The Specific Event

WIRED reported this week that workers at three major publishing houses are raising internal alarms about the quiet adoption of large language models across core editorial and marketing functions. According to staff accounts, LLMs are being used to draft publicity materials, generate cover art concepts, write back cover copy, and compose outgoing emails. More pointedly, junior staff are being pressured by executives to actively champion the technology internally, even when those same junior staff bear the greatest professional risk from its substitution effects. This is not a story about AI adoption. It is a story about organizational power, information asymmetry, and what happens when the people least protected by an institution are tasked with legitimizing the tool most likely to displace them.

The Coercion Structure Hidden Inside "Championing"

The detail that stands out most to me is the directive for junior staff to become internal advocates. This is not a neutral training initiative. When an organization asks its most precarious workers to publicly endorse the technology being used to restructure their roles, it is doing something specific: it is manufacturing consent while insulating senior leadership from visible accountability for the decision. Rahman (2021) describes a structurally similar dynamic in gig platform governance, where algorithmic control is rendered invisible partly because workers are induced to narrate their own dependence as autonomy. The publishing case involves a human management chain rather than an algorithm, but the underlying logic is identical. The coercive element is obscured by the language of participation and initiative.

Why "Awareness" of the Threat Does Not Produce Power

Staff at these publishing houses clearly know what is happening. The WIRED report could not exist otherwise. Workers are aware that AI is being used, aware that their functions are being targeted, and aware of the asymmetry in who is being asked to carry the reputational and advocacy burden of adoption. Yet awareness does not translate into effective organizational response. This maps directly onto what algorithmic literacy researchers have documented at the platform level. Kellogg, Valentine, and Christin (2020) demonstrated that workers in algorithmically managed environments develop sophisticated awareness of the systems governing them without necessarily developing the structural leverage to respond effectively. The awareness-capability gap is not a cognitive failure. It is a structural condition produced by the distribution of formal authority within the organization.

In the publishing context, the gap between knowing and being able to act is sustained by employment precarity. Junior staff understand precisely what is being asked of them and why it is problematic. Their awareness is not the missing variable. What is missing is any organizational mechanism that converts that awareness into upward influence on governance decisions. Schor et al. (2020) argue that dependence and precarity are not incidental features of platform work but constitutive of how platforms extract labor on favorable terms. The same logic applies here. Precarity is not a background condition in this story. It is the active mechanism that makes coercive championing viable as a management strategy.

The Information Asymmetry Is Bidirectional and Deliberate

There is a second layer worth examining. Executives pushing AI adoption for publicity copy and back cover text are almost certainly not the people evaluating whether the outputs are editorially adequate, legally compliant, or consistent with author relationships. Junior staff are positioned closest to those quality and relationship concerns, which means they hold material information that leadership formally lacks. In organizational theory terms, this is a classic principal-agent information gap. What is unusual here is that management's response to that gap is not to solicit the expertise of the agents who hold it but to instruct those agents to suppress their expert judgment in favor of institutional enthusiasm for the technology.

This has direct implications for failure risk. Asonye (2021) documented in a nursing context that organizational factors - specifically those governing how competence and concern are communicated upward - are primary predictors of whether organizations successfully avert consequential failures. The structural suppression of ground-level expertise does not eliminate the information that expertise represents. It eliminates the organizational channel through which that information would otherwise travel.

What This Reveals About Governance, Not Technology

It would be easy to frame this story as being about AI adoption speed, or about the cultural fit between LLMs and editorial craft. I think that framing misses the operative mechanism. The technology is downstream of a governance decision about whose judgment counts and who bears the cost of being wrong. Publishing houses asking junior staff to champion AI are not primarily making a statement about artificial intelligence. They are revealing something about how authority, risk, and information are distributed inside those organizations. That distribution predates AI and will persist after the current adoption cycle regardless of what the tools produce. The AI is the occasion for the revelation, not its cause.

References

Kellogg, K. C., Valentine, M. A., and Christin, A. (2020). Algorithms at work: The new contested terrain of control. Academy of Management Annals, 14(1), 366-410.

Rahman, H. A. (2021). The invisible cage: Workers' reactivity to opaque algorithmic evaluations. Administrative Science Quarterly, 66(4), 945-988.

Schor, J. B., Attwood-Charles, W., Cansoy, M., Ladegaard, I., and Wengronowitz, R. (2020). Dependence and precarity in the platform economy. Theory and Society, 49(5), 833-861.

Asonye, C. C. (2021). Organizational factors associated with nurses' competence in averting failure to rescue in acute care settings. Journal of Client-Centered Nursing Care, 7(1). https://doi.org/10.32598/jccnc.7.1.358.1