The Human Is in the Loop. But Is the Organisation Really Safe?

For years, HR leaders have relied on increasingly sophisticated enterprise platforms to help manage payroll, talent, workforce planning, performance, employee data and organisational decision-making.

These systems create enormous value. They bring structure, consistency, information and efficiency to processes that would be extraordinarily difficult to manage manually across large, complex organisations.

And as AI becomes embedded more deeply into enterprise technology, those capabilities are expanding.

But I keep coming back to a question:

Does sophisticated technology sometimes create an illusion of organisational safety?

The question is not whether enterprise platforms have controls. The question I am asking is different:

What responsibility still belongs to the organisation using the technology?

The dashboard is not the decision

My HR career has taught me something very simple about enterprise systems: the output is only part of the story.

Whether the platform is an ERP, HRIS, workforce analytics system or increasingly an AI-enabled system, leaders still need to understand what sits behind the output.

What information went in? Was the data accurate? How was the system configured? What assumptions were embedded? Which internal policies apply? Which jurisdiction applies? Does a collective agreement change the answer? Who has authority to make the decision? What role did technology actually play? And who remains accountable for the consequence?

A polished dashboard can make something look complete. That does not necessarily mean the organisational decision behind it is complete.

Human-in-the-Loop can mean many things

This becomes particularly important when organisations say: “Don’t worry. There is a human in the loop.”

I support meaningful human oversight. But the presence of a human somewhere in a workflow does not automatically tell us whether the oversight is meaningful.

Where is the human in the process? What information does that person receive? Does that person understand what the AI system has done? Does that person have the expertise and authority to question the recommendation? Can the person stop or reverse an action? Is the person exercising judgement, or simply approving what the system has already placed in front of them? And who owns the outcome?

Does the organisation have an architecture of meaningful human accountability?

What leaders may not see beneath the dashboard

When configuration begins influencing human decisions. A system can support performance management, talent identification, workforce planning or organisational restructuring. But leaders should understand how system configuration influences the information placed in front of decision-makers.

When organisational context is missing. Enterprise platforms operate across organisations with different policies, collective agreements, jurisdictions, workforce structures and legal obligations. A general capability may not automatically understand the specific context of the organisation using it.

When human oversight becomes a checkbox. A human approval button does not necessarily demonstrate human judgement. If the person approving the outcome does not understand the evidence, cannot meaningfully challenge the recommendation, or lacks the authority to alter the outcome, the organisation should ask whether Human-in-the-Loop has become procedural rather than substantive.

Vendor responsibility does not eliminate organisational responsibility

The argument I am making is not that enterprise technology vendors are unsafe.

The stronger question is:

Have organisations mistaken strong vendor controls for sufficient organisational governance?

A vendor can build responsible controls into its product. The organisation still has decisions to make about data quality, configuration, permissions, internal policies, decision rights, human judgement, monitoring, escalation and accountability.

Governance is not something that happens once when the system is purchased. It lives with the system.

This is also why I am going deeper

On October 20, 2026, I begin MIT Professional Education’s Applied Agentic AI for Organizational Transformation programme.

Alongside it, I am deliberately developing proficiency in Microsoft Foundry.

Not because I am trying to become an AI engineer or software developer. And not because decades of organisational leadership experience suddenly stopped mattering when AI arrived.

Quite the opposite.

I want to understand the technology well enough to connect what it can do with what organisations should authorise it to do.

That means going beneath the executive summary. What is the agent being asked to do? What knowledge is it using? What has actually been tested? What can go wrong? What technical controls exist? What remains a human decision? Who is accountable? How will we know whether value is actually being created? When should we stop? And where does my expertise end and the expertise of engineering, cybersecurity, privacy, Legal or another specialist need to go deeper?

Proficiency should improve collaboration with specialists, not diminish their contribution.

My cup is deliberately half full

I do not enter this learning journey empty. I bring senior HR and enterprise leadership experience, ER/LR depth, organisational transformation, judgement, relationships and thousands of hours supporting leadership development, change and transformation.

That is one half of the cup.

I am deliberately protecting the other half. For what I do not yet understand. For terminology whose Microsoft-specific meaning may be different from mine. For technology that may challenge assumptions built through experience. For evidence that may prove part of my current thinking incomplete.

And for the possibility that what I believe about Human-in-the-Loop today may look different after I have spent the next several weeks learning, experimenting, testing and questioning.

That is not uncertainty I need to hide. That is the point of learning.

Meeting People Where They Are. Creating Meaningful Impact.