AI conversation has shifted dramatically over the last year. In many cases, we’ve moved from AI pilots to AI that is really here in production. But how and where its truly driving value is far more complex than most AI-enabled businesses let on. Today, we face an AI adoption picture that’s not binary (did we or didn’t we), but fragmented.
Most programs can be defined by ambitious strategy at the top, patchy deployment in the middle and inconsistent impact on the ground. In HR, where jobs revolve around people, this uneven adoption isn’t just a technical issue. It’s a people problem.
The ‘say-do’ gap
In boardrooms, the language around AI has shifted from if to how fast. But recent research reveals a stark maturity gap. In a 2025 industry roundup, 88% of HR tech leaders reported no significant ROI from their AI initiatives, despite positive adoption indicators — a sign that implementation isn’t delivering measurable business results yet.
Similarly, a global workforce survey from EY found that while roughly nine out of ten employees use AI tools at work, only about 28% of organizations translated those deployments into high-value outcomes.
What this tells us is that while HR organizations may deploy AI tools, they’re not yet achieving the integrated process change necessary to shift key outcomes such as talent acquisition efficiency, retention or workforce planning. This disconnect mirrors broader enterprise patterns that include rapid experimentation without aligned operating models or measurement frameworks.
Shadow AI: The manifestation of a broader enterprise challenge
One of the most persistent barriers to controlled adoption is shadow AI. Research shows that 80% of companies experience the use of shadow AI at work, with HR among the functions most likely to expose sensitive employee data to insecure platforms. According to IBM, only 37% have policies to manage AI or detect shadow AI.
In HR, this can take shape in several ways:
- Recruiters using public generative AI to rewrite job descriptions or screen resumes
- Managers inputting employee performance data into ungoverned tools for quick insights
- Teams experimenting with AI assistants to manage employee questions without compliance vetting.
Shadow AI isn’t inherently malicious, and in many scenarios, its use (when paired with human scrutiny, not blind trust) is fine. It is, however, an indicator of a demand outpacing formal capability, and the risk is real.
Unauthorized use opens the door to data leakage, compliance violations and unethical decision-making. This can be especially apparent in functions like talent acquisition, where fairness is deeply important.
The real bottleneck: Change management
Another clear pattern emerged in 2025: AI adoption is outpacing training. Multiple sources show that while adoption among HR professionals is rising (surpassing 70%, according to a recent survey), only a fraction have received job-specific training on how to use AI effectively.
Our partner, Workday, has been one of the rare companies to excel in this area. HR leaders succeeded in getting nearly 80% of employees to use AI tools by investing deeply in tailored, multi-stage learning journeys rather than one-off sessions (Fortune).
This insight is critical for CIOs. AI isn’t the bottleneck; human preparedness is. Organizations that invest upfront in layered skills development and role-specific support create far more impact than those that treat AI like another software rollout.
Beyond training, cultural support matters deeply. A global PwC survey underscored that trust, clarity and workforce alignment around AI shape whether adoption sticks or stalls. This comes from the top down.
The rise of technical HR roles
A positive evolution, yet also a sign of uneven adoption, is the emergence of hybrid technical titles within HR. Titles like People Analytics Engineer, HR AI Product Manager and Talent Technology Lead are becoming more common. It’s a sign of the times, as organizations try to bridge the gap between traditional and modern HR.
As organizations move beyond experimentation, they’re realizing that neither generic data science teams nor traditional HR roles can fully own AI-enabled systems. Deep HR domain knowledge and technical fluency are now critical. These roles sit at the intersection of human judgment and machine intelligence, translating AI outputs into decisions leaders can trust and act on.
The important lesson here is that effective AI adoption isn’t just about deploying the technology itself, but surfacing new kinds of expertise that can operationalize it responsibly and at scale.
Key AI takeaways for HR leaders
AI adoption is uneven in HR, even where use cases are well-defined, data is rich and ROI levers are clear. This is a sign that similar challenges exist across industries. Here are several takeaways that span across enterprises adopting AI:
- Governance needs to balance control with speed. Shadow AI is a product of unmet need and rapid innovation. Governance frameworks must enable safe experimentation, not just restrict tools.
- Change management trumps technology choice. The tools themselves are rarely the root of the problem. Investing in human learning, process redesign, and clear incentives leads to far more value than chasing the most advanced models.
- Roles must evolve. Traditional org charts don’t account for hybrid thinkers who can navigate human complexity and algorithmic logic. Identifying, recruiting and developing those skills (and fast) is essential to stay competitive.
AI in HR is shaping the future of work in real ways. But that impact isn’t evenly distributed across organizations or use cases. The companies that will win in the long run are those that built capability, culture and governance frameworks to deploy strategic, responsible AI programs.
This article is published as part of the Foundry Expert Contributor Network.
Want to join?
Read More from This Article: The state of AI in HR: Big promises, uneven reality
Source: News