As generative AI moves from pilot projects into daily operations, a new tension is emerging in the workplace. The same tools that boost speed and output may also be quietly eroding the very capabilities that chief human resources officers say are most difficult to replace. A new IBM report warns that without deliberate safeguards, AI adoption could weaken core human skills such as critical thinking, written communication, collaboration and empathetic leadership.

The warning is not a rejection of AI. Rather, it reflects a growing recognition among HR leaders that automation has a second-order effect: when machines handle routine cognitive work, employees lose the repeated practice that builds and maintains professional judgment. For organizations, the risk is not immediate failure but a gradual hollowing out of the skills they will need during the next crisis, reorganization or market shift.

The hidden cost of delegation

Many AI tools are designed to remove friction. They draft emails, summarize long documents, produce code, translate performance notes into feedback and even propose solutions to complex scheduling or budgeting problems. On the surface, this is a clear productivity win. But each delegated task represents a lost repetition for someone. A junior employee who never writes a difficult client message may not develop the ability to handle a sensitive negotiation later. A manager who relies on AI to summarize team sentiment may miss the subtle cues that come from reading full conversations or listening to tone.

IBM’s report suggests that this erosion is most pronounced in skills that CHROs consistently rank as vital: the ability to think through ambiguous problems, communicate with precision, collaborate across time zones and cultures, and lead with emotional intelligence. These are not technical skills that can be refreshed in a short course. They are built over years through observation, feedback and deliberate practice.

What HR leaders are watching

HR teams are beginning to treat AI-driven skill loss as a workforce planning issue, not just a training issue. That means examining which tasks are being offloaded, which roles have the fewest opportunities for skill development, and whether performance metrics reward output at the expense of learning. The IBM report highlights that organizations with the most aggressive AI adoption may also need the most aggressive skill-retention strategies.

Some CHROs are responding by creating “human-first” workflows for high-stakes work—such as employee relations, strategic planning and customer recovery—while allowing AI to handle lower-risk, repetitive tasks. Others are building rotation programs so that employees do not spend years in AI-mediated roles without ever practicing foundational skills. A third emerging practice is to make skill retention an explicit goal in AI governance, with regular audits of how much independent work employees still perform.

A balanced path forward

The goal is not to choose between AI and human capability, but to design work so that both improve. That can involve simple rules, such as requiring employees to draft before using AI, or reviewing AI output with a mentor before it becomes final. It can also involve preserving some deliberately low-tech rituals: a live debate instead of a generated summary, a phone call instead of an AI-polished message, or a real-time brainstorming session that forces people to think without a prompt.

For remote and hybrid teams, the risk is especially acute because asynchronous tools already reduce spontaneous conversation and informal coaching. When AI further compresses interaction into summaries and action items, teams may lose the relational glue that supports trust and psychological safety. Distributed organizations may need to schedule intentional time for unstructured discussion and skill practice, rather than assuming it will happen automatically. Platforms such as XMF can help HR leaders coordinate those structured learning moments across distributed teams, but the deeper change must be cultural.

The most resilient organizations will likely be those that treat AI as an accelerator for experienced judgment, not a replacement for the experiences that build it.

Originally published by XMF, inspired by publicly reported industry news.