The messy business of managing people at work: Is AI the solution?

TEXT ANALYSIS: ILO Article on AI in HR

TEXT START:

"AI and work: The messy business of managing people at work: Is AI the solution?"

THE DISSECTION

This is a policy-level comfort artifact — a technically rigorous paper that performs institutional seriousness about AI in HR while surgically avoiding the one question that renders the entire exercise irrelevant: what happens when the function being managed no longer requires managed humans?

The piece methodically dissects implementation failures — bad objectives, garbage data, opaque algorithms, stakeholder exclusion — and concludes that better-designed AI systems, built with meaningful human participation, can preserve or improve HR decision-making. This is institutional coping at its most sophisticated. The ILO is an organization built around the premise of human labor. Of course it will find the problem in bad design rather than in the design itself.

THE CORE FALLACY

The paper assumes the central dysfunction is poor implementation rather than structural displacement. It treats AI in HR as a tool optimization problem: get the objectives right, feed good data, ensure transparency, involve stakeholders, and you get better human-centered HR.

This is the same category error the paper itself commits in critiquing the "growth mindset" word-frequency algorithm. You cannot operationalize a genuine human quality through reductive metrics. The ILO commits an equivalent error in the other direction: you cannot humanize a displacement mechanism by making it more accurate.

The "paradox of automation" framing at the opening — each solution creates a new problem — is gestured at but then immediately defanged. Berg does not follow the logic to its terminus: the paradox doesn't resolve; it accelerates. The ILO paper wants to negotiate better terms with a tide.

HIDDEN ASSUMPTIONS

1. Human labor remains the substrate of the employment relationship. The entire framework assumes that organizations will continue to need, hire, manage, evaluate, compensate, and schedule human workers. No acknowledgment that AI in recruitment is the first step toward eliminating recruitment as a function entirely.

2. Stakeholder participation is a meaningful corrective. The case studies of co-design (BT technicians, multinational hybrid model) are presented as replicable solutions. They are boutique interventions. At scale, competitive pressure will always favor the employer who automates fastest and cheapest, not the one who spends two years iterating with HR professionals.

3. "Meaningful" data is recoverable. The paper correctly identifies that keystrokes and balloon games don't measure job performance. But it assumes the problem is finding better proxies. The DT frame says: the meaningfulness of any proxy collapses when the underlying productive function itself becomes automatable.

4. HR as an institution is worth preserving. The paper implicitly treats improved AI-assisted HR as a goal. But HR's core function — managing the human employment relationship — is precisely what AI automates. The paper is arguing for better management of an activity in terminal decline.

SOCIAL FUNCTION

Transition Management / Prestige Signaling — dressed in the language of empirical rigor.

The ILO occupies a specific institutional niche: it must appear to be doing serious work on AI and labor while not threatening the framework of global labor standards that defines its existence. This paper performs exactly that balance. It acknowledges the harms, calls for better governance, advocates stakeholder participation, and ultimately leaves the structural dynamic untouched.

The paper's footnotes are excellent. Its conclusions are theater.

THE VERDICT

The ILO has produced a meticulous autopsy of the wrong corpse. It is documenting the failure modes of AI-assisted human resource management as if the problem were bad design rather than intentional displacement. The framework of "objective + data + programming" is a sophisticated way to miss the point: these systems are not failing to achieve human-centered goals; they are succeeding at replacing the humans those goals were designed to serve.

The paper concludes that investment in better AI-HR systems is "necessary and worthwhile." Under the DT lens, this is equivalent to recommending better management practices for passengers on a sinking ship — technically not wrong, structurally irrelevant, and terminally behind the rate of water intake.

The messy business of managing people at work is not a problem AI can solve or fail to solve. It is a function AI is eliminating. The ILO is advising on how to better decorate a room that is being demolished.