How To Use Bureaucracies

[Cross­posted from Note­wor­thy]

When we en­counter un­sa­vory fea­tures of re­al­ity, it can be tempt­ing to look away. In­stead, we should ask, “What pur­pose does this serve?”

With this in mind, let’s look at bu­reau­cra­cies. Some peo­ple fear bu­reau­cra­cies; they fear “the Ma­chine.” Others are both­ered by the bu­reau­cra­cies ’ ap­par­ent dys­func­tion. With a bet­ter un­der­stand­ing of bu­reau­cra­cies — what they are, why they’re here, and how they work — both of these re­sponses evap­o­rate, be­cause the re­al­ity is this: bu­reau­cra­cies aren’t al­to­gether bad. In fact, bu­reau­cra­cies can be in­cred­ibly use­ful.

What is a bu­reau­cracy?

A bu­reau­cracy is an au­to­mated sys­tem of peo­ple cre­ated to ac­com­plish a goal. It’s a mech suit com­posed of peo­ple. The owner of a bu­reau­cracy, if an owner ex­ists, is the per­son who can effec­tively shape the bu­reau­cracy. Bureau­crats are peo­ple who are part of a bu­reau­cracy (ex­clud­ing the owner).

Not all or­ga­ni­za­tions are bu­reau­cra­cies. Most or­ga­ni­za­tions are mixed — they have both bu­reau­cratic and non-bu­reau­cratic el­e­ments.

The Pur­pose of Bureaucracies

The pur­pose of a bu­reau­cracy is to save the time of a com­pe­tent per­son. Put an­other way: to save time, some com­pe­tent peo­ple will cre­ate a sys­tem that is meant to do ex­actly what they want — noth­ing more and noth­ing less. In par­tic­u­lar, it’s nec­es­sary to cre­ate a bu­reau­cracy when you are both (a) try­ing to do some­thing that you do not have the ca­pac­ity to do on your own, and (b) un­able to find a com­pe­tent, al­igned per­son to han­dle the pro­ject for you. Bureau­cra­cies ame­lio­rate the prob­lem of tal­ent and al­ign­ment scarcity.

Fea­tures of Bureaucracies

Bureau­crats are ex­pected to act ac­cord­ing to a script, or a set of pro­ce­dures — and that’s it.

Own­ers don’t trust that bu­reau­crats will be com­pe­tent or al­igned enough to act in line with the owner’s wishes of their own ac­cord. Given this lack of trust, own­ers should be try­ing to dis­em­power bu­reau­crats. Bureau­cra­cies are built to al­ign peo­ple and make them suffi­ciently com­pe­tent by chain­ing them with rules. When bu­reau­cra­cies de­liber­ately re­strict in­no­va­tion, they are do­ing it for good rea­son.

Bureau­crats are meant to have only bor­rowed power (power that can eas­ily be taken away) given to them by the owner or op­er­a­tor of the bu­reau­cracy.

Effec­tive Bureaucracies

What is an effec­tive, owned bu­reau­cracy? Why are effec­tive bu­reau­cra­cies owned? To be­gin, we must make two im­por­tant dis­tinc­tions: one be­tween owned and aban­doned bu­reau­cra­cies, and one be­tween effec­tive and in­effec­tive bu­reau­cra­cies.

Owned bu­reau­cra­cies are bu­reau­cra­cies with an owner; they’re bu­reau­cra­cies that some­one can shape. Aban­doned bu­reau­cra­cies are bu­reau­cra­cies with­out an owner.

If a bu­reau­cracy is owned, the bu­reau­cracy’s cre­ator is likely the owner. The cre­ator will have knowl­edge about the setup of the bu­reau­cracy that is nec­es­sary for prop­erly re­form­ing it. Others, un­less given this in­for­ma­tion, will not un­der­stand the bu­reau­cracy well enough to prop­erly re­form it.

The per­son tech­ni­cally in charge of the bu­reau­cracy (e.g. the C.E.O. of a com­pany who is not its founder) might not be its owner sim­ply be­cause he or she doesn’t have suffi­cient in­for­ma­tion about the bu­reau­cracy’s setup to guide it. As a re­sult, the offi­cial head of a given bu­reau­cracy may just be an­other bu­reau­crat.

While the owner is typ­i­cally the cre­ator, this needn’t be true, as long as the new owner has come to un­der­stand enough of the func­tion of the bu­reau­cracy to make effec­tive adap­ta­tions to its pro­ce­dures.

Effec­tive bu­reau­cra­cies are bu­reau­cra­cies that are han­dling the pro­ject they were cre­ated to han­dle. Ineffec­tive bu­reau­cra­cies are bu­reau­cra­cies that are not han­dling the pro­ject they were cre­ated to han­dle.

Bureau­cra­cies that are prop­erly set up will be effec­tive at the start. Changes in re­al­ity re­quire changes in pro­ce­dures, how­ever, so a bu­reau­cracy’s pro­ce­dures in­evitably need to be al­tered ap­pro­pri­ately for it to re­main effec­tive. Over time, aban­doned bu­reau­cra­cies, hav­ing no per­son who can func­tion­ally shape the bu­reau­cracy to make these changes, quickly be­come in­effec­tive bu­reau­cra­cies.

Owned bu­reau­cra­cies, on the other hand, have a shot at mak­ing these adap­ta­tions to pre­vent de­cay. If the owner is skil­led, the bu­reau­cracy’s pro­ce­dures can be mod­ified, and the bu­reau­cracy will con­tinue serv­ing its origi­nal pur­pose. If the owner is un­skil­led, it is as if the bu­reau­cracy is aban­doned — the owner’s efforts to change the bu­reau­cracy’s strate­gies won’t yield suc­cess­ful adap­ta­tion, and the bu­reau­cracy will be­come in­effec­tive. As a re­sult, for a bu­reau­cracy to re­main effec­tive over time, it must be an owned, not aban­doned, bu­reau­cracy with a suffi­ciently ca­pa­ble owner.

Los­ing and Dis­man­tling Bureaucracies

Bureau­cra­cies are best thought of as an ex­ten­sion of their cre­ator and as a source of power for him or her. How­ever, the owner can lose con­trol of the bu­reau­cracy over time, as bu­reau­crats con­vert bor­rowed power into owned power by ex­ploit­ing in­for­ma­tion asym­me­tries. While own­ers will try to limit the owned power of their bu­reau­crats, the bu­reau­crats will have more than enough time to study the in­stru­ments of their con­trol and will learn what is re­warded and what isn’t.

Imag­ine a bu­reau­crat that is sup­posed to be an as­sis­tant to the ab­sen­tee owner of an in­sti­tu­tion. This se­nior as­sis­tant is sup­posed to re­search solu­tions to key prob­lems, and then pre­sent sev­eral op­tions to the owner, who then se­lects one. The as­sis­tant is then re­quired to im­ple­ment the one that was cho­sen. There is a very de­tailed doc­u­ment de­scribing their job and re­quire­ments at ev­ery step of this pro­cess.

The key prob­lem is that a very com­plex set of rules can be eas­ily bent to ac­quire an ar­bi­trary out­come. The out­come will be com­pletely valid from the rule set. This is analo­gous to how in sci­ence a very com­plex model, that fits the data, is not very im­pres­sive. As Von Neu­mann put it: “With four pa­ram­e­ters I can fit an elephant, and with five I can make him wig­gle his trunk.” Let’s walk through the de­scribed pro­cess the se­nior as­sis­tant is sup­posed to fol­low to demon­strate how bu­reau­crats wig­gle their trunks.

You might re­quire the as­sis­tant to not en­gage in origi­nal re­search, but rather work as a search en­g­ine through more ob­jec­tive aca­demic liter­a­ture or best prac­tices in a par­tic­u­lar in­dus­try. The as­sis­tant, how­ever, can cherry pick seem­ingly ob­jec­tive aca­demic pa­pers to ar­gue for their preferred policy out­come. It is ac­tu­ally much eas­ier to start with a pre­con­ceived opinion and then find work con­firm­ing it, rather than re­view a liter­a­ture as a whole. The plau­si­bil­ity of this short­cut should be in­ti­mately fa­mil­iar to any uni­ver­sity stu­dent who worked un­der the pres­sure a dead­line for a class pa­per they didn’t much care about.

The chief as­sis­tant can craft sev­eral op­tions. They can make op­tion B, their fa­vorite, the most ap­peal­ing, and crip­ple op­tions A and op­tions C. Maybe even in­clude point 14, their core agenda, into all three pro­pos­als that vary on points 1 to 13 they don’t much care about. What­ever the im­ple­men­ta­tion of the se­lected solu­tion is, the let­ter of the law can be bent and eas­ily di­verge from the spirit of the law.

In such a cir­cum­stance, an owner can lose con­trol of the bu­reau­cracy and the power that comes with it.

It is of­ten benefi­cial for own­ers to dis­man­tle bu­reau­cra­cies af­ter they have served their pur­pose to avoid los­ing own­er­ship of them due to these in­for­ma­tion asym­me­tries. Bureau­cra­cies of this type might grow to be in­de­pen­dent pow­ers that in­terfere with your plans.

Aban­doned bu­reau­cra­cies might also be vi­able tar­gets for out­side takeover. Such takeovers can be a se­ri­ous prob­lem if un­der­taken by your op­po­si­tion. Bureau­cra­cies nearly always carry a heavy legacy doc­u­ment foot­print; when ex­am­ined this foot­print can not only pro­duce, but also be used to carry out le­gal at­tacks. If the in­sti­tu­tion is vested with an au­thor­ity or rep­u­ta­tion, this can also be turned against you.

If it is too hard to re­gain own­er­ship, dis­man­tling the in­sti­tu­tion for re­sources may be the best op­tion. Th­ese re­sources might be quite eas­ily quan­tifi­able, such as use of real es­tate or key em­ploy­ees. They might also be less tan­gible, such as the at­ten­tion of your al­lies. Un­less you for­mally re­tire a ve­hi­cle, these al­lies might mis­tak­enly be­lieve it ac­tive, caus­ing com­mu­ni­ca­tion is­sues or mi­s­un­der­stand­ings of your key pri­ori­ties.

In short, when han­dling mul­ti­ple or­ga­ni­za­tions ty­ing up loose ends be­comes very im­por­tant.

How to ac­com­plish tasks in an in­sti­tu­tional landscape

Build­ing a bu­reau­cracy is an effec­tive way to ac­com­plish your goals un­der the right cir­cum­stances, but it’s not the best op­tion. In or­der of effec­tive­ness, here are gen­eral op­tions for get­ting things done:


If you can find a com­pe­tent, al­igned per­son who will do the pro­ject in ques­tion for you — let’s call them a del­e­gate — then let them. This per­son can cre­ate a bu­reau­cracy for you, if nec­es­sary, as pro­jects of a cer­tain scale will re­quire bu­reau­cra­ti­za­tion. Un­for­tu­nately, be­cause of the harsh tal­ent and al­ign­ment scarcity men­tioned ear­lier, find­ing del­e­gates can be challeng­ing. Fur­ther­more, cor­rectly as­sess­ing whether some­one is a wor­thy del­e­gate takes skill. Fre­quently peo­ple will ac­ci­den­tally del­e­gate a pro­ject to some­one who is in­suffi­ciently com­pe­tent or al­igned. Failed del­e­ga­tion is worse than build­ing your own bu­reau­cracy, be­cause it will lead to pro­ject failure.

If you have ac­cess to a del­e­gate, don’t treat them like a bu­reau­crat. This wastes a valuable re­source: a del­e­gate can perform tasks you didn’t know needed do­ing and build al­igned sys­tems be­yond your de­sign, a bu­reau­crat can­not.

Fur­ther such treat­ment in­vites dis­al­ign­ment with your del­e­gate. It isn’t just a mat­ter of in­ter­per­sonal grace and re­spect, so it can­not be over­come with kindly man­age­ment; rather if you are at­tempt­ing to closely pro­ce­du­ral­ize the ac­tions of a com­pe­tent del­e­gate, they might ac­cu­rately con­clude the best way to perform their job is to at­tempt to by­pass your con­trol. If you picked them well, they will be rather effec­tive in do­ing so. They don’t need a script — if they’re com­pe­tent enough for your pur­poses, they’ll be able to figure out how to do the pro­ject.

Give them owned power, oth­er­wise you might run them off.


If you can’t find a del­e­gate, then build­ing your own bu­reau­cracy (even if it’s small) is the best bet. Bureau­cra­tiz­ing some things and not oth­ers, on the ba­sis of whether the task can be pro­ce­du­ral­ized, is typ­i­cally more effec­tive than bu­reau­cra­tiz­ing ev­ery­thing by de­fault. Figure out when us­ing an au­to­mated sys­tem is the best op­tion.

Do it yourself

While do­ing it your­self may be most likely to re­sult in a well run pro­ject, it is not always fea­si­ble — you have limited time and ca­pac­ity. Without del­e­gates or bu­reau­cra­cies, the am­bi­tious­ness of the pro­jects you can suc­cess­fully ex­e­cute will be bounded.

Don’t do it

Some things, though use­ful, aren’t worth do­ing…

Un­der­stand­ing the World Around You

Assess­ing People

An un­der­stand­ing of bu­reau­cra­cies lets you an­a­lyze a given per­son’s power: is some­one act­ing as a del­e­gate or a bu­reau­crat? Is some­one cre­at­ing del­e­gates or bu­reau­crats? If some­one has cre­ated a bu­reau­cracy, do they un­der­stand the func­tion of bu­reau­cra­cies? Do they own their bu­reau­cracy, or is it aban­doned? If they own their bu­reau­cracy, is it effec­tive or in­effec­tive? Are they cre­at­ing bu­reau­cra­cies un­der the right con­di­tions? What is the role of bu­reau­cra­cies in their plan?

If a per­son is pow­er­ful, what does it mean if he’s cre­ated many bu­reau­cra­cies? In some cases, the cre­ation of many bu­reau­cra­cies in­di­cates the owner is ex­tremely good at build­ing au­to­mated sys­tems. Alter­nately, he might have trou­ble del­e­gat­ing — per­haps be­cause he can’t find com­pe­tent, al­igned peo­ple, or be­cause he can’t as­sess peo­ple well. Peo­ple who can work well with oth­ers and have ac­cess to suffi­ciently tal­ented al­igned peo­ple need fewer bu­reau­cra­cies. In­stead, they’ll del­e­gate to oth­ers, who can ei­ther do the pro­ject them­selves or cre­ate a bu­reau­cracy of their own.

On the other hand, if a per­son is pow­er­ful, what does it mean if he’s cre­ated few or no bu­reau­cra­cies? If he isn’t del­e­gat­ing, it means that he’s do­ing ev­ery­thing him­self and pos­si­bly doesn’t know how to de­sign au­to­mated sys­tems. If he is del­e­gat­ing, he’s likely to be good enough at find­ing com­pe­tent, al­igned peo­ple that he doesn’t need a bu­reau­cracy. Pow­er­ful peo­ple who don’t cre­ate bu­reau­cra­cies can be just as pow­er­ful as peo­ple who do.

Assess­ing Organizations

The frame­work can be ap­plied to eval­u­at­ing or­ga­ni­za­tions. For a given or­ga­ni­za­tion, be­gin by ask­ing if it’s a bu­reau­cracy. If it is, ex­pect it to be­have in highly stereo­typed ways, it will not be very adap­tive to new challenges and will not ac­cu­rately eval­u­ate things out­side the as­sumed on­tol­ogy of its pa­per­work and in­ter­nal di­vi­sion of la­bor.

If it’s a bu­reau­cracy, we can ask: is it an owned or aban­doned bu­reau­cracy? If it is owned, ex­pect that a large enough challenge will even­tu­ally cause it to re­or­ga­nize. You’ll also be able to reach out to the owner to re­solve prob­lems or find a way to co­op­er­ate that the bu­reau­cracy it­self doesn’t un­der­stand.

Is it an effec­tive or in­effec­tive bu­reau­cracy? If it is effec­tive, you can rely on the in­ter­face it offers you to achieve the goal it claims to achieve. Ineffec­tive ones will provide a some­times be­wil­der­ing ser­vice that might only tan­gen­tially be re­lated to their efforts.

Re­mem­ber that not all or­ga­ni­za­tions are bu­reau­cra­cies.

Some non-bu­reau­cratic in­sti­tu­tions will have to pre­tend they are bu­reau­cra­cies on pa­per for le­gal com­pli­ance. This is an ex­am­ple of a more gen­eral prin­ci­ple: in­de­pen­dent or­ga­ni­za­tions in­ter­pret ex­ter­nally im­posed reg­u­la­tion as dam­age, and route around it.

Or­ga­ni­za­tions can be tightly co­or­di­nated groups the fea­ture a lot of del­e­ga­tion and defer­ence. In these, ex­pect adap­tive be­hav­ior; the on­tol­ogy they are work­ing in might rapidly change to re­spond to ei­ther your challenge or offer of co­op­er­a­tion. Most im­por­tantly there will be in­di­vi­d­u­als be­yond merely the leader who can ex­er­cise their own judge­ment.

Effec­tively In­ter­act­ing with Ex­ist­ing Organizations

If an or­ga­ni­za­tion is not a bu­reau­cracy but rather a tightly co­or­di­nated group, talk to the del­e­gates if you want to get things done; they will have free­dom to act com­pe­tently within their own do­main and will be eas­ier to reach than lead­er­ship.

The key ad­van­tage of talk­ing to peo­ple over en­gag­ing with au­to­mated sys­tems is that you can bring con­sid­er­a­tions from out­side their im­me­di­ate in­sti­tu­tional con­text into con­sid­er­a­tion. While the lo­cal bal­ance of power might still be in the way of such con­sid­er­a­tions, it is sur­pris­ingly of­ten vi­able to have them taken into ac­count.

If it’s a bu­reau­cracy, you can ei­ther (1) go along with it, (2) figure out how to by­pass it, or (3) co­or­di­nate with its owner, if it is owned. You may pre­fer to by­pass (or game) the bu­reau­cracy if it is aban­doned and thus dys­func­tional, or if you aren’t al­igned with its owner.


The ori­gin of bu­reau­cra­cies lies in them ex­tend­ing power and effects far be­yond what a sin­gle in­di­vi­d­ual can do. They can do so in the ab­sence of ex­pen­sive and difficult co­or­di­na­tion, or difficult to train and eval­u­ate in­di­vi­d­ual tal­ent.

Much like fac­to­ries can pro­duce cheap prod­ucts at scale with un­skil­led la­bor, dis­plac­ing crafts­men, so have bu­reau­cra­cies dis­placed lo­cal so­cial fabric as the gen­er­a­tors of so­cial out­comes.

We find our­selves em­bed­ded in a bu­reau­cra­tized land­scape. What can or can­not be done in it, is de­ter­mined by the or­ga­ni­za­tions com­pos­ing it. The con­stant drive by tal­ented in­di­vi­d­u­als to both ex­tend power and make due with un­skil­led white col­lar la­bor (a cat­e­gory that economists should rec­og­nize and talk more about) have lit­tered the land­scape with many large or­ga­ni­za­tions. Some re­main pi­loted, oth­ers are long aban­doned. Some con­tinue to perform vi­tal so­cial func­tions, oth­ers lum­ber about mak­ing life difficult.

Much as we might be­moan the very real hu­man cost bu­reau­cra­cies im­pose, they cur­rently provide ser­vices at economies that are oth­er­wise sim­ply not pos­si­ble. We must ac­knowl­edge our col­lec­tive and in­di­vi­d­ual de­pen­dence on them and plan to in­ter­act ac­cord­ingly.