ConnectDevs

Tag: AI recruiting automation

  • AI Hiring Laws in 2026: What Recruiters Need to Know Before Using AI Tools

    AI Hiring Laws in 2026: What Recruiters Need to Know Before Using AI Tools

    AI Hiring Laws in 2026: What Recruiters Need to Know Before Using AI Tools

    AI hiring laws now apply to ordinary recruiting work such as resume ranking, candidate scoring and automated interviews. The rules differ by location, several deadlines moved during 2026, and many online summaries are already out of date.

    This guide summarizes what is in force and what is coming, based on the sources linked below as of October 2026. It is a plain-English overview for recruiters and founders, not legal advice. Always confirm details with counsel.

    TL;DR

    • NYC Local Law 144 has been enforced since July 5, 2023 and requires bias audits and candidate notice for automated employment decision tools.
    • Illinois (HB 3773) and Texas (TRAIGA) took effect January 1, 2026. California’s automated-decision regulations took effect October 1, 2025.
    • Colorado’s rewritten AI law applies to decisions made on or after January 1, 2027.
    • The EU’s high-risk hiring obligations have been postponed to December 2, 2027, though many summaries still say August 2026.
    • Practical steps help in every jurisdiction: inventory tools, give notice, keep humans in the loop, keep records and ask vendors for audit evidence.

    Table of Contents

    Snapshot: what applies where

    The table below summarizes the main rules. Sources are secondary trackers, so verify dates against the official text before relying on them.

    Jurisdiction Law Status Main duty
    New York City Local Law 144 Enforced since July 5, 2023 Bias audit, public summary, candidate notice
    Illinois HB 3773 Effective January 1, 2026 No discriminatory AI use; notice to applicants
    California Civil Rights Council ADS regulations Effective October 1, 2025 Anti-discrimination rules; four-year records
    Texas TRAIGA (HB 149) Effective January 1, 2026 No AI built or used with intent to discriminate
    Colorado AI Act as rewritten by SB 26-189 Applies to decisions on or after January 1, 2027 Disclosure and consumer rights
    European Union AI Act, high-risk systems Obligations from December 2, 2027 Risk management, transparency, human oversight

    New York City Local Law 144

    Local Law 144 requires employers in New York City to audit automated employment decision tools for bias and to tell candidates about their use. According to the NYC Department of Consumer and Worker Protection, the key requirements are:

    • The tool must have had an independent bias audit within one year of its use.
    • Information about the audit must be publicly available.
    • Candidates must receive notice, and the department has clarified that this must be at least 10 business days before the tool is used.

    Penalties are $500 for a first violation and $500 to $1,500 for each later violation, and each day of use and each missed notice can count separately, according to a Metaview summary of 2026 and 2027 rules. A study cited in HeroHunt’s guide found that only 18 of 391 covered employers had posted the required audit, so awareness remains low.

    Illinois, California, Texas and Colorado

    Illinois

    HB 3773 amends the Illinois Human Rights Act. As of January 1, 2026, employers may not use AI in ways that discriminate against protected classes, may not use ZIP codes as a proxy for protected classes, and must notify applicants and employees that AI is used. Enforcement runs through the existing civil-rights complaint process.

    California

    The Civil Rights Council’s automated-decision regulations, effective October 1, 2025, apply to employers with five or more employees. Liability can extend to vendors and consultants acting as agents. Employers must keep records of the system, its selection criteria and its outputs for four years, and anti-bias testing evidence is relevant to a defense.

    Texas

    TRAIGA took effect January 1, 2026. It prohibits developing or deploying AI with the intent to unlawfully discriminate, a narrower standard than outcome-based tests. The Texas Attorney General has exclusive enforcement, there is no private right of action and a 60-day cure period applies.

    Colorado

    Colorado’s original AI Act was delayed more than once. According to Metaview and HeroHunt’s state-by-state guide, SB 26-189, signed in May 2026, rewrote the law. It removed the duty to prevent algorithmic discrimination, the mandatory impact assessments and the deployer risk-management program, and replaced them with a disclosure-and-rights model enforced by the Attorney General. Its duties apply to consequential decisions made on or after January 1, 2027. A federal court order in April 2026 also limits enforcement for a period tied to a pending ruling, so watch for updates.

    The EU AI Act

    The EU AI Act treats AI used in recruitment and candidate evaluation as high-risk. Many articles still say the high-risk obligations begin on August 2, 2026. According to Metaview, Regulation (EU) 2026/1744, adopted July 8, 2026 and in force since July 27, 2026, moved the start date for high-risk hiring obligations to December 2, 2027.

    Systems already on the market before that date are covered only if their design is subject to significant changes afterward. Penalties for breaching the Act’s prohibited practices can reach 7% of worldwide annual turnover or €35 million, whichever is higher. This is a single secondary source for the new date, so confirm it against the Official Journal before you plan around it.

    If you recruit in the EU, the practical impact is on vendors: expect providers to offer risk documentation, transparency information and human-oversight features before the deadline.

    A practical compliance checklist

    1. Inventory every AI tool in your hiring process. Include sourcing rankings, resume screening, chatbots, interview scoring and assessments.
    2. Decide where decisions are automated. A tool that only suggests is different from one that rejects candidates. See our guide to AI candidate screening software.
    3. Give notice. Tell candidates when AI is used, in clear language and early enough to meet local rules.
    4. Keep a human in the loop. Have a person review rejections and be able to override outputs.
    5. Request bias-audit evidence. Ask vendors for audit results or documentation, and get them in writing.
    6. Keep records. Store criteria, outputs and decisions. California requires four years for covered systems.
    7. Offer accommodations. Give candidates a way to request an alternative to automated steps.
    8. Review annually. Laws and deadlines are still changing, and this article will be out of date in months.

    For a broader view of what to automate and what to leave to people, see what AI recruiting agents actually automate.

    Questions for AI recruiting vendors

    • Does the product score, rank or reject candidates, or only organize information?
    • Do you have an independent bias audit, and can we see the summary?
    • What data does the tool use, and does it infer protected characteristics?
    • Can recruiters see and override the reasoning?
    • How long do you store candidate data, and how can candidates opt out?
    • Who is responsible if a regulator asks for records?

    Evaluation matters at the interview stage too. Our guide to AI interview software covers what structured reporting should include. ConnectDevs’ AI interviewer produces a structured report for each interview that a recruiter reviews, but it does not replace your own compliance review of how you use it.

    FAQ

    Is it legal to use AI in hiring?

    Generally yes, but a growing set of laws governs how. Requirements include notice, bias audits, records and human oversight, depending on where candidates and employers are located.

    What is NYC Local Law 144?

    It is a New York City law, enforced since July 5, 2023, that requires employers using automated employment decision tools to have an independent bias audit, publish a summary and give candidates notice.

    When does the EU AI Act apply to hiring?

    Reports indicate high-risk hiring obligations now begin on December 2, 2027, after a 2026 regulation postponed the earlier August 2026 date. Confirm the current date with official sources.

    Do these laws apply to small companies?

    Some do and some do not. California’s regulations apply to employers with five or more employees, while NYC’s law applies to covered employers and agencies using tools for NYC candidates. Check each law’s scope.

    Do I have to tell candidates I use AI?

    In several places yes, including NYC and Illinois. Even where it is not required, clear notice builds trust. Gartner found only 26% of candidates trust AI to evaluate them fairly.

    Who is liable, the employer or the vendor?

    Usually the employer remains responsible, and in California liability can also extend to vendors acting as agents. Ask vendors for contractual commitments and documentation.

    Conclusion

    You do not need to avoid AI to stay compliant. You need to know which tools influence decisions, tell candidates, keep people in charge and keep records. Start with an inventory this month, because the rules will keep moving.

    Want to see how an AI interview report works in practice? Book a demo and walk through it with your own role.

  • AI Recruiting Automation: Which Hiring Workflows to Automate First (and Which to Leave Alone)

    AI Recruiting Automation: Which Hiring Workflows to Automate First (and Which to Leave Alone)

    AI Recruiting Automation: Which Hiring Workflows to Automate First (and Which to Leave Alone)

    AI recruiting automation pays off when you automate the right workflows in the right order. Most teams do the opposite. They start with the flashy step, such as AI screening, and skip the boring ones that save the most time with the least risk.

    This guide gives you a four-factor scoring framework, a workflow-by-workflow table, ROI math with labeled assumptions, and the compliance guardrails that matter in 2026. It is built for TA leaders, agency owners and founders who need a sequence, not a tool list.

    TL;DR

    • Score every workflow on volume, repeatability, trust risk and ease of verification. Automate the high scorers first.
    • Scheduling, outreach follow-ups and ATS hygiene come first. Screening and interviews come later, with human review. Offers and closing stay human.
    • In our illustrative model, four low-risk workflows save about 165 recruiter hours a month. Replace the numbers with your own.
    • Regulators treat tools that filter or evaluate candidates as the risky category. Plan audits, notices and human oversight before you switch those on.

    Table of Contents

    What is AI recruiting automation?

    AI recruiting automation is the use of software, including AI models, to run repeatable hiring tasks with little manual effort, while people keep control of decisions that affect candidates.

    It covers a wide range. Some tasks are simple rules, like sending a calendar link. Others use AI, like ranking profiles or running a structured interview. The difference matters, because the risk grows as the task moves from administration to judgment.

    Adoption is already common. SHRM’s 2025 Talent Trends report found that 51% of organizations use AI in recruitment. So the question is no longer whether to automate. It is where to start.

    If you want a deeper look at what agents can and cannot do, read our guide to what AI recruiting agents actually automate in 2026. This post takes a different angle. It ranks workflows by how safe and how valuable they are to automate.

    The four-factor prioritization framework

    Score each workflow from 1 to 5 on four factors: volume, repeatability, trust risk and ease of verification. A higher total means automate sooner.

    The scores are a judgment call, not a law of nature. Run them with your team and argue about the numbers. The argument is the useful part.

    The four factors

    • Volume. How often does the task happen? 5 means hundreds of times a month. 1 means a few times a quarter.
    • Repeatability. Does the task follow the same steps each time? 5 means a fixed pattern. 1 means every case is different.
    • Trust risk (inverted). What happens to candidate trust if the automation fails? 5 means low risk, because a failure is invisible or trivial. 1 means a failure could harm a candidate or your employer brand.
    • Ease of verification. Can a human check the output quickly? 5 means one glance. 1 means you cannot tell if it is wrong without redoing the work.

    Trust risk deserves extra weight. Gartner surveyed job candidates and found that only 26% trust AI to evaluate them fairly, and 25% trust employers less when AI evaluates their information. Candidates accept AI that schedules their interview. They are far more wary of AI that judges them.

    How to read the total

    • 16 to 20: automate now.
    • 12 to 15: automate with human review on every output.
    • Below 12: pilot carefully or keep human-led.

    Workflow-by-workflow table

    Scheduling, outreach follow-ups and ATS hygiene score highest. Screening and interviews score lowest, because trust risk and verification are hard.

    These scores are our own illustrative judgment for a typical mid-size recruiting team. Adjust them to your context.

    Workflow Volume Repeat-ability Trust risk (5 = low) Verification Total /20 Verdict
    Interview scheduling 5 5 5 5 20 Automate now
    Outreach follow-ups 5 4 4 5 18 Automate now
    Reporting and ATS hygiene 4 5 5 4 18 Automate now
    Candidate enrichment 4 4 4 3 15 Automate, spot-check
    Sourcing 5 3 4 3 15 Automate, review shortlists
    Application screening 5 3 2 2 12 Human in the loop
    First-round interviews 4 3 2 2 11 Pilot with oversight
    Final interviews, offers, closing 2 1 1 1 5 Keep human

    The rollout order, in detail

    Start with administrative workflows where a mistake is cheap and obvious, then move toward judgment workflows only after you have measured results.

    1. Interview scheduling

    Scheduling is pure coordination. It has high volume, a fixed pattern and an instant check: the meeting happens or it does not. Automate it first.

    2. Outreach follow-ups

    Most replies come after a follow-up, and most recruiters stop too early because follow-ups are tedious. Automate the sequence, but make sure it stops when a candidate replies. Nothing erodes trust faster than a “just checking in” email sent to someone who already said no.

    Pilot, our outreach product, runs multi-touch outreach from your own Gmail, stops follow-ups on reply and sorts replies. If you are building this yourself, those two behaviors are the minimum bar.

    3. Reporting and ATS hygiene

    Stage updates, notes, duplicate records and weekly reports eat hours and add no candidate risk. Automate the data entry. Keep a person responsible for the numbers leadership sees. For benchmarks on what to track, see our cost per hire benchmark for engineering roles.

    4. Candidate enrichment

    Enrichment fills in missing profile data, such as role history and contact details. It is repeatable and high volume. Verification is harder, since stale data can look right. Spot-check a sample each week. ConnectDevs offers candidate enrichment for this step.

    5. Sourcing

    AI sourcing finds and ranks candidates faster than manual Boolean strings. It is not a decision about a person, because a recruiter still chooses who to contact. Review every shortlist before outreach. Our guide to finding passive candidates beyond LinkedIn covers the channels. For a tool example, Scout runs natural-language search over a published 800M+ public-profile graph and returns ranked matches with reasoning.

    6. Application screening

    Screening is where volume tempts teams to go all in. It is also where bias, error and regulation concentrate. Use AI to organize and summarize, and let a recruiter make the reject decision. Our walkthrough on how to screen high-volume job applicants shows a workable split.

    7. First-round interviews

    Structured AI interviews can standardize first-round questions and produce transcripts. They are also the most visible AI touchpoint for a candidate. Disclose them, offer an alternative where you can, and keep a human reviewing the results. SAM, the ConnectDevs AI interviewer, runs voice-based structured screening interviews with transcripts and ranked reports. To compare options, see our roundup of the best AI interview software in 2026.

    What to leave alone

    Keep humans in charge of final hiring decisions, offer negotiation, closing, and any conversation where a candidate is making a life choice.

    Those moments run on persuasion, empathy and context. They are low volume, so automation saves little. They carry the highest trust cost if you get them wrong. That combination is a clear “leave alone.”

    Also leave alone anything you cannot verify. If you cannot explain why the tool rejected a candidate, do not let it reject candidates.

    Fraud is a related reason to keep humans involved. Gartner predicts that by 2028, one in four candidate profiles worldwide will be fake. Automated pipelines that never meet a human are easier to game.

    ROI math with illustrative assumptions

    Even conservative assumptions show four low-risk workflows saving roughly 165 recruiter hours a month, but the numbers below are illustrative, not benchmarks.

    Every figure here is an assumption chosen to make the math easy to follow. Swap in your own volumes, time estimates and costs. None of this is a ConnectDevs customer result.

    Assumptions

    • Fully loaded recruiter cost: $60 per hour (illustrative).
    • Team of five recruiters.
    • Automation platform costs: $1,500 per month in total (illustrative).
    Workflow Monthly volume (assumed) Manual time per unit (assumed) Manual hours Time cut (assumed) Hours saved
    Interview scheduling 200 interviews 12 min 40 75% 30
    Outreach follow-ups 1,500 prospects x 2 follow-ups 2 min 100 80% 80
    Application screening (admin only) 600 applicants 4 min 40 50% 20
    Reporting and ATS hygiene 5 recruiters x 13 hours n/a 65 50% 32.5
    Total 245 162.5

    The math

    • Hours saved: 162.5, or about 165.
    • Value of time saved: 162.5 x $60 = $9,750 per month.
    • Less tool cost: $9,750 – $1,500 = $8,250 net per month.
    • Annualized: roughly $99,000.

    Two cautions apply. First, saved hours are only worth money if you redirect them to higher-value work, such as talking to candidates. Second, the screening row assumes AI handles organization and humans make the decisions. If you automate the decision itself, the compliance cost changes the math.

    The strongest payoff is usually speed. Faster scheduling and follow-ups shorten the process, which we cover in our guide to reducing time to hire for technical roles. For a view of output per recruiter, see recruiter productivity and placements per recruiter.

    Compliance guardrails

    The more a tool filters, ranks or evaluates candidates, the more regulation applies. Administrative automation faces far less scrutiny.

    This section reflects our reading of public sources as of October 2026. It is not legal advice. Confirm your obligations with counsel.

    New York City Local Law 144

    NYC requires employers and agencies using automated employment decision tools to have a bias audit within one year of using the tool. They must also publish the audit results and give candidates notice at least 10 business days before use. Enforcement began on July 5, 2023.

    Enforcement is still evolving. A New York State Comptroller audit published in December 2025 reviewed DCWP’s enforcement from July 2023 to June 2025 and found gaps, and DCWP committed to implementing most of its recommendations. Expect scrutiny to rise, not fall.

    EU AI Act

    The AI Act classifies systems used to recruit or select people, including to analyse and filter job applications and to evaluate candidates, as high-risk. The original deadline for those obligations was August 2, 2026. The Digital Omnibus has since been published in the Official Journal, and high-risk obligations now apply from December 2, 2027.

    That delay is not a free pass. Rules on prohibited practices and AI literacy already apply, and high-risk compliance takes time to build.

    A practical guardrail checklist

    • Keep a human decision-maker for every reject and advance decision.
    • Disclose AI use to candidates before it touches their data.
    • Get audit evidence from vendors for any tool that screens or ranks.
    • Log outputs so you can explain a decision later.
    • Test for bias on your own pipeline data, not just vendor claims.
    • Set stop rules, such as halting automation when error rates pass a threshold.

    Smaller teams can follow the same logic. Our piece on AI recruiting software for startups covers how to apply it without a big compliance budget.

    FAQ

    What should I automate first in recruiting?

    Automate interview scheduling first. It has high volume, a fixed pattern, low candidate risk and an instant check. Outreach follow-ups and ATS hygiene come next.

    Should I automate candidate screening?

    Automate the organizing, not the deciding. Let AI summarize and rank, and have a recruiter make the final call. Screening tools can trigger bias-audit and notice rules, so check your local law first.

    How do I calculate the ROI of recruiting automation?

    Multiply monthly volume by minutes per task, estimate the share you can cut, and price the saved hours at your loaded recruiter cost. Subtract tool costs. Then check that the saved time goes to higher-value work.

    Does NYC Local Law 144 apply to scheduling or outreach tools?

    The law targets automated employment decision tools, which are tools that support hiring decisions. Pure scheduling and follow-up sequencing generally do not make those decisions. Confirm the scope with counsel for your specific tool.

    When do EU AI Act rules for hiring tools apply?

    High-risk obligations for recruitment and selection systems now apply from December 2, 2027, after the Digital Omnibus moved the original August 2026 date. Prohibited-practice and AI literacy rules already apply.

    Will AI replace recruiters?

    Not for the work that matters most. AI absorbs repetitive tasks. Recruiters keep relationships, judgment and closing. Our guide to what AI recruiting agents still cannot do goes deeper.

    Conclusion

    Good AI recruiting automation is a sequencing problem. Score your workflows, start with the high-volume, low-risk ones, and add judgment workflows only after you can measure and verify them. Keep people in charge of decisions and closing.

    Run the scoring exercise with your team this week. Then pick the top two workflows and set a 30-day target for each.

    Want to see how this fits your pipeline? Book a demo and we will walk through sourcing, enrichment, outreach and structured interviews on your own roles. ConnectDevs works alongside your ATS and does not replace it.