commentary both treat "substantially assist" broadly — a tool doesn't need to make the final call to be in scope, it just needs to meaningfully shape who advances.

Answering is my hiring tool an AEDT honestly means walking all three parts for every distinct tool — not the vendor relationship, the individual capability. One vendor can sell you three tools, and only one of them might clear this bar.

Walking Five Common Hiring Tools Through the Test

Here's how the test plays out against tools most 50–250 person employers actually run.

Keyword-matching ATS filters. A static Boolean search ("must contain: CPA, 5+ years") that a human wrote and doesn't learn from outcomes generally fails part one — there's no machine learning or statistical modeling doing the work. Usually out of scope, but confirm the vendor hasn't quietly added a "smart match" or "AI-ranked" feature to the same module — many ATS platforms have layered scoring on top of what used to be plain filtering.

Resume-ranking or "fit score" tools. These typically clear all three parts: they compute a score from resume data using a model, they output that score as a simplified number or tier, and recruiters routinely rely on that ranking to decide who to call. This is close to the paradigm case the law was written for.

Video-interview scoring software. If the tool analyzes recorded responses and produces a competency score or recommendation, it clears parts one and two almost by design. Part three depends on your workflow — if hiring managers see and weigh that score before deciding, it's in scope.

Chatbot pre-screening / scheduling assistants. A chatbot that only schedules interviews and answers FAQs, with no scoring or ranking output, generally fails part two — there's no simplified output about the candidate's suitability. But a chatbot that also asks knockout questions and produces a qualify/disqualify flag is a different tool entirely, even if it's badged the same way in your vendor's marketing.

Background-check and reference-check automation. These often produce a report rather than a simplified suitability score, which keeps many of them outside the AEDT definition — but any product that converts a background report into a risk score or hire-recommendation crosses back into scope.

Every one of these determinations should be treated as your determination, for your deployment, on a specific date — not a permanent label borrowed from a vendor's marketing page.

Tools That Usually Land Outside AEDT Scope — and Why "Usually" Matters

Static rules engines, plain keyword search, calendar and scheduling tools, and reporting dashboards that display data without producing a decision-shaping score generally sit outside the AEDT definition. So do tools used purely for internal analytics that never touch an individual hiring decision.

The caveat that matters: vendors update products constantly. A tool that was a plain filter last year can ship a "smart ranking" feature this year without you noticing, because it often arrives as a checkbox in an admin settings panel, not a new contract. This is exactly why a one-time classification exercise isn't enough — the scope test needs to be something you can re-run whenever a vendor pushes an update, not a decision you make once and file away.

Recording the Verdict: What a Defensible Scope File Looks Like

Running the test in your head isn't the same as running it on paper. If DCWP ever asks why you treated a tool as out of scope, "we looked at it and decided it was fine" is not a record — it's a memory, and memories are exactly what a regulator will ask you to produce evidence for.

A defensible scope file, at minimum, records: the tool's name and vendor, the specific feature evaluated (not the whole product suite), the date of the review, who conducted it, the answer to each part of the three-part test above, and the resulting verdict — in scope, out of scope, or needs vendor follow-up. Re-date the entry whenever the vendor changes the feature set or you change how the output is used in your workflow.

A tool doesn't have to make the final hiring decision to be in scope — it only has to substantially shape who advances. That's the line most employers underestimate.

This is the exact structure behind the "Is This Tool an AEDT?" Classification Decision Workbook — a downloadable worksheet built to walk each tool in your stack through the same three-part test and produce a dated, defensible verdict you can point to later, rather than reconstruct from memory after the fact.

This Is an Operations Framework, Not a Legal Opinion

Everything above is a structured way to organize the analysis LL144 asks employers to do — it is not a legal determination, and it isn't a substitute for one. Whether a specific edge-case tool in your stack meets the statutory definition is ultimately a legal question, and borderline calls (a "smart" filter, a hybrid scoring/scheduling tool, a vendor feature update) are exactly the kind of thing worth confirming with outside counsel or directly with DCWP before you commit to a verdict in writing. The workbook, and this article, help you document your reasoning and organize the operation around it — they do not perform a legal review, and they never score any individual candidate.

Getting the scope call wrong in either direction carries real cost. Misclassifying an in-scope tool as out of scope skips the annual bias audit, the public summary posting, and the ten-business-day candidate notice the law requires (Crowell & Moring LLP, 2023; Epstein Becker Green, 2023) — and civil penalties for violations accrue per violation, per day, escalating for repeat violations (Office of the NY State Comptroller, 2025). Confirm the current penalty schedule with DCWP before relying on any specific figure in an employer-facing document.

Your First Action Item

Don't start with a spreadsheet of tool names and a gut-feel column. Start by running each hiring tool in your stack through the three-part test above, one feature at a time, and write the verdict down with a date attached. The "Is This Tool an AEDT?" Classification Decision Workbook is built to structure exactly that exercise before you move on to notice language, audit scheduling, or vendor documentation requests.

Once you have verdicts for every tool, the next question is whether your organization as a whole falls under Local Law 144's jurisdiction at all — worth confirming even for employers who are sure they use an AEDT. And if you haven't yet nailed down the statutory definition itself in full, that's the right place to start before running any tool-by-tool test. For pricing on the full workbook library, including the classification workbook alongside the audit calendar and vendor tracker, see the pricing page.