Agency

Opening signal

There is a question that sounds sensible and closes the conversation: can it be done?

It gets asked over a sheet of paper listing the reasons why it probably cannot. The mortgage. The children. The market. Age. A contract until the end of the year. Health. Three people who would have to be convinced. The list is honest and most often true — and that matters, because I do not intend to claim here that wanting is enough. Constraints can be hard. People really do have obligations that no amount of resolve will lift.

The problem is not the list. The problem is the shape of the question. A binary question gives a binary answer and ends the thinking. “Can it be done” has two answers, both useless: “no” closes the matter, “yes” does not tell you what to do on Monday.

There is a second layer. “Can it be done” is grammatically a question about possibility and psychologically a question about permission. We address it to someone: to the market, to the boss, to the family, to an imagined authority that would tell us we are allowed. That authority never answers.

On either side of this question stand two lies. The first says that nothing depends on you — convenient, because it excuses you from trying. The second says that everything depends on you — expensive, because it turns every failure into shame and effort into the only available currency. The agency spoken of here is neither. It begins with a third question: where is the lever.

Source

The Corpus Hermeticum is a collection of seventeen Greek treatises, written down roughly between the first and third centuries AD in Hellenistic Egypt, translated into Latin by Marsilio Ficino in 1471 and for a century and a half taken to be older than Moses — until Isaac Casaubon showed on linguistic grounds in 1614 that they were composed considerably later. They are not, then, a record of Egypt’s secret knowledge. They are a record of how, in late antiquity, people thought about a human being caught up in an order he had not chosen.

In the first treatise, the Poimandres, the human being is a double creature: mortal as to the body and something else as to essence. The Asclepius, a text preserved in Latin, puts it more strongly still. What matters, though, is not the claim itself but what follows from it in the text: a person regains power at the moment of recognition, not at the moment of effort. He does not acquire agency through work. He remembers what arrangement he is in.

These texts also have their own name for the force that carries us along: heimarmene — fate understood as a weave of causes, an order of necessity. Hermeticism does not claim that there is no fate, and it does not sell control. It claims something more modest: that recognition changes one’s relation to the arrangement.

Hence the tradition’s insistent opposition. The opposite of gnosis is not ignorance but sleep, drunkenness and forgetting. A sleeping person also acts — only to someone else’s script, and without knowing that it is a script.

What is not in these texts should also be said. There is no programme for changing social conditions, no technique of influence, and no ethics of power. Later centuries repeatedly read into them a promise of power over others — that is a later reading, not the source.

Our reading

We take one tool from this, not a metaphysics one has to believe in. In its working version it goes: before you ask what you can do, check what arrangement you are in and whose script you are currently running. Who set your calendar today. Who defined what is urgent. Whose criterion of success is in your head, and when you adopted it. Very often that single recognition releases more power than a month of discipline — not because it is magical, but because it shows that some of the constraints that looked hard are inherited.

Recognition does not, however, remove the arrangement. The mortgage does not disappear because you see why you took it out. This tradition does not promise a way out of your conditions. It promises that you will stop being merely their function. So we add the second half of the definition ourselves:

Agency is not willpower. It is the recognition of the place where a small input produces a disproportionate change of state — and the readiness to bear that input.

A lever in this sense is a technical notion, not a motivational one, and it has four properties, each of them counter-intuitive.

It rarely lies where it hurts most. Pain indicates the symptom, not the fulcrum. Someone drowning in work looks for the lever in the pace of the work, and it lies in what he takes on in the first place.

It is always narrower than the problem. “Improve sales” is not a lever. “Stop replying to enquiries below threshold X” is.

It requires the right to leave. If you cannot get out of something, you cannot really negotiate it either. That is why the first investment in agency is often not courage but a buffer: savings, a second client, a skill somebody else would buy. Never out of desperation is not a slogan about dignity — it is a description of the conditions under which a lever works at all.

It is local. You have no influence over the market, or over whether language models will eat your profession in three years. You have influence over a dozen or so things within a radius of one month. All the useful work happens inside that radius, and most of the despair outside it.

And a warning built into the notion itself: a lever works both ways. The same asymmetry that lets a small input gain a great deal lets a small mistake lose a great deal. Hence the rule of distribution: protect the core, experiment at the edges.

Operational translation

In practice, levers fall into four types. They are worth knowing, because the search goes faster when you know where to look.

• Entry — whom and what you let in. Clients, projects, people, commitments. The cheapest lever in any system, because it acts before the cost, not after it.
• Terms — price, scope, deadlines, rules of collaboration. One sentence in a contract changes a whole year of work.
• Sequence — order, not quantity. The same tasks arranged differently give a different result, because some of them unblock the rest.
• System — a thing built once that works every week. The slowest lever, and usually the largest.

Work

The typical question is “how do I send out more proposals”. The lever usually sits at entry: which enquiries are worth handling at all, and how you will recognise them in the first three minutes of a conversation. Changing the entry threshold can change a year more than doubling the number of proposals — and it costs one sentence written in advance.

Money

Raising your rate by fifteen per cent or so with three existing clients is usually easier and more certain than winning two new projects. It is not more pleasant — it is cheaper in energy. Agency often looks like an uncomfortable conversation instead of three weeks of work.

Relationships

A year of guesswork against one honest conversation. The cost of the conversation is high in the moment and low across the year; the cost of the guesswork is the reverse. That is exactly the profile of a good lever: a concentrated, short cost and a gain spread over time.

Technology

The lever in AI lies in the repeatable, not the spectacular. Do not ask what the model can do. Ask what you do by hand every week that has a clear criterion of correctness — because that is where automation pays, and where you can check whether the output is any good. Beware the opposite trap: a tool that produces more material to review is not a lever, only a way of moving the work one step further on.

False levers

Three of them impersonate the real thing most often, and all share one feature: they give an immediate sense of movement. A tool bought instead of a decision made — a new system, a new app, when the problem was one subject nobody would talk about. Learning as postponement — one more course before the first attempt; you will recognise it by the fact that you cannot say what specifically that knowledge unblocks. Effort as the answer to every question — the most popular, because it is socially rewarded, and the only one that scales strictly linearly. There is one test: after a week, is something in the arrangement different, or are you just more tired.

Embodied check

Agency has an energy price, and that is an operational fact, not a poetic aside. Pushing on a lever almost always means a conversation you have been avoiding, and conversations of that kind cost disproportionately much relative to their length. Twenty minutes can take a day.

So before any such move, check three things. Sleep over the last three nights — decisions made in sleep debt shift systematically towards accepting terms you would not accept rested. Your state before the conversation, not during it — if on the way to the meeting your breath is shallow and your jaw tight, you will set the terms either too softly or punitively; rarely accurately. The hour of the day — levers are chosen in the morning, not after ten at night, when everything looks structural.

There is also the opposite trap, more common among competent people. In exhaustion one chooses levers that punish oneself. A tired person will almost always conclude that the solution is more discipline, earlier mornings and less rest — because that is the only lever that requires a conversation with nobody. That recognition is false in roughly nine cases out of ten.

It is worth testing levers in an order that is easy to dismiss: sleep, food, movement, people — and only then tools and strategy. One more hour of sleep beats any productivity system you could implement this week. The reverse order is popular because tools can be bought and sleep cannot.

Practice

The agency protocol — a seven-day experiment

Five fields on one sheet of paper, spread over a week. Do not skip the order — most mistakes come from people starting at the lever.

Day 1 — the goal

One sentence describing the end state and a date. Not a direction, not a value, not “to develop”. What has to be true on 30 November. If you cannot write that sentence, you do not yet have a goal, only a mood, and no lever will help.

Day 2 — the obstacle

Not the loudest — the truest. Write out three, and of each ask: if this one disappeared tomorrow, would the goal be achieved? Usually two drop out. The one that remains is the real opponent. It tends to be unpleasant, because more often it concerns what you agree to than what happens to you.

Day 3 — the way round

Three possible solutions, one of which must be unfairly easy: one that seems too simple, or slightly embarrassing. Just ask. Buy instead of building. Drop half the scope. Ask someone who already has it. That third option is in the protocol because it most often wins, and we reject it by reflex, because it does not look earned.

Day 4 — the lever

Choose one place and justify it by a ratio: what it costs in hours, money and discomfort, and how much it changes. Check the type — entry, terms, sequence, system. If your lever is “work more”, you have not found a lever; go back to day two.

Days 5–7 — the experiment

One action, carried out within those three days, with the criterion of the result written down in advance: what I will count as a signal that this works, and what as a signal that it does not. A criterion written after the fact always confirms what you wanted to believe.

A worked example

Goal: by 30 November I have two clients on retainer and no project below the threshold. Obstacle: not a shortage of enquiries, but the fact that I take everything on, because I am afraid of an empty December. Ways round: build a funnel; raise the threshold and wait it out; ask three existing clients outright whether they want an ongoing arrangement — that is the unfairly easy option. Lever: terms and entry at once, at a cost of three conversations and the risk of one refusal. Experiment: three conversations by Sunday; a positive signal is one “yes” or two “tell me more”; a negative signal is three polite refusals — in which case I go back to day two, because the obstacle was somewhere else.

What this experiment does not promise. It does not remove constraints and it does not turn hard situations into easy ones. It does not work well in exhaustion. It does not replace the decision: it will show you the fulcrum, but you have to push yourself.

Limits

The greatest risk in this chapter is not passivity. It is the opposite extreme: the conviction that since I know how to look for levers, I ought to rearrange every situation. Which is why three questions matter here more than all the rest.

When not to act. When you do not yet have the information and acting closes options. When the arrangement is changing on its own and in three weeks the problem will be a different one. When the only available moves are costly for other people and the gain is yours. And when you want to act mainly in order to stop feeling uncertain — that is the most common reason for bad decisions among competent people, because action anaesthetises effectively before it does harm.

Not everything requires action. Not everything that can be changed should be changed.

Which problems are not mine. This distinction saves more energy than any system. Someone else’s problem becomes yours the moment you take it on — and we take them on by reflex, because we know how to solve things and because being needed is pleasant. The test is simple and uncomfortable: did someone ask me for this, or did I volunteer so as not to watch someone cope worse than I would. Taking on other people’s problems can also be a form of control served as help.

When to accept the absence of control. Sometimes there is no lever, and that too is a result. Illness, mourning, an institution’s decision, the market, someone else’s free will. Agency then consists of something else: shortening the horizon, protecting the core, and building the buffer that will give you the right to leave in six months. This is not a failure of analysis. It is an answer: not here, not yet.

And the last limit, because it concerns the notion itself: agency is not a moral virtue. People in harder conditions do not have less of it because they ask worse questions. Treating levers as a measure of a person’s worth is the same mistake as treating wealth as a measure of diligence.

Ethical question

This is the most important question in the whole book, because it concerns the one place where the method described here can genuinely do harm — and not to you.

Agency increases responsibility. The more you can do, the more carefully you must choose.

The ability to find levers is neutral. The same habit of thought that lets you rearrange your own working conditions lets you rearrange the conditions of someone who cannot see that they are being rearranged. That is the real cost of this competence, and good intentions will not get you round it. Only distinctions will — and the distinctions have to be named in advance, because in the situation itself it is already too late to invent them.

Agency versus power over others. Agency is the widening of your own field of choice. Power is the narrowing of someone else’s. Very often the two look identical from the outside and differ in one thing: whether, after your move, the other side has more options or fewer. A move whose effect is that someone must requires a different justification from a move after which someone may.

Influence versus manipulation. The line is sharper than is usually admitted, and it runs through openness. Influence works when the other side sees what you are doing and still agrees. Manipulation stops working the moment it is disclosed — and that is the whole test. If your tactic requires the other side not to know that it is a tactic, it is not influence but a way round their judgement. It makes no difference whether the outcome is good for them. Good intentions do not change the status of a bypass.

Strategy versus treating people instrumentally. Thinking strategically about relationships is fine — we plan conversations, choose the moment, prepare arguments. Instrumentalisation begins where a person becomes solely a position in your arrangement: a contact, a channel, a resource, a point of entry. The practical test: would this relationship survive if it stopped being useful. If you do not have an answer to that, you have your answer.

Asymmetry of information. Agency grows fastest where you see the arrangement better than the other side does. That is precisely the moment of greatest temptation and greatest responsibility. The greater your advantage in recognition, the more has to be said, not less — because an unspoken advantage always turns into a condition the other side never had a chance to negotiate.

And a check question worth asking once a quarter: would the person on the other side of my last three levers have agreed to them if they had seen everything I saw. If yes — that was agency. If no — that was an advantage, and it should be called by its name.

Closing formula

It is worth noticing how little this chapter promises. It does not say that everything is possible. It does not even say that your situation can be improved this quarter. It says something more modest: that the question you ask determines the size of the answer you can get — and that one of those questions systematically returns zero.

This is exactly the hermetic operation, only on the material you have to hand. Recognise the arrangement. See where in it you are and whose script you are running. Find the one place where a small push changes all the rest. And then — because without this step everything above is only a view — push, knowing what it costs and whom.

Someone who asks “can it be done” is waiting for a permission that nobody has the right to grant. Someone who asks “where is the lever” starts from the same material and comes away with a map — and with a bill they have to be able to read.

Do not ask for permission. Ask for the fulcrum.