When the Data is Perfect and the Humans Are Not

I walked out of the senior executive meeting knowing we had not succeeded. Not because our analysis was flawed. Not because the numbers were wrong. We had spent weeks on a rigorous pay parity study — gender, levels, skill set, experience level — and the findings were unambiguous. Gaps existed and in certain pockets, they were significant. The budget to fix the problem existed too.

And yet, when we presented the data to the executives who held the decision, something invisible pushed back. The conversation drifted. Objections appeared from nowhere. Should we really pay a junior employee with high skills as much as someone with years of experience? Is the data accounting for everything? The meeting ended without resolution.

I realised, standing in the corridor afterward, that my team and I had made a classic mistake. We had assumed that better data produces better decisions. We had forgotten that decisions are made by human beings — with emotions, biases, and decades of preconditioning that no spreadsheet can touch. We were trying to solve the human conditioning with analytics and that does not work all the time.

I share this story not as a cautionary tale about pay parity — though it is that too. I share it because it is the most important lesson I carry into the age of artificial intelligence: data is not wisdom, and intelligence is not understanding.

Right now, across boardrooms in Mumbai, Bengaluru, Singapore, and New York, a particular anxiety is spreading. Call it AI FOMO — the fear of missing out on artificial intelligence. Organisations are rushing to implement AI agents for recruitment, onboarding, performance management, and employee engagement. The pressure to adopt is enormous. The implicit message is clear: if you are not moving fast, you are falling behind.

And the technology deserves some of this excitement. Large language models — Claude, ChatGPT, Microsoft Copilot — have made genuine, extraordinary progress. Their ability to synthesise information, draft communications, and analyse people data already exceeds what most employees can do manually. This is not hype. It is happening, and HR teams who ignore it entirely do so at their peril.

But in our rush toward adoption, we are skipping a question that only experience can teach you to ask: Is this tool actually equipped to operate in the reality of my organisation?

These models are not plug-and-play. An AI onboarding agent deployed in an American technology company should not communicate the same way as one deployed in a Japanese manufacturing firm or an Indian services company. The tone, the warmth, the pace of relationship-building, the degree of directness — these are not cosmetic preferences. They are the difference between an employee feeling welcomed and an employee feeling processed by the agent.

The implementation gap — between a model's raw capability and its contextually calibrated deployment — is one of the most underappreciated challenges in AI adoption today. And it is a challenge that requires human-centric thinking, not just technical expertise.

But even perfectly implemented AI hits a deeper wall. And this is where we need to challenge the FOMO most directly.

There is a meaningful difference between an AI agent answering "What is my parental leave policy?" and an AI agent creating the experience of being genuinely heard. Employees — especially in their first weeks with an organisation — are not simply downloading information. They are forming an emotional contract. They are asking, beneath every interaction: Does this place see me? Do I belong here?

An agent can simulate warmth. It can personalise responses. It can be available at 2am when a new hire in a different time zone has a question. But it cannot yet replicate the felt experience of human presence — the sense that another conscious being is paying attention to you with genuine care.

And then there is the deepest question, the one that the technology industry is racing toward without fully understanding: What happens when AI becomes truly general? When machines match human intelligence across every domain?

The debate around Artificial General Intelligence — AGI — is fierce, and notably, there is no consensus even among those building it. What is conspicuously absent from this conversation is serious engagement with the question of consciousness. Every definition of AGI measures what a system does. None grapples with what it is.

The philosophical traditions of India have thought about consciousness longer and more rigorously than any AI thought leader. Advaita Vedanta teaches that consciousness is not a product of intelligence — it is beyond intelligence. Krishnamurti asked whether a mind conditioned by accumulated knowledge can ever truly understand. These are not mystical abstractions. They are precise challenges to the assumption that scaling compute will eventually produce wisdom.

If consciousness cannot be engineered, then the human at work is not threatened with obsolescence — they are elevated in their irreplaceability. The HR leader becomes not the administrator of process, but the guardian of what no algorithm can replicate: the soul of the organisation.

I am not arguing against AI. I am arguing against the FOMO driven abdication of human judgement and contextual reality.

The pay parity story I opened with did not end in failure. The data eventually got through — not because we improved the analysis, but because we changed how we had the conversation. We spent more time with the humans in that room. We understood their fears. We addressed the preconditioning, not just the pay gaps.

That is what great HR teams have always done. And it is what AI, for all its extraordinary capability, cannot yet do alone.

The question for every leader reading this is not "Are we using AI?" It is "Do we understand what AI cannot do — and are we protecting that space?"

That space is where your cultural reality lives. Guard it carefully.

(This article is written by Avadhesh Dixit, Former Chief Human Resources Officer. This is an opinionated article; EPN has nothing to do with this editorial.)