The loudest conversation about AIin HR is often about what technology will eliminate.

Meidan Keidaris asking a different question: What if AI could make HR more human?

As Global Director of HR at Lumana, Meidan Keidar built eight AI agentsin her first week using them to organize benefits information, policies, communications and other HR infrastructure.

But she isn’t building an automated HR department. She is building the infrastructure that could allow humans to do better HR.

AI is not the HR department

Keidar described employees and leaders as HR’s customers. Her philosophy was simple: when an employee has a meaningful HR issue, she wants that person to be able to reach an HR business partner, not be pushed into an AI system because it is cheaper. That distinction matters.

McKinseyrecently reported that roughly two-thirds of today’s HR activities could be fully automated or delivered through automation by 2030. The report also emphasized that humans would continue to own policy, outcomes and exceptions in many automated workflows.

The implication is bigger than automation. If AI takes over more of the transactional work, the uniquely human parts of HR become more important, not less.

The power of treating agents like employees

Keidar’s approach to AI was surprisingly human. She described treating her agents almost like direct reports. She trained them, gave them information, and expected them to perform specific responsibilities. The difference, she joked, was that agents could not get promoted.

That mindsetreveals a potentially important shift in how leaders may manage AI: not as software they occasionally open, but as a workforce they intentionally design, supervise and govern. And that creates an entirely new leadership responsibility.

Who does what? Which decisions can an agent make? What information can it access?

When must a human intervene? How do multiple agents work together without creating conflicting or unintended outcomes?

The future of the HR operating model

Keidar’s agents were not isolated experiments. She was creating a knowledge center containing policies, culture, operational information and communication systems that could continue functioning even as people moved in and out of the organization. That knowledge center hasn’t stayed static.

The architecture keeps evolving, splitting into more specialized agents as real usage reveals where it needs to — the same way a growing team hires a specialist instead of overloading a generalist. That is organizational infrastructure. And it could become one of the most important applications of agentic AI in HR.

McKinsey has described the emerging workplace as a partnership between people and AI agents, while Deloittereported in 2026 that only 21 percent of surveyed organizations had mature governance for agentic AI.

That gap should get HR leaders’ attention. Building agents may be relatively easy. Building an organization that knows how those agents should operate is much harder.

Automate the infrastructure. Elevate the human.

The most interesting part of Keidar’s approach was not the eight agents. It was what she refused to automate.

She used AI to organize knowledge, support communication, and strengthen the operating infrastructure of HR. But she wanted humans to remain at the center of the moments that require judgment, empathy and relationships.

That’s the real opportunity for HR in the AI era. Not fewer humans. Better use of humans.

The rise of the AI workforce architect

The future HR leader won’t be the person who automates the most work. It will be the leader who knows exactly what should be automated and what should become even more human because of it.

Keidar has a name for this discipline: AI Workforce Architect. It’s not an HR generalist who happens to use AI, but a practitioner who deliberately designs how work divides between people and agents. It’s a role most companies don’t know they need yet.

Keidar is already building toward that future.

This was published on Inc.com first.

Click the Inc.com link to read the full article: https://www.inc.com/netta-jenkins/she-built-8-ai-agents-in-one-week-what-she-didnt-automate-matters-more/91407664