From HMI to MHI: When the Human Becomes the Interface

There's a term every IT professional knows: HMI, Human-Machine Interface. Since the 1980s, the entire discipline of software design has been built around a simple question: how do you enable a human to talk to a machine? Buttons, menus, forms, touchscreens — each generation produced its own translations.

But something has changed. And this change is so natural that we haven't named it yet.

The Flow Has Reversed

In a traditional HMI, the human initiates. They click, type, select. The machine executes. The interface is a translator of human intent into machine language.

With generative AI, it's the machine that proposes. It drafts an analysis, suggests a structure, generates a report. And the human? They read, evaluate, correct, validate. They become the filter between what the machine produces and what is actually relevant.

We haven't created a better human-machine interface. We've created a machine-human interface.

What This Changes in Practice

At Quanthome, we live this inversion every day. When a real estate fund manager queries our AI about rental yields for a portfolio, they don't fill out a form. They ask a question. And the AI doesn't return a raw data field — it produces a structured analysis, with context, comparisons, and recommendations.

The manager's job is no longer to find the information. It's to judge whether the analysis is accurate, complete, and relevant to their decision.

This is a fundamental shift in competence. We're moving from mastering the tool to mastering judgment.

The Human in the Loop, Not at the Controls

The English-speaking world has a term for this: Human-In-The-Loop, or HITL. The human is no longer the pilot giving orders to the machine. They are a component in a loop that the machine orchestrates. An essential component — the most important one, even — but a component nonetheless.

In the context of Swiss real estate, this loop takes a very concrete form. Quanthome's AI can cross-reference yield, vacancy, renovation, and ESG compliance data across thousands of buildings. No analyst can do this work manually in a reasonable timeframe. But no AI can decide on its own whether a building deserves an investment — because that decision involves fiduciary responsibility, knowledge of the local market, and judgment that remains irreducibly human.

The loop works precisely because each party does what the other cannot.

The New Bottleneck

For decades, the productivity bottleneck was production capacity: how many reports can an analyst write per week? How much data can they process?

That's no longer the problem. AI can produce in seconds what used to take days. The new bottleneck is judgment capacity: how many AI proposals can a human correctly evaluate per day? With what depth? With what level of critical thinking?

This is a question our clients are beginning to ask. And it's a healthy question. Because it's no longer about technology — it's about human competence in the age of AI.

Knowing How to Say No

If the human has become the interface, then their core competence is no longer knowing how to formulate the right query. It's knowing how to say no.

No, this analysis is incomplete. No, this figure doesn't match the reality of the Geneva market. No, this recommendation ignores a regulatory factor. It's in this "no" that human added value resides. Not in production — the machine handles that. But in filtering, correction, and reframing.

At Quanthome, we design our tools with this reality in mind. Our AI doesn't present its results as truths. It presents them as proposals — with sources, reasoning, and explicit limitations. Because a good machine-human interface is one that facilitates judgment, not one that replaces it.

A Vocabulary Yet to Be Invented

We don't yet have the right vocabulary to describe what's happening. "Human-machine interface" assumes an active human and a passive machine. "Artificial intelligence" assumes an intelligence comparable to our own. Both terms are legacies of an era when the machine was an inert tool.

Perhaps the term "MHI" — Machine-Human Interface — will never catch on. But the inversion it describes is already here. And the professionals who understand it early will have a decisive advantage: not because they master the technology better, but because they better master their own role in the loop.

In Swiss real estate as elsewhere, the future doesn't belong to those who ask the best questions of AI. It belongs to those who best know how to judge its answers.