How AI is changing the HR function

Human resources has long been a function where technology promised more than it delivered. Applicant tracking systems streamlined administration without improving the quality of hiring decisions. HR information systems digitized records without meaningfully changing how people management happened. AI is different in scale and in where it actually changes the work, but it is also different in the new failure modes it introduces.

The organizations that are handling AI in HR well are being specific about where AI helps and where it introduces problems. The ones struggling are treating AI as a general-purpose HR improvement, without accounting for the parts of HR where AI amplifies existing biases, creates compliance exposure, or removes the human judgment that the work actually requires.

Where AI is changing HR operations

Talent acquisition has more AI deployment than any other HR function. Resume screening and initial candidate ranking are the most common applications: AI systems score resumes against job descriptions, filter for qualifications, and rank candidates for recruiter review. The efficiency gains are real. Recruiting teams handling high-volume hiring can process far more candidates per recruiter hour than before.

The compliance exposure is also real. AI-based screening systems have produced well-documented patterns of disparate impact, filtering out candidates from protected groups at different rates than human reviewers would. The legal risk of deploying AI in hiring decisions without rigorous bias testing and ongoing monitoring is significant, and several large employers have faced regulatory action and litigation over AI-driven hiring tools. Organizations deploying AI in talent acquisition without robust evaluation of disparate impact are taking on risk they may not have modeled.

Employee onboarding has become an area of AI investment, particularly for companies with high-volume new hire processes. AI-powered systems handle administrative tasks: document collection, policy acknowledgment tracking, benefits enrollment guidance, and answering common new hire questions. The efficiency gains are meaningful for HR teams processing hundreds of new hires per month. The limitation is that onboarding experiences that are entirely AI-mediated often produce lower engagement and retention than those where new hires have substantive human contact early. The balance between AI efficiency and human connection matters more in onboarding than in most other HR processes.

Performance management is changing more slowly than recruiting. AI tools assist with performance review analysis, helping managers identify patterns across teams, flag inconsistencies in ratings, and surface insights from engagement survey data. But the core of performance management, setting expectations, giving feedback, and making judgments about people’s development and advancement, remains a human activity. AI tools that attempt to automate performance ratings have generally produced outcomes that both managers and employees find less credible than human judgment.

Workforce planning and analytics is an area where AI genuinely adds analytical capability. HR teams can now analyze workforce data at a scale and speed that was not practical before: modeling attrition risk, identifying skill gaps against future business needs, analyzing compensation equity, and forecasting headcount needs. These analyses used to require specialized analysts and weeks of work; they now run faster and are available to HR business partners without data science expertise. The quality of decisions downstream from this analysis depends on how well the people making decisions understand both the analysis and its limitations.

The compliance landscape

AI in HR sits in one of the more complicated compliance environments of any business function. Equal employment opportunity law, data privacy regulation, and an evolving regulatory environment around algorithmic decision-making all intersect in HR AI applications.

Several jurisdictions have enacted or are considering regulations specifically governing AI use in employment decisions. New York City’s Local Law 144 requires bias audits and candidate notification for AI tools used in hiring. The European Union’s AI Act classifies employment-related AI systems as high-risk, requiring substantial documentation and human oversight. Other jurisdictions are following with their own requirements.

Organizations using AI-based tools in talent acquisition need to understand which tools they are using, what decisions those tools influence, and what audit and disclosure obligations apply in the jurisdictions where they operate. Vendors selling AI HR tools have an incentive to characterize their products as decision-support rather than decision-making, which can obscure the actual compliance obligations that apply.

The safe path on HR AI compliance is not avoiding AI; it is deploying AI with documented testing, ongoing monitoring, human review of consequential decisions, and clear records of what role AI played in specific outcomes. This is more work than deploying AI without these practices, but it is substantially less work than managing the legal exposure when an AI-assisted hiring or termination decision is challenged.

What HR vendors are doing

The HR technology market is responding to AI pressure in ways that affect which vendors to trust with strategic HR functions.

Established HR platforms (applicant tracking systems, HRIS platforms, learning management systems) are adding AI features rapidly. The quality varies substantially. Some AI capabilities represent genuine advances; others are surface-level feature additions that use AI terminology without meaningfully changing what the product does. Evaluating AI features in HR platforms requires looking past vendor claims to what the features actually do in practice, what data they train on, and what the vendor’s approach to bias testing and transparency is.

AI-native HR tools are emerging across every HR function. Some are building genuinely new capabilities; others are replicating what established platforms do with a better AI wrapper. The risk with AI-native tools is vendor stability. HR systems hold sensitive data and are embedded in core people processes; a vendor that fails or pivots creates a disruption that is expensive to manage.

The due diligence questions for HR AI vendors are more extensive than for most software categories. Beyond the standard questions about data security and integration, organizations should understand: how was the AI trained and on what data, what bias testing has been conducted and by whom, what human review exists in the decision flow, what the vendor’s approach to regulatory compliance is across the jurisdictions where the customer operates, and what happens to customer data if the vendor relationship ends.

The HR function itself

AI is also changing what HR professionals need to know and do. The transactional and administrative work that has historically occupied a significant portion of HR team time is being automated. HR professionals who have built careers primarily around administrative expertise are facing real displacement pressure from AI and from HRIS platforms that are making administration increasingly self-service.

The HR work that is growing in value is different: expertise in employment law and regulatory compliance, organizational design, change management, workforce planning that connects people data to business strategy, and coaching and development for managers and leaders. These areas require human judgment, relational skills, and domain expertise that AI augments rather than replaces.

HR functions that are investing in building these capabilities, and restructuring around them rather than around administrative work that AI is absorbing, are likely to emerge from this period with more strategic influence than they entered it with. HR functions that are waiting to see how AI plays out, without building the capabilities that will matter when it does, are likely to find themselves competing for resources in a function whose traditional value proposition has been substantially automated.