Answer
What changes about AI in hiring depending on where you operate?
Whether you need an independent audit, what you must publish, and how much notice a candidate gets before a tool assesses them.
Whether an independent bias audit is required, whether its results must be published, how much notice a candidate must receive, and whether they can request a human alternative. These are operating requirements, not disclosures.
Hiring is the area where automated decision-making has attracted the most specific regulation, and the requirements are unusual in shape. They are not primarily about what the tool may do; they are about what must exist around it. An independent evaluation performed within a defined period. A published summary of that evaluation. Notice to the candidate before the tool is used, with a stated minimum period. And in some cases the right to request an alternative process.
That shape matters because it changes the cost. A disclosure requirement is a line of text; an audit requirement is an engagement with an independent party, a set of calculations across demographic categories, a published result and a repeat before it expires. A business adopting a screening tool in a jurisdiction with these rules is adopting an ongoing obligation rather than a piece of software, and that obligation is frequently larger than the licence fee.
The scope question is the one that catches employers out. These requirements generally attach based on where the candidate is rather than where the employer is, so a business hiring remotely can fall within a city's rules without any presence there. For any role advertised beyond a local area, the question is not where you are but where your applicants are, and that is a question about your advertising rather than your premises.
The definition of the covered tool is worth reading rather than assuming. The rules are written around tools that substantially assist or replace a discretionary decision, which is broader than the obvious screening product and can reach a ranking, a score, or a filter applied to applications. A business using something informally to sort a pile of applications may be inside a definition it never considered, and the informality of the use does not remove it.
Beyond the specific rules, ordinary discrimination law applies to the outcome regardless of the mechanism. A process that produces a disparate result is a problem whether it was produced by a person or a tool, and the tool does not supply a defence. This is the constraint that applies everywhere, and it is the reason the audit requirements exist rather than an alternative to them.
The practical position for a small business is that automated screening carries an obligation cost that scales badly at low volume. Using a tool to help write a job description, to draft interview questions or to summarise a conversation afterwards raises none of this, because none of them assesses a candidate. The line is whether the tool contributes to the decision about a person, and staying on the useful side of it is straightforward once the line is stated.
In some places the tool is not the thing being regulated; the process around it is, and that process has to exist before the tool runs.
Siddharth Sharma, Context Theory
Related questions
Does this apply to using an assistant to read CVs?
Read the definition rather than assuming it does not. If the output ranks, scores or filters candidates and that materially assists the decision, it may be within scope, and the informality of the arrangement does not exempt it. Summarising a CV for a human who then assesses it is a different activity, and the distinction is worth being able to state.
What should a small business do if it hires nationally?
Either avoid tools that assess candidates, which is straightforward and costs little at small hiring volumes, or build to the strictest requirement among the places applicants come from. The first is what most small employers should choose, because the obligation cost does not scale down with the number of hires.
METHOD
Every figure below carries its source and the date it was verified. Nothing on this page is asserted.
The numbers on this page.
| What | Value | Specific to |
|---|---|---|
| Close rate — response under 5 minutes vs over 24 hours | 32% vs 12% | Category-wide |
| Sub-15-minute compliance — automated routing vs manual only | 62.5% vs 39.1% | Category-wide |
Optifai speed-to-lead benchmark · n=939 companies · Q2 2025–Q1 2026 · verified
2026 speed-to-lead benchmark · verified
What is specific to this page.
| Kind | Claim | Check it against |
|---|---|---|
| Regulation | Requirements in this area govern the process around the tool rather than the tool's behaviour: an independent evaluation within a defined period, a published summary of its results, advance notice to the candidate, and in some cases a right to request an alternative process. | The automated employment decision tool rules published by the New York City Department of Consumer and Worker Protection. |
| Constraint | Coverage generally attaches to the candidate's location rather than the employer's, so an employer hiring remotely can be within a jurisdiction's rules without any presence there, making advertising reach the determining factor. | The residency-based scope language in the applicable rules, read against where the role was advertised. |
| Licensing | Covered tools are defined around substantially assisting or replacing a discretionary decision, which can reach an informal ranking, score or filter rather than only a purpose-built screening product. | The definition of an automated employment decision tool in the applicable rules, compared against the informal use in question. |
| Workflow | Ordinary discrimination law applies to the outcome regardless of the mechanism producing it, so the tool provides no defence and the audit requirements supplement rather than replace that constraint. | The applicable employment discrimination statutes, which are framed around effects rather than instruments. |
Each row would be wrong on another industry's page. Where a sourced figure exists it is in the table above instead; these are the constraints that shape the work and do not happen to be numbers.
Start with the measurement.
Reading about a benchmark is not the same as knowing your own number. The audit produces yours, measured rather than estimated.
$497 · delivered in 5 business days · credited against month one