HiTM: Human in The Middle — Why the Future of AI in Pharma Isn't Fully Automated

There is a quiet war being fought inside every pharmaceutical company, every medical communications agency, and every regulatory submission team on the planet right now.

On one side: the breathless promise of fully automated AI — faster, cheaper, infinitely scalable.

On the other: the stubborn, irreplaceable reality of human expertise — contextual, ethical, accountable.

At CORTiX.io, we refuse to pick a side. Instead, we built something better. We call it HiTM.

What Is HiTM — Human in The Middle?

HiTM, or Human in The Middle, is CORTiX.io's founding philosophy and operational framework. It holds a single, non-negotiable conviction: the most powerful point in any AI workflow is the human standing in the centre of the AI workflow. 

Not before the AI. Not after it. In the middle of it.

HiTM is not a feature toggle or a compliance checkbox. It is the architectural decision — baked into every tool CORTiX.io has ever built — that places a trained, experienced medical mind as the active intelligence through which AI output must pass before it becomes deliverable content.

Compare it to an airline pilot in charge of the airplane, even though the autopilot is on, the flight crew still bear the ultimate responsibility over the plane. You are always the driver.

Why This Matters More Than Ever in 2026

The AI landscape has shifted dramatically. Search engines now reward what practitioners call people-first AI — content and tools that place genuine human expertise above algorithmic velocity. The EU AI Act has made meaningful human control over AI systems a legal obligation in high-risk sectors. And in pharma, the stakes have never been higher: a hallucinated reference in a clinical summary or a misformatted citation in a submission document isn't just embarrassing — it can derail approvals, compromise patient safety, or expose organisations to serious regulatory censure.

This is precisely the environment that gave birth to HiTM.

The broader AI ethics and governance community has been circling this idea for years, using a constellation of terms: human-in-the-loop (HITL), human-centered AI (HCAI), responsible AI, human-AI collaboration, augmented intelligence, AI guardrails, and human oversight. These are all valid, important concepts. But they describe the problem space. HiTM is the answer — at least for the pharma and medical communications world.

The HiTM Spectrum: Where Do You Sit?

The AI governance literature defines three broad models of human involvement with AI systems:

Human-in-Command (HIC) — The human holds ultimate authority and approves every decision. AI plays a purely advisory role. Effective in the most critical, highest-stakes scenarios but slow and resource-intensive.

Human in The Middle (HiTM) — The human is embedded within the AI workflow as an active co-processor. AI handles the heavy lifting — literature retrieval, formatting checks, transcription, abbreviation scanning — while the human applies domain expertise, clinical judgment, and ethical reasoning to shape, correct, and validate outputs in real time.

Human-on-the-Loop (HOTL) — The human monitors AI output and can intervene if something goes wrong, but the system largely runs on its own. Useful for lower-risk, high-volume tasks, but dangerous when applied to regulated content.

For medical writers, regulatory affairs specialists, and pharmaceutical communications professionals, HiTM is the rational operating mode. It captures the efficiency of AI without surrendering the judgment that no algorithm has yet earned the right to replace.

KEYWORD RESEARCH TABLE:

How the World Describes Human-Centric AI

What HiTM Looks Like Inside CORTiX.io

CORTiX.io was built by medical editors, for medical peeps. That is not a marketing line. It is the reason every tool in the platform is engineered around the HiTM principle.

Take CORTIX.io Research. When you retrieve PubMed publications or chat with a manuscript, the AI is not generating conclusions for you. It is surfacing, synthesising, and structuring — so that you, the expert, can draw the conclusion that matters. The human stays in the middle of the knowledge loop.

Take CORTIX.io Editing. Grammar checks, footnote validation, reference format verification, abbreviations audits — these are tasks where AI achieves extraordinary speed and consistency. But every flag it raises, every suggestion it surfaces, passes through your hands before it reaches the page. AI spots the pattern. You decide what it means.

Take CORTIX.io Audio. The medical-trained transcription model diarises speakers and renders spoken science into structured text. But it does not interpret what the medical advisor said, weigh it against clinical context, or decide how it fits your publication plan. You do. Human in The Middle.

And underneath all of this sits a commitment that is rare in the AI-as-a-service world: CORTiX.io does not store your data and does not use it to train AI models. Your clients' confidential data stays confidential — because in a world where human oversight matters, the human's professional obligations must be protected too.

Why "Human in The Middle" Is More Than a Metaphor

In network security, "Man in The Middle" (MiTM) refers to a malicious actor intercepting communication between two parties. We flipped that concept deliberately.

In HiTM, the human is not an intruder in the AI pipeline. The human is the pipeline. The AI sends, the human receives, processes, enriches, and transmits — and only then does the output reach its destination.

This reframing matters because it changes how teams relate to AI tools. HiTM is not about limiting AI. It is about amplifying the human. Every minute the AI saves on formatting checks is a minute the medical writer spends on scientific rigour. Every citation retrieved automatically is a reference the researcher can critically evaluate rather than hunt for. AI handles the mechanical. Humans handle the meaningful.

This is what the growing movement around human-first AI, responsible AI design, and ethical AI in healthcare is reaching for — a model where automation serves human capability rather than substituting for it.

HiTM and the Regulatory Horizon

Regulators are paying attention. The EU AI Act classifies AI systems used in medical and pharmaceutical contexts as high-risk, explicitly requiring that they be designed to support human oversight and allow for human intervention. The U.S. FDA's emerging AI/ML framework similarly centres on the question of human control over algorithmic output.

For any organisation working in drug development, medical communications, or clinical research, this means the compliance question is no longer whether to implement human-centric AI, rather how to implement it at scale.

HiTM is CORTiX.io's answer to that question. It is not a theoretical framework. It is a live, operational philosophy running inside tools used every day by medical writers, MSLs, pharmacovigilance teams, and pharmaceutical communications agencies across Asia-Pacific and beyond.

The Bottom Line: AI Should Work for the Expert, Not Replace Them

The most dangerous misconception in pharmaceutical AI today is that human review is an obstacle to efficiency. It isn't. It is the source of the efficiency's value.

An AI that produces a manuscript in four minutes is impressive. An AI that produces a correct, compliant, submission-ready manuscript in four minutes — because a medical expert was positioned in the middle of the process — is transformative.

That is HiTM.

That is CORTiX.io.

Ready to put the human back in the middle of your AI workflow?

Explore CORTiX.io tools → | Get started for free →

CORTiX.io is an AI-powered SaaS platform for pharma and medical communications, developed by medical editors for medical professionals.