How to Build an HR and Recruiting App with AI
Contents
Last updated: August 2026.
If a recruiting app scores, ranks, or flags candidates in any way, even a simple 1-to-100 ranking or a "strong fit" label, it very likely counts as an Automated Employment Decision Tool under New York City's Local Law 144, and that law reaches any employer evaluating a candidate who lives in NYC, regardless of where the company itself is based.
What counts as an AEDT
Local Law 144 defines an AEDT as any computational process that issues a score, classification, or recommendation used to "substantially assist or replace" a hiring decision. The bar for "substantially assist" is broad on purpose: a human making the final call doesn't exempt you if that human is heavily relying on the tool's ranking to narrow the pool. A resume screener that ranks candidates 1 to 100, a system that flags someone "strong" or "weak," or an ATS feature that auto-advances the top tier all count as the kind of simplified output the law is built around. If your recruiting app has a ranking or scoring feature at all, assume it applies and confirm with a legal opinion, don't self-determine your way out of it. Huxly can generate the interface for that scoring feature quickly, but the bias audit and legal classification of the feature itself have to happen separately, no app builder resolves that question for you.
Three separate obligations, not one
Employers using a covered tool need an independent bias audit completed within the past year, calculating impact ratios across sex, race, and ethnicity categories. They need to publicly post a summary of that audit's results. And they need to give each candidate at least 10 business days' notice before the tool is used, along with the option to request an alternative assessment method. These are three distinct requirements, and one of the most common compliance gaps is publishing the audit results while forgetting the separate candidate-notice requirement entirely, or vice versa.
"Our vendor handles compliance" is the most common mistake
Local Law 144 places the obligation on the employer and employment agency using the tool, not on the vendor that built it. Even if a platform like HireVue or Pymetrics runs its own bias audit, the employer deploying that tool in NYC still carries an independent compliance obligation. Building a recruiting app for other companies to use doesn't transfer your customers' legal exposure to you, and assuming it does is exactly the gap that shows up in enforcement reviews.
The map is bigger than NYC now
Colorado's AI Act regulates high-risk AI systems, employment included, with its own algorithmic discrimination duties. Illinois has regulated AI video interviews specifically since 2020. Maryland requires consent for facial recognition used in pre-employment interviews. And the EU AI Act classifies recruitment and selection AI as high-risk under Annex III, with obligations, human oversight, system monitoring, log retention of at least six months, and disclosure to affected candidates, phasing in as of August 2026. The bias-audit infrastructure built to satisfy NYC's law is largely reusable across these other regimes, which is the practical reason to treat Local Law 144 as the baseline rather than a one-off requirement.
Enforcement is getting stricter, not looser
A December 2025 New York State Comptroller audit found the city agency's enforcement of Local Law 144 ineffective in specific, documented ways, three-quarters of complaint calls about AEDT issues never reached the right office, and when the Comptroller's own auditors reviewed 32 companies the city had cleared, they found at least 17 potential violations the city's review had missed. That kind of finding tends to produce tighter enforcement, not looser, going forward, worth planning around rather than assuming a quiet compliance environment continues.
The same "build compliance in from day one" logic that shapes fintech and healthcare apps applies here too, generating the interface through Huxly (Expo, Flutter, or Swift, flat $25 or $99 a month) covers the app-building side while the audit and legal work stay a separate line item regardless. For the broader cost picture, see our breakdown of what it costs to build a mobile app in 2026.
FAQ
Does a simple candidate ranking feature really count as an AEDT?
Yes, in most cases. A 1-to-100 score, a "strong fit" or "weak fit" label, or a feature that auto-advances top candidates all count as the simplified output Local Law 144 targets, even if a human makes the final hiring decision.
Do I need to worry about NYC's law if my company isn't based there?
Yes, if you evaluate any candidate who lives in New York City. The law's reach follows the candidate's location, not the employer's headquarters.
What are the actual penalties for non-compliance?
$500 to $1,500 per violation, with each day of an ongoing violation and each un-notified candidate counted separately, which can add up quickly across a real candidate pool.
If my vendor already did a bias audit, am I covered?
Not automatically. The compliance obligation sits with the employer or employment agency using the tool, not the vendor. A vendor's own audit doesn't substitute for the employer's independent obligation.
Can Huxly help with the bias audit requirement itself?
No. Huxly generates the native app, including a scoring or ranking feature once you've specified it, but the independent bias audit, impact-ratio calculations, and candidate notice process are legal and statistical work that has to be handled separately, by an auditor, not an app builder.
Are the bias audit and the candidate notice the same requirement?
No, they're separate. You need a published audit summary and a 10-business-day advance notice with an opt-out option for each candidate. Satisfying one doesn't satisfy the other.
Is NYC's law the only one I need to worry about?
No. Colorado, Illinois, and Maryland all have their own AI-in-hiring requirements, and the EU AI Act classifies recruitment AI as high-risk with obligations phasing in through 2026. NYC's framework is a reasonable foundation, not the whole picture.
Is enforcement of this law being taken seriously?
Increasingly, yes. A December 2025 city audit found real enforcement gaps and under-detected violations, which historically leads to stricter follow-through rather than the issue quietly fading.
