How Should HR Leaders Think About Low, Medium, and High-Agency AI?

For the past two years, HR leaders have focused on one question: Should we adopt AI? By 2026, that question will be outdated. The real challenge now is not whether to use AI, but how much authority AI should have inside the organization.

Not all AI is the same. Some tools help employees work faster. Others can run workflows. A new generation of “agentic AI” can plan, act, and optimize with minimal human prompting. That means HR leaders need a smarter framework for adoption - one based on AI agency levels.

What Is AI Agency?

An AI agency refers to how independently an AI system can operate, make decisions, and take action.

Think of it in three levels:

Low-Agency AI = Assists humans

Medium-Agency AI = Executes tasks within rules

High-Agency AI = Acts autonomously toward goals

This lens helps HR leaders decide where AI adds value—and where human judgment must remain central.

Low-Agency AI: The Safe Starting Point

Low-agency AI works like an assistant. It responds to prompts, creates drafts, summarizes data, or helps employees complete tasks faster.

Examples in HR include:

• Writing job descriptions

• Drafting policies

• Summarizing interviews

• Creating training content

• Rewriting internal communications

This is already delivering measurable value. SHRM’s 2026 research found that most HR professionals report medium or high productivity gains from AI tools. (SHRM)

For HR leaders, low-agency AI should become the baseline. It reduces administrative burden and frees teams to focus on higher-value work such as coaching, culture, and workforce planning.

Medium-Agency AI: Where ROI Accelerates

Medium-agency AI goes beyond assistance. It can take action within defined guardrails.

Examples include:

• Scheduling interviews automatically

• Routing employee helpdesk tickets

• Screening resumes with oversight

• Personalizing onboarding journeys

• Flagging attrition risks

• Recommending learning paths

This is where many organizations will see the biggest operational returns. It reduces cycle times, improves employee experience, and scales HR support without linear headcount growth.

However, medium-agency AI also introduces real risks – bias, opaque decision-making, and employee distrust if systems are not transparent.

HR must ensure that workflows remain explainable and auditable.

High-Agency AI: High Reward, High Responsibility

High-agency AI is the newest frontier. These systems can plan, coordinate across tools, and pursue goals with limited supervision.

Potential HR use cases include:

• Workforce planning simulations

• Internal talent marketplace management

• Strategic hiring forecasts

• Organization redesign modelling

• Personalized AI career coaches

This is powerful – but it will be dangerous if misused.

Recent executive surveys show many companies are adopting AI faster than they can govern it, with 80% saying they would likely fail an AI governance audit. (Axios)

That is a warning sign.

High-agency AI should never be given unchecked control over promotions, layoffs, compensation, or disciplinary decisions. These are human accountability zones.

The Rule HR Leaders Need to Follow

The higher the people impact, the lower the acceptable AI autonomy.

For example:

• Drafting policy copy = Low agency

• Interview scheduling = Medium agency

• Succession planning insights = Medium with review

• Compensation decisions = Human-led

• Terminations = Human-led only

AI can support decisions. It cannot own them.

Why HR Must Lead This Shift

AI governance is not just an IT issue. It is an HR issue.

Because HR already owns:

• Fairness

• Trust

• Change management

• Workforce design

• Leadership behaviour

• Employee experience

As AI becomes embedded into work, HR leaders must define the rules of engagement: where AI assists, where it executes, and where it stops.

Final Thought

AI will not be leading the future of HR. However, HR leaders who understand AI better than everyone else will be leading the future.

The winning organizations will know when AI should assist, when it should operate, and when it should never be left alone.

In the years ahead, the competitive advantage will not be AI adoption.

It will be an AI authority design.