39% of HR Teams Use AI. Most Risk an Audit.

Correction, 3 September 2026: an earlier version of this post attributed an “88% use AI for screening” figure and a “93% plan to expand” figure to SHRM. Neither appears in SHRM’s published research. Every number below is taken from SHRM’s State of AI in HR 2026 (1,908 HR professionals) and its accompanying analysis.

SHRM’s 2026 State of AI in HR report contains two numbers that should concern every HR leader running AI-assisted hiring: just 39% of organisations have implemented AI in their HR functions, and among HR professionals in US states with workforce AI regulations, 57% are not aware of those rules.

The concerning part isn’t the 39%. It’s what the 39% are running, and how few of them could defend it.

The Governance Gap

The same report found that 49% of organisations have policies in place to regulate AI use among employees, and only a quarter believe those policies are future-proof. Among the HR professionals who do know their state’s AI rules, only 12% have taken steps to ensure compliance.

That is the gap: organisations using AI in hiring that have not formally evaluated whether what they’re using is compliant, auditable, or even working as intended.

Most of these tools were purchased in 2024 or early 2025, when the regulatory environment was quieter, the Mobley v. Workday class action hadn’t been certified, and the EU AI Act enforcement date felt distant. The procurement was driven by speed and cost. The governance followed — or didn’t.

What “Agentic AI” Means for This Gap

The April 2026 AI news cycle has been dominated by agentic AI entering the HR operating layer. GPT-5.4’s one-million-token context window has enabled autonomous multi-step workflows. Microsoft has rebuilt significant HR infrastructure around AI agents.

This is relevant to compliance because agentic AI doesn’t just screen candidates — it makes sequential decisions: flagging, scheduling, shortlisting, and sometimes rejecting, with minimal human touchpoints at each step.

The legal frameworks being built around AI hiring — ADEA, FCRA, EU AI Act, NYC Local Law 144 — were designed for a world where AI was one tool in a human-supervised process. In an agentic pipeline, identifying where the human oversight point is requires intentional design.

If your 2024 AI screening tool is now connected to a 2026 agentic workflow, you may have inadvertently created a decision pipeline with no documented human review step. Under California and NYC law, that’s not compliant.

The 3-Question Audit

Here is a 10-minute audit you can run this week on your current AI screening setup:

Question 1: Can you explain any specific rejection? Pick a candidate your AI screened out in the last 90 days. Can you — without asking the vendor — explain in job-relevant terms why they were rejected? If the answer is “the system gave them a low score,” you have an auditability problem.

Question 2: When were your screening criteria last reviewed? If your AI screening tool is using evaluation criteria set at deployment and never revisited, it may be optimising for an outdated job profile. Worse, those criteria may not be documented anywhere. Undocumented criteria cannot be defended in an adverse action dispute.

Question 3: Where is the human in your pipeline? Map your candidate journey from application to shortlist. Identify every point where a human reviews, approves, or overrides an AI decision. If there are segments with no human checkpoint — particularly around rejection — document them and assess regulatory exposure.

What the numbers say to do next

The report also found that by 2025, 73% of HR directors and above were using AI in their work, ahead of managers and individual contributors at roughly 65%. Leadership is adopting faster than the governance beneath it, and SHRM’s guidance to CHROs is explicit: balance innovation with risk mitigation through a principle-based AI governance framework supported by training, audit mechanisms and cross-functional oversight.

Read plainly, SHRM is arguing that the next wave of AI adoption in HR must be supervised, not autonomous. Recruiters as oversight roles, not execution roles. The tools should do more. The humans should be more thoughtful about what the tools are doing.

HireQwik’s design reflects this directly: AI scores + reasoning + HR review dashboard. Every shortlist decision involves a human reviewer looking at structured AI output — not a pipeline that auto-advances candidates without review.

The Competitive Implication

There is a near-term competitive advantage available to HR teams that close this governance gap before their peers do.

The enterprises that can demonstrate auditable, bias-tested, structured AI screening will have a recruiting advantage in markets where candidates are increasingly aware of AI bias risks. They’ll have a compliance advantage when the regulatory environment tightens further. And they’ll have a talent quality advantage because structured screening produces better shortlists.

The 61% who haven’t formally adopted AI — and the slice of the 39% who adopted it without adequate governance — have a narrowing window to get this right before it becomes an enforcement issue rather than a best-practice issue.

The 3-question audit takes 10 minutes. The legal exposure it might surface could cost considerably more to fix later.

Want to see what a structured, auditable AI screening process looks like in practice? Start a HireQwik pilot and we’ll walk you through the evaluation framework, scoring logic, and HR review dashboard.