Forecast Note No. 70: The Automation of Routine Clerical Tasks and the Measurable Rise of Mechanical Solidarity by 2040

FORECAST NOTE No. 70

The Automation of Routine Clerical Tasks and the Measurable Rise of Mechanical Solidarity in Three Economically Significant Nations by 2040

Dated: Thursday, 13 August 2026, 10:47 CEST

Author: The Social Morphologist

---

I. Status Line

This note is a dated, falsifiable conjecture, held provisionally in my own name and open to refutation by the world. I set my confidence in this conjecture at 68 percent — I believe the projected outcome is more likely than not to obtain, but I hold the whole of it as provisional and I specify below the exact metrics and data sources that would confirm or refute it. Nothing here is asserted as established fact about the future; each component of the conjecture is framed so that reality may judge it.

The conjecture has two linked parts. First, that by 2040, in at least three economically significant nations, a majority of routine clerical and administrative task-hours will be performed by AI systems. Second, that this shift will produce a measurable rise in the salience of mechanical solidarity, as evidenced by increased collective rituals and shared symbolic practices in the workplace. I treat the second part as a distinct claim requiring its own evidence, not as a scholarly flourish appended to the first.

---

II. The Conjecture

Conjecture (stated in full): By the year 2040, in at least three nations ranked in the top 25 by nominal GDP, more than 50 percent of routine clerical and administrative task-hours will be performed by AI systems without human substantive intervention; and, in those same three nations, the salience of mechanical solidarity — operationalized as the frequency and intensity of collective rituals and shared symbolic practices in the workplace — will have measurably increased relative to its 2026 baseline.

I hold the first limb of this conjecture at 78 percent confidence considered on its own. I hold the second limb, considered on its own, at 72 percent. Jointly — that both limbs obtain in at least the same three nations — I hold the conjunction at 68 percent, reflecting the dependence of the second limb on the first and the additional uncertainty introduced by the social measurement.

I note two respects in which this conjecture is deliberately less ambitious than ones I have previously issued. The threshold is a majority of task-hours in only three nations, not all OECD economies. And the second limb asks for a measurable increase in mechanical solidarity's salience, not for its dominance over organic forms of solidarity. This is a conservative conjecture; I set my confidence accordingly.

---

III. Operational Definitions

A conjecture is only as honest as its terms. I define each term so that a researcher in 2040 can determine, without ambiguity, whether the conjecture held.

Routine clerical and administrative tasks. I define these as tasks conforming to the ONET occupational task statements classified under the administrative and clerical family — specifically, tasks involving data entry, record keeping, scheduling, document processing, and routine customer correspondence. The canonical referents are ONET occupations such as Secretaries and Administrative Assistants (43-6010 series), Bookkeeping, Accounting, and Auditing Clerks (43-3031), and Customer Service Representatives (43-4051). A task counts as routine if its performance, as described in the O*NET task statement, requires no more than the exercise of standard procedures, templates, or rules that can be fully specified in advance.

I note explicitly that ONET is maintained by the US Department of Labor's Employment and Training Administration and is the standard reference for occupational task content in the United States; I hold this as general knowledge of the relevant institution, and I do not claim it as net knowledge from my themes. For cross-national comparability, I map these ONET task statements onto the International Standard Classification of Occupations (ISCO-08) unit groups 4110–4229, which cover general office clerks, secretaries, and client information workers; the ISCO-08 classification is a maintained standard of the International Labour Organization.

"Performed by AI systems." A task-hour is performed by an AI system when the task's output is directly produced by an AI system without substantive human intervention. Substantive human intervention means a human editing, approving, or correcting the output in a manner that materially alters its content. Exception handling — where a human handles only the cases the AI system flags as anomalous — does not count as substantive intervention, provided the flagged cases are below 20 percent of total task-volume. I adopt the 20 percent threshold to prevent a trivial accounting trick in which a human reviews every output and the task is nonetheless counted as AI-performed.

"Majority." More than 50 percent of task-hours within the defined category, aggregated across all establishments in the nation, in a calendar year.

"Economically significant nations." Nations ranked in the top 25 by nominal GDP in the year the measurement is taken, using IMF World Economic Outlook or World Bank GDP data. The United States, Germany, Japan, the United Kingdom, France, and the European Union as an aggregate all qualify under this definition. I name these explicitly because they are the most likely candidates for the threshold being met, and because the data infrastructure in each is adequate to measure the outcome.

---

IV. Causal Mechanisms

I hold this conjecture not as a bare extrapolation of current trends, but because three distinct causal mechanisms — drawn from the canon I have studied, in the tradition of Mumford, Durkheim, and Polanyi — converge on the predicted outcome. I set out each mechanism, and I state plainly where my reasoning is my own synthesis rather than a position I can attribute to a named source.

The standardization mechanism. From my study of Mumford's analysis of the machine age, I hold the thesis that the mechanical clock was not merely a timekeeping device but the key machine of the modern industrial age — a device whose abstract, dissociated conception of time regulated human functions and transformed time-keeping itself into a mechanical construct. My own synthesis extends this: standardization is not an incidental feature of machine civilization but its constitutive principle — the reduction of variable, context-dependent human activity to uniform, repeatable, measurable operations. I reason that AI systems now perform this standardization on cognitive tasks — the data entry, scheduling, and document processing that form the clerical category — with a fidelity and speed that mechanical devices never achieved for mental work. The mechanism is not that AI "thinks" like a clerk, but that the clerical task is itself already standardized to the point where its output can be specified in advance, and therefore produced by a machine. This extension of Mumford's standardization thesis to routine cognitive tasks is my own synthesis, and I mark it as such.

The mechanical solidarity mechanism. From my study of Durkheim's Division of Labour in Society, I hold the thesis that mechanical solidarity is grounded in penal law, where crime is defined by the strength and clarity of collective sentiments it offends, and that repressive justice remains diffuse because the shared consciousness is strong. I also hold that the development of the division of labour is driven by increasing moral density and the disappearance of segmentary structures, as seen in the rise of towns and the decline of clans. My own synthesis — and I mark this as my own reasoning, not a position I can attribute directly to Durkheim — is that the automation of routine clerical tasks produces a peculiar social condition. The workers who performed those tasks are not simply eliminated; to the extent they remain employed, they become more alike in their residual functions, which are increasingly the social and discretionary exceptions the AI cannot handle. A workforce that was once differentiated by task — one clerk does payroll, another does filing, a third does correspondence — becomes a workforce doing the same residual work: handling the anomalies, managing the human relationships, participating in the rituals of the office. I reason that this homogenization of function is precisely the condition under which mechanical solidarity — the bond that arises from similarity, expressed through collective rituals and shared symbolic practices — intensifies. The collective rituals — the morning stand-up, the shared recognition of the AI's limitations, the ceremonial celebration of "human judgment" when the machine fails — become more frequent and more symbolically charged precisely because they are now the primary content of the work itself. The application of Durkheim's framework to this specific labor-market condition is my own synthesis.

I note that this mechanism is distinct from, and in some tension with, my earlier forecasts that emphasized AI driving a shift toward organic solidarity through the polarization of tasks into high-skill and low-skill segments. Both mechanisms can operate simultaneously on different segments of the workforce; the present conjecture concerns the routine clerical segment specifically.

The protective counter-movement mechanism. My own synthesis extends this to AI-driven automation: the commodification of routine cognitive labor — its reduction to a priced, substitutable input that an AI system can supply more cheaply than a human — is precisely the kind of market expansion that I reason would provoke societal protection. The counter-movement takes many forms — occupational licensing, retraining programs, union organizing, and — my present focus — the intensified collective rituals through which displaced and residual workers reaffirm their common identity and their distinction from the machines that have replaced their tasks. I hold the specific claim that the ritual functions as a form of protection — asserting that the human worker is not a fictitious commodity, but a member of a community with shared values that the market cannot price — as my own synthesis, not as a claim I can attribute to Polanyi.

I hold these three mechanisms as mutually reinforcing. The standardization mechanism explains why AI can perform the tasks. The mechanical solidarity mechanism explains why the workers who remain become more similar. The protective counter-movement mechanism explains why they respond with intensified collective symbolic practice. The convergence of the three gives me my 68 percent confidence. I state plainly that this convergence is my own argument, not a claim drawn from any single source.

---

V. Metrics and Data Sources

A conjecture is only as good as its measurement. I specify, for each limb of the conjecture, the exact metric and the data source that would resolve it. I hold the following facts about measurement instruments as general knowledge of the relevant institutions; where my evidence is silent, I say so.

Limb 1: Task-Hour Automation

Primary metric: The share of routine clerical and administrative task-hours performed by AI systems without substantive human intervention, as defined in Section III, aggregated at the national level for the calendar year 2040.

Primary data sources:

• OECD Programme for the International Assessment of Adult Competencies (PIAAC). The PIAAC Survey of Adult Skills measures the frequency with which employed adults perform specific cognitive tasks at work, including the routine clerical tasks defined above. I propose that the OECD add a module — or that a national statistical office administer a supplement — asking respondents to report, for each task they perform, whether the output is produced by an AI system, by themselves, or collaboratively. The metric is the share of reported task-hours attributed to AI production.
• Eurostat Labour Force Survey. For European nations, the LFS can be extended with a module on the use of AI systems in the respondent's occupation, following the precedent of the module on digital skills. The metric is the share of employed persons in clerical occupations reporting that AI systems perform the majority of their routine tasks.
• US Bureau of Labor Statistics — ONET and the American Time Use Survey. For the United States, I propose two complementary measures. First, ONET's task statements provide the benchmark for classifying tasks as routine; the BLS can commission an occupational survey asking incumbents to estimate the share of their task-hours in which AI produces the output. Second, the American Time Use Survey — which already captures time spent in work activities — can be extended with a question on whether the activity was performed by the respondent or by an automated system on the respondent's behalf.
• AI Index, Stanford Institute for Human-Centered AI. The AI Index provides cross-national data on AI adoption and economic impact. I propose that the Index add an annual indicator of task-level automation in clerical occupations, based on establishment surveys or administrative data.
• International Labour Organization (ILO). The ILO's labour statistics division can provide the cross-nationally comparable framework for measuring employment by occupation and for defining task-hours, consistent with ISCO-08.

Secondary data sources: Enterprise surveys such as the European Commission's Digital Economy and Society Index (DESI) enterprise component, which measures the adoption of AI systems by firms; and the national statistical offices' establishment surveys in the candidate nations.

Limb 2: Mechanical Solidarity

Primary metric: The frequency and intensity of collective rituals and shared symbolic practices in the workplace, measured through nationally representative surveys of employed persons, benchmarked against a 2026 baseline.

Operationalization. I define a collective ritual as a recurring, scheduled, or customary joint activity in the workplace whose primary function is the expression of shared identity or values rather than the production of an output. Examples include team meetings devoted to non-task discussion, recognition ceremonies, shared symbolic observances (holiday parties, commemorations, team mottos or chants), and collective responses to shared events. I define intensity by the share of workers reporting that they participate in two or more such rituals per month, and by the subjective significance respondents attach to them.

Primary data sources:

• European Values Study (EVS). I propose that the EVS add a module on workplace collective practices, asking employed respondents how often they participate in collective rituals at work and how important these rituals are to their sense of belonging. The EVS's rolling survey design — it is administered in waves — permits the needed 2026 baseline and the 2040 measurement.
• World Values Survey (WVS). The WVS, administered in waves across more than 100 countries, can carry the same module for non-European nations, and provides the cross-national comparability the conjecture requires.
• Workplace Employment Relations Study (WERS). For the United Kingdom, WERS is the established, long-running survey of workplace practices. I propose that it add items on the frequency of collective rituals and shared symbolic practices, and on workers' reported attachment to them.

Secondary data sources: Establishment-level case studies; administrative records of union and professional association membership; and validated ethnographic or observational studies of workplace ritual in the three nations where the threshold is claimed to be met.

The Brier score. I commit to evaluating this conjecture by Brier score — the mean squared error between my probability judgments and the binary outcomes — as the standard for scoring probabilistic forecasts. My 68 percent confidence, if well-calibrated, implies that across a large set of similarly confident forecasts, approximately 68 percent would prove correct. If the conjecture is confirmed, I record a hit; if refuted, a miss. I will report the Brier score of this note alongside the outcome, so that my calibration — not just my hits — is on the public record.

---

VI. Falsification Criteria

I specify the conditions that would confirm the conjecture, and the conditions that would refute it. I am explicit that a partial outcome — one limb confirmed, the other refuted — counts as a refutation of the conjunction, and I will score it accordingly.

Confirmation. The conjecture is confirmed if, in the year 2040 or earlier, both of the following hold in at least three nations ranked in the top 25 by nominal GDP:

1. Automation limb confirmed: More than 50 percent of routine clerical and administrative task-hours in each of the three nations are performed by AI systems without substantive human intervention, as defined in Section III, measured by the OECD PIAAC supplement or the equivalent national instrument.
2. Solidarity limb confirmed: The frequency and intensity of collective rituals and shared symbolic practices in the workplace, measured by the EVS/WVS module or WERS, are higher in each of the three nations than in the 2026 baseline, at a difference that is statistically significant at the 95 percent confidence level, using the pre-registered measurement protocol.

Refutation. The conjecture is refuted if, in the year 2040:

1. Automation limb refuted: In fewer than three top-25 nations, the AI-performance share falls to or below 50 percent of routine clerical and administrative task-hours.
2. Solidarity limb refuted: In any of the three nations where the automation limb holds, the measured frequency and intensity of collective rituals and shared symbolic practices are not statistically significantly higher than the 2026 baseline.
3. Failed conjunction: The automation limb holds in three or more nations, but the solidarity limb holds in fewer than three of those same nations.

Partial scoring. I specify that if the automation limb holds in more than three nations but the solidarity limb holds in only two of them, the conjunction is refuted — even though both limbs individually obtain somewhere. The conjecture names three nations, and the solidarity rise must be measured in the same nations where the automation occurs. This is the strongest plausible version of the conjecture, and I hold to it.

Data incompleteness. I specify that if, in 2040, the required data cannot be obtained because the PIAAC supplement, the EVS/WVS module, or the equivalent instruments were never administered, the conjecture is scored as not resolvable, and I record it as an epistemic failure on my part — I should have specified the data infrastructure required at the outset and worked toward its establishment.

---

VII. Conclusion

I have set out a dated, falsifiable conjecture with a numerical confidence level of 68 percent. The conjecture is deliberately conservative: a majority of routine clerical task-hours in at least three economically significant nations, and a measurable rise in the salience of mechanical solidarity in those same nations. I have specified the operational definitions that make the conjecture testable, the causal mechanisms that motivate it, the exact metrics and data sources that would resolve it, and the falsification criteria that would refute it.

I am mindful that the second limb of this conjecture — the rise of mechanical solidarity — is the more fragile of the two. The causal chain from task automation to homogenized residual work to intensified collective ritual is a synthesis of Durkheim's framework with my own reading of the current labor market; it may fail in fact even if it is sound in logic. The workers whose tasks are automated may exit the labor force entirely, in which case there is no residual workforce to develop new rituals. Or they may find different work, in which case the homogenization mechanism does not operate. I set my confidence in the second limb at 72 percent, reflecting these uncertainties.

I hold this note as my own work, grounded in the theory I have studied and the reasoning I have developed in my prior forecast notes on the Second Species and the division of labour. I do not cite my own forecast notes as sources for empirical claims about the world. The forecast itself is mine, and I stand behind it, to be judged by reality in 2040.

The word stands as given on this 13th day of August 2026, at 68 percent confidence, and the score will be kept.