Inventors of Problems

Every large organization contains a department that grows during hiring freezes. It has no revenue, no margin, no unit economics and no natural size, and it is the only function in the building whose output is consumed almost entirely by itself. It has been renamed three times in forty years, which is the surest available sign that something is wrong with the underlying thing, since functions confident of their contribution do not require rebranding. Personnel became human resources, human resources became people operations, people operations became talent, and at no point did the renaming follow a demonstrated improvement in the quality of anybody's working life.

The complaint that follows is not that the people in these roles are bad people. That version of the argument is popular, satisfying and analytically useless, because it cannot explain why the pattern reproduces itself identically across firms, sectors and continents staffed by entirely different individuals. Something that consistent is not a personnel problem. It is a structural one, and the structure was built deliberately, by executives who wanted what it produces.

Begin with the arithmetic that governs everything else. A department that cannot point to revenue must justify its headcount by some other measure, and the only measures available to it are activity and risk avoided.

Activity is easy to produce. You can always run another survey, publish another framework, add a stage to the review cycle, require another certification, or convene a working group on the findings of the last working group. None of this requires anyone's permission, because none of it appears as a line item anyone defends. The cost lands as time, distributed thinly across thousands of people who have no mechanism for refusing it and no forum in which to price it.

Risk avoided is better still, because it is unfalsifiable. Nobody can demonstrate that the lawsuit which did not happen would have happened, and so any expenditure justified by its absence is immune to review. A function whose value is measured in counterfactual disasters has discovered the perfect business model: unlimited demand, no accountability, and a customer who cannot cancel.

The order Rodentia takes its name from the Latin for to gnaw, and the defining feature of its members is dental rather than dietary. A rodent's incisors never stop growing. They are open rooted, they advance continuously through the animal's life, and they must be worn down against something or they curve back on themselves until the animal can no longer eat. The gnawing is therefore not a response to hunger. A well fed rodent gnaws exactly as much as a starving one, because the behavior is maintenance of the apparatus rather than pursuit of a meal. Deprive it of anything worth chewing and it will chew the cage.

This is the most accurate available description of a staff function whose headcount grows independently of the work. The output is not a response to demand. It is what the teeth require, and the volume stays constant whether or not there is anything to eat.

The predictable consequence is that problems get invented. Not maliciously, and usually not consciously. It is simply what a system does when growth requires demand and demand can be generated internally at no cost. Engagement scores decline slightly, so a program is required. A manager gives blunt feedback, so a communications standard is required. Two people in a company of nine thousand behave badly, so nine thousand people complete a ninety minute module. Each intervention is defensible in isolation. In aggregate they constitute an enormous, unbudgeted tax on the productive time of the firm, levied by a department that has never once been asked to demonstrate a return.

The remarkable thing about the last decade of corporate people programs is not that they were political. It is that they were, by the standards of any other capital allocation decision, evidentially unserious.

The research on mandatory diversity training has been unflattering for a long time, and the most cited work in the area, published in the Harvard Business Review in 2016 by two sociologists who had spent thirty years assembling data across hundreds of firms, found that the compulsory version frequently produced no improvement and sometimes produced backlash and worse representation outcomes. The same body of work found that things which do help are mostly boring and structural: mentorship, rotational programs, actual accountability for managers who hire and promote. The industry took the finding, kept selling the training, and quietly ignored the rest.

Implicit bias testing enjoyed a comparable run. The instrument's ability to predict individual behavior is weak, its test-retest reliability is poor, and the meta-analytic work indicates that interventions which shift the score do not reliably shift the conduct. This was known and published while procurement departments were still writing checks for firm-wide rollouts.

Consider what that combination means. A staff function persuaded boards to commit hundreds of millions of dollars and tens of millions of employee hours to interventions whose supporting literature was, at best, contested and, in the mandatory case, actively negative. Any operating executive who proposed a capital program on that evidentiary basis would be asked to leave the room. The programs survived because they were not, in fact, being evaluated as interventions. They were being purchased as insurance and displayed as signals.

Then came the part nobody in the field likes to discuss.

Between roughly 2020 and 2022, a great many large firms announced ambitious commitments, appointed executives with new titles, published targets and produced reports. Between roughly 2023 and 2025, a substantial share of those same firms cut the teams, retired the titles, removed the targets from their filings and revised the language, in several prominent cases within a single quarter of the political weather changing.

Take the two announcements together and the conclusion is unavoidable. If the commitment was correct in 2021, abandoning it in 2024 was a serious failure of principle, and if it was incorrect in 2021, it was adopted without analysis in the first place. Both readings describe an organization that was never reasoning about the substance. It was reading the room, and it will read the room again.

That is the actual objection to what gets called wokeness in the corporate setting, and it is not a complaint about values. It is a complaint about sincerity. A firm that adopts a moral position because it is currently cheap, and drops it the moment it becomes expensive, has not held a moral position at all. It has run a marketing campaign in the vocabulary of conscience, and it has done so using the working hours of employees who were required to attend.

The people who suffer most from this cycle are the ones who took it seriously, including, conspicuously, the intended beneficiaries. Being hired into a role created for a press release, and discovering in the third year that the role was a line item in a budget nobody intended to defend, is a specific and avoidable humiliation. Cynical adoption is not kindness. It never was.

The cleverest member of the order is worth a longer look, because it explains the mechanism rather than merely describing it.

Beavers are rodents, the largest in North America, and they are genuine engineers. The dams they build create wetlands, moderate floods, recharge groundwater and support an enormous amount of life that would not otherwise be there. Nobody sensible disputes their competence.

What triggers the building is the interesting part. In a series of classic experiments, the trigger was isolated: not the sight of a gap, not a rising water level, but the sound of running water. Play a recording of a flowing stream through a loudspeaker and a beaver will attempt to dam the loudspeaker. It will do this on dry land. It will pack mud and sticks around a machine that is impounding nothing, in a place where no water has ever flowed, and it will do so with the same care and the same competence it would bring to an actual breach.

The animal is not stupid, and this is the point that matters. Over a long time in an environment where that sound reliably meant a leak, responding to the proxy rather than to the underlying condition was cheaper and faster than assessing each situation on its merits. The heuristic was correct until somebody introduced a speaker.

A large organization is full of speakers. An engagement score ticks down two points. A manager sends a blunt message that gets forwarded. A competitor announces a policy and a headline follows. An anonymous comment appears in a survey. Each of these is the sound of running water: a signal that once correlated with a real breach, arriving now with no breach attached. And the response is immediate, competent, and thorough. A program is designed. A standard is drafted. A training is scheduled. Mud is packed around a loudspeaker with great professional skill.

Watch what actually gets dammed and the pattern is unmistakable. The interventions cluster around whatever is audible rather than whatever is costly. Nobody builds anything in response to the best engineer in the firm quietly deciding, over eighteen months, that the place is not serious. That produces no sound at all until the resignation letter, and by then the water is somewhere else.

Here is the part the critics usually miss, and it is the part that matters most, because it identifies who is actually responsible.

Almost every pathology attributed to the people function is a service being purchased by executives who want the outcome and not the exposure.

A senior manager knows an employee should be let go. Doing it requires a conversation, a judgment he must own, and a small risk that he is wrong. Instead he initiates a performance improvement plan, a documented process of predetermined outcome and six weeks' duration, which converts a two minute act of professional courage into a bureaucratic procedure with a paper trail and someone else's signature at the bottom. He has not avoided the decision. He has purchased deniability, and the department that sold it to him is the same one he complains about at dinner.

Multiply that transaction by every difficult conversation in a large firm and you have explained the entire apparatus. Compensation bands exist so that no manager must defend paying one person more than another. Calibration sessions exist so that no individual owns a rating. Approval matrices exist so that no single name appears next to a decision. Anonymous surveys exist so that nobody has to say the thing to a face. The function did not impose this on unwilling leaders. It was built to their specification, because they wanted a mechanism that produced decisions without decision makers.

Any critique that stops at the department is therefore a critique that lets the buyer off. The department is a supplier. The demand comes from the top.

Honesty requires the other half of the ledger, and it is not small.

Payroll runs, or nobody eats. Benefits get administered, and when someone's child is seriously ill the quality of that administration is the difference between a catastrophe and a manageable year. Immigration filings are made correctly or your best engineer is deported. Employment law across a dozen jurisdictions is complied with, and the cost of failure is real money. Genuine harassment cases arrive, and they must be investigated by someone competent, discreet and unafraid, which is a difficult skill possessed by a minority of practitioners and worth paying a great deal for.

Every one of those is a real service with an identifiable customer and a legible failure mode. Notice that they are also the least glamorous things the function does, the parts least likely to appear in a conference talk, and typically the parts staffed most thinly.

The dysfunction is not the existence of the function. It is the layer built on top of the load-bearing work, which grew because nothing constrained it, and which now consumes the majority of the budget and nearly all of the attention.

The remedy is unexciting and entirely available, and it consists of subjecting the function to the same questions every other cost center answers annually.

What is the ratio of staff in this function to employees, how has it moved over five years, and what changed in the business to justify the movement? What is the median time from approved requisition to accepted offer, and what does each additional week cost in candidates who took another job? How many hours per employee per year are consumed by mandatory processes originating in this department, what is that at fully loaded cost, and what evidence exists that any of it changed behavior? Which programs launched in the past three years have been discontinued, and who decided? When a manager wants to remove someone who is clearly not working out, how many days does it take and how many signatures? What is regretted attrition among the top decile of performers, and did any of them cite process, compression or the impossibility of rewarding excellence on the way out?

The last question is the one that exposes the true cost. Every mechanism described above is a compression mechanism. Bands compress pay, calibration compresses ratings, uniform process compresses managerial discretion, and the aggregate effect is a firm that is systematically better at retaining its median employee and systematically worse at retaining its best one. That trade is never stated, never voted on, and never measured. It is simply the output of a system optimizing for the absence of complaints.

Nobody in this argument is asking for a firm with no people function. They are asking for one with a small, senior, expensive people function that does hard things, and no layer above it inventing work.

Concretely: pay the compliance and payroll professionals well and staff them properly, because those are the parts with genuine downside. Employ two or three genuinely excellent investigators rather than twenty coordinators. Return hiring and firing authority to managers, along with the accountability that makes it real, and accept that some will use it badly and be replaced for it. Delete every mandatory training that cannot show evidence of behavior change, which will be most of them. Publish the fully loaded cost of every process the department imposes. And require the function's leader to defend headcount in the same forum, on the same terms and with the same skepticism applied to a request for a new sales region.

When beavers dam a culvert and flood a road, the amateur reaches for a trap and discovers that the colony is replaced within a season, because the habitat is still good and the sound is still there. The professional installs a flow device: a pipe run through the dam, or a fence that moves the intake far enough upstream that the animal cannot hear the water leaving. The beaver stays, the pond stays, the road stays dry. Nothing is killed and nothing is lectured. The stimulus is simply decoupled from the response.

That is the entire remedy for this function, and it is worth stating in the same terms. Firing the coordinator changes nothing, because the position refills and the sound is still playing. What works is turning off the speakers: stop commissioning surveys whose findings nobody intends to act on, stop treating a competitor's press release as a signal about your own firm, stop escalating every awkward exchange into a policy, and stop rewarding the department for the volume of its response to noise. Leave the real breaches, which are few and which deserve the full competence of a serious professional, and remove the phantom ones.

The department did not seize this territory. It was handed the territory by executives who found it convenient to have somewhere to put the decisions they did not want their names on. The problems it invents are real problems now, because a sufficiently large organization will always make its fictions true. But they were invented, they were paid for, and the invoice was signed at the top.

Anyone serious about the cost should start there, in the room where the demand originates, rather than with the coordinator sending the calendar invitation for the workshop nobody wanted.