Artificial intelligence (AI) is in your applicant tracking system, your HR information system, your benefits platform and your payroll tool. It has been bolted on, bundled in and integrated into almost every piece of HR technology you use. And yet, for many HR leaders, something feels off.

The gap between what AI promised and what it’s actually delivering is real and the ones doing the work feel it most. According to the Society for Human Resource Management (SHRM), 40% of organizations using AI in HR cite concerns about security and privacy and 21% point to employee resistance and lack of trust as a barrier to wider adoption. Among those yet to adopt AI at all, the most common reason is not cost but simply not knowing which tools actually fit their needs.

AI is everywhere. Quality is not.

The pace of adoption has been striking. SHRM data shows 43% of organizations used AI in HR tasks in 2025, up from 26% the year before. Industries most exposed to AI saw productivity growth jump from 7% (2018–2022) to 27% (2018–2024), while the least-exposed barely moved. The direction of travel is clear.

But adoption is not the same as value. Most AI being deployed in HR today is built on general-purpose large language models which are broad, fast and impressively fluent — but trained on public internet data with no grounding in the specific regulatory environment, validated compensation benchmarks or people practices that HR professionals depend on. Ask a generic AI tool about state-level employment legislation or whether a benefits package is competitive and you may get a confident, well-structured answer… that may also be wrong, outdated or both.

For HR professionals, that’s a liability waiting to happen.

The trust deficit is well-founded

HR.com research puts the concern bluntly:

• 53% of HR professionals cite vulnerability to bias as a top concern in AI-assisted recruiting
• 47% worry about liability for unintentional discrimination
• 44% report low trust in AI-driven decisions outright
• 41% flag inaccuracy and hallucinations as a key risk

These are not the worries of technophobes. They are the rational concerns of professionals who are being asked to rely on tools they cannot fully audit, drawing on sources they cannot fully verify, in a function where getting it wrong has consequences for real people.

Nowhere is this pressure felt more sharply than in mid-market organizations. ADP data shows that 44% of businesses with 50 to 999 employees have implemented or are piloting generative AI but they are doing so with a fraction of the resources of larger enterprises. These teams face the same complex challenges — compensation strategy, compliance, talent acquisition, workforce planning — yet with smaller teams that are already stretched; checking the answers AI gives can also mean more time than is saved with AI in the first place.

What actually makes AI work for HR

The AI tools that earn genuine trust in HR share a common set of characteristics. They are domain-specific, not generic. They draw on curated, verified content rather than the open internet. They are kept current as legislation, market data and best practice evolve. And they are designed to support human judgment, not replace it.

This is the principle behind Expert. Unlike the AI bolted onto existing HR platforms as an afterthought, Expert is built from the ground up for HR professionals combining the speed of generative AI with WTW’s proprietary research, exclusive benchmarking data and deep expertise in HR, compensation and benefits. Responses are grounded in curated WTW content and trusted third-party sources.

That means, when an HR professional uses Expert to benchmark a salary, query U.S. employment legislation or summarize a policy, the answer reflects what WTW actually knows and not what the internet happened to say at some point in the past.

The question worth asking

The real value of AI in HR is well-documented. SHRM reported that 89% of HR professionals at AI-using organizations report time savings or efficiency gains. The potential to recover nearly 400 hours per year per HR coordinator from routine task reduction alone makes a compelling case. But that value only materializes when the AI can be trusted.

The question worth asking has moved beyond whether your HR stack includes AI. It almost certainly does. The question is whether any of it is built on the insight, expertise and rigor HR professionals actually need and whether it will still be accurate tomorrow.

Stats have been sourced from Society for Human Resource Management