SHRM 2026: 52% of Companies Leave HR Out of AI Strategy

Correction, 3 September 2026: an earlier version of this post attributed a “67% cite lack of awareness”, a “92% of CHROs” and a “50% plan autonomous screening” figure to SHRM. None of them appears in SHRM’s published research. Every SHRM number below is taken from State of AI in HR 2026 and its April 2026 analysis.

52% of organisations do not involve HR in AI strategy or vision. Only 28% have HR leading AI upskilling, the one area where HR should obviously own the work.

SHRM’s State of AI in HR 2026, one of the most comprehensive surveys of the space, covering 1,908 HR professionals across 138 distinct tasks, carries a finding that should reset how we talk about AI adoption in hiring: the barrier isn’t data, budget, or legal risk. It’s ownership.

More than half of organisations (52%) do not involve HR in AI strategy at all. Leadership of AI initiatives typically sits with IT, legal or cross-functional teams. Even in upskilling, where HR should play the central role, only 28% say HR is leading the effort, against 29% led by cross-functional teams. SHRM’s own reading: this lack of ownership risks AI implementations that are disconnected from workforce strategy and employee experience. The single largest barrier to good AI adoption in HR in 2026 is that the people who understand the workforce are not the people choosing the tools.

There’s an addendum that makes this sharper: 57% of HR professionals working in AI-regulated US states aren’t aware of the policies governing their use of AI in hiring. Among those who are aware of the policies, only 12% have taken steps to ensure compliance.

This isn’t a technology problem. It’s an ownership problem.

What HR Leaders Actually Need to Know

The SHRM data draws a clear picture: adoption is real but early (39% of organisations use AI in their HR functions, another 23% elsewhere in the business), HR leaders are the heaviest personal users (73% of HR directors and above by 2025), but the people who will actually evaluate, procure and operate these tools are mostly not in the room when the AI strategy is set. That is a knowledge gap by design.

That gap creates two failure modes:

Failure mode 1: Over-trusting AI. Teams that don’t understand the technology’s limits deploy it without governance. They set up automated rejection flows without human review. They use opaque scoring systems they can’t explain. They become the Mobley v. Workday defendant.

Failure mode 2: Under-trusting AI. Teams that don’t understand the technology’s capabilities continue screening 3,000 candidates manually, burning HR bandwidth on work that takes 18 hours instead of 2. They miss the April–June peak season window. They lose to competitors who figured it out.

The fix for both failure modes is the same: specific, practical knowledge about what AI screening can do, what it can’t do, and how to evaluate any vendor claiming to offer it.

The 5-Question AI Screening Discovery Sheet

These are the five questions every HR team should be able to answer — either about their current AI screening tool or about any vendor they’re evaluating. We’ve designed them to surface the real picture quickly, without requiring a technical background.

Question 1: What does the AI actually assess — and what does it not assess?

What you’re looking for: A clear, specific answer. “Communication quality, role understanding, and response coherence” is a real answer. “Overall candidate fit” or “cultural alignment” is a vague answer that should prompt follow-up.

Why it matters: AI screening tools vary widely in what they actually measure. Resume parsers measure keyword match. Chat-based tools measure engagement and response speed. Voice-based tools can measure communication clarity, follow-up coherence, and anti-scripting performance. These are genuinely different things — and only one of them tells you whether a candidate can do a communication-heavy job.

What good looks like: The vendor can name the specific dimensions being assessed and explain how each one produces a score or recommendation.

Question 2: Who makes the final decision — the AI or a human?

What you’re looking for: Explicit confirmation that no candidate receives a rejection communication before a human HR professional has reviewed the AI’s recommendation.

Why it matters: This is the Mobley v. Workday question. Automated rejection with no human review is the central mechanism in the leading AI hiring bias class action in US federal court. Even if your company is India-only, this question determines whether your process is defensible.

What good looks like: A defined workflow with documented human review at every decision tier — not just for edge cases, but as standard operating procedure.

Question 3: Can you see and explain the scoring rubric?

What you’re looking for: A document or dashboard showing exactly how the AI generates its recommendation. Dimensions assessed, how follow-up questions trigger, how scores combine into a tier.

Why it matters: If you can’t explain to a candidate or a regulator why they received a No Go, you have a black-box problem. The EU AI Act — enforceable August 2, 2026 for high-risk HR applications — explicitly requires transparency in automated decision-making.

What good looks like: HR can read the rubric, understand it without technical help, and explain it in plain language.

Question 4: What does the AI do when a candidate gives a scripted-sounding answer?

What you’re looking for: Evidence of dynamic follow-up questioning — not a static script.

Why it matters: As covered in our AI resume crisis analysis, 67% of HR leaders in Robert Half’s 2026 survey say AI-generated applications are already slowing their hiring. A voice screening tool that asks three static questions and scores the answers is not meaningfully more robust than a keyword filter. The anti-scripting moat is real-time adaptive probing.

What good looks like: The AI generates follow-up questions based on what the candidate just said — probing the specific claim rather than moving to the next pre-set question.

Question 5: What does the full audit trail look like?

What you’re looking for: Complete logs of every interaction — AI prompts, candidate responses, scoring decisions, timestamp of candidate disclosure, and the human review record.

Why it matters: EU AI Act compliance, Mobley-style litigation defence, and internal quality review all require the same thing: a complete, exportable record of what happened in every candidate interaction.

What good looks like: Logs retained for a defined period, exportable on demand, covering both the AI’s side and the candidate’s side of the conversation.

How to Use This With Your Team

These five questions work in three contexts:

Vendor evaluation: Send them in writing to any AI screening vendor before a demo. Vendors who can’t answer all five in specific, documented terms have a compliance gap.

Internal audit: Run them against your current AI screening setup, if you have one. If you can’t answer Question 3 or Question 5, you have work to do before August 2.

CHRO briefing: Use the SHRM data (52% of organisations leave HR out of AI strategy, 57% unaware of regulations) to frame why this matters to leadership, and use the five questions to show HR has a practical response.

The Ownership Gap Is Closeable

The SHRM data is a problem, but it’s a solvable one. HR teams that are left out of AI strategy aren’t opposed to AI: 73% of HR directors and above already use it in their own work. They need a concrete way to earn the seat.

These five questions are that starting point. They don’t require you to understand large language models or EU regulatory Annexes. They require you to ask your vendor specific questions and evaluate whether the answers are real.

See how HireQwik answers all five — we’ll walk you through the rubric, the audit trail, and the human review workflow in 20 minutes.