The New Shape of Leadership

Consider a founder who built an agency from a spare room to two hundred people. When it was six of them around a kitchen table, she could lead the way the best small-firm leaders always have: by being in the room. Every price, every hire, every awkward client conversation passed across her desk or within her earshot. She did not manage the work so much as inhabit it. If a decision was going wrong, she felt it before it had a name.

Now there are two hundred. She sees perhaps a tenth of what matters. The rest happens in client meetings she is not in, in message threads she will never read, in the split-second judgement of an account manager deciding whether to absorb a scope change or push back on it. She is a clever woman and she noticed the drift years ago. Her response was the natural one: reporting. Weekly status. A pipeline review. A utilisation dashboard. A project tracker with a column for risk, rated red, amber, or green.

Here is the strange thing she cannot quite explain to her board. The more she reports, the less she feels she knows. The decks are immaculate. The dashboards are green, mostly. And yet the surprises keep coming from inside the building: the account that churned with no warning, the project that was green until the week it went red, the margin that evaporated somewhere between the quote and the invoice. She has more information than any leader in the firm's history and less confidence than she had at the kitchen table.

She has not failed at leadership. She has run into the limits of a model of leadership that was built for a different kind of work.

The question underneath the dashboards

The instinct, when control slips, is to reach for more of what used to give you control. More reporting, more oversight, more status, more sign-off. It is just aimed at the wrong target, because it treats a leadership problem as a visibility problem and then tries to solve visibility by watching activity.

The real question this chapter answers is harder and more useful than "how do I get more visibility?" What does leadership actually become when you can no longer make, or even witness, most of the decisions that determine whether your firm wins? And what does it become when some of those decisions are increasingly made, shaped, or accelerated by machines that work faster than any human can supervise?

The answer is that the shape of leadership is changing. The old shape, control through hierarchy and visibility through reporting, was built for work you could see. The new shape is built for work you mostly cannot, and it rests on three things the old model never needed: judgement, transparency, and the ability to make a good decision legible to someone who was not in the room.

The gantry is gone

The industrial model of management was a genuine achievement. On a factory floor the work is visible, sequential, and largely standardised. A manager standing on a gantry could see the whole line: where it was moving, where it had stopped, who was behind. Control was a matter of specifying the task precisely, watching it performed, and correcting deviation. The person who designed the process knew the best way to do the job, and the person on the line executed it. Supervision worked because a weld either happened or it did not, and you could see which.

The industrial model was engineered deliberately, early in the last century, to wring waste and variation out of manual work, and it succeeded so completely that we forgot it had ever been a choice. Its logic seeped into everything: the org chart, the reporting line, the monthly review, the assumption that a manager plans the work and then checks that the work matches the plan. All of it was built on the belief that the work can be seen. When that belief held, the furniture was excellent. We kept it long after the work stopped being visible.

Almost none of that holds for knowledge work, and a people business is knowledge work all the way down. The value your firm sells is not a weld. It is a judgement: how to frame the client's problem, which two people to put on it, whether to challenge the brief or accept it, when to escalate and when to hold. That judgement is invisible. It happens inside a person's head in the seconds before they reply to an email. It is non-standard, because the whole point of hiring judgement is that each situation is a little different. And very often the person doing the work knows more about the specific call than the leader above them does.

Command-and-control assumes the boss knows the best move and the worker executes it. In knowledge work that assumption is frequently false, and building a management system on a false assumption produces exactly the paradox our founder is living: more oversight, less control. She is gathering more evidence of activity and mistaking it for a view of the work.

Now add machines to the floor. Increasingly, the decisions that move a people business are not purely human. A model drafts the proposal, ranks the candidates, suggests the price, flags the at-risk account. Sometimes a person reviews the suggestion with care. Often, under deadline, they wave it through. The result is that judgement is now distributed not only across many people but across people and software together, and it moves faster than any gantry-style supervision could ever track. You cannot watch a thousand small decisions a week. You certainly cannot watch them when a fair number of them are being made at machine speed. The supervisory model does not merely strain at this scale. It becomes a fiction that everyone maintains and nobody believes.

And the machines change the stakes. When a junior made a questionable call, the damage was bounded by how much work one person could do in a day. When a model makes the same call, it makes it a thousand times before lunch, in the same confident register, and it never gets tired, or suspicious, or struck by the feeling that something is off. Speed without judgement is not a gift to a leader who already cannot see the decisions. It is an accelerant. The firms that will handle machine judgement well are not the ones with the best models but the ones that had already learned to see and shape the decisions their people were making, because a machine dropped into a firm that cannot see its own decisions simply makes more of them, faster, in the dark.

From managing activity to managing decisions

If you cannot supervise the work directly, what do you manage instead? The honest answer, in most firms, is that you manage activity. You count the things that are easy to count: hours logged, utilisation, tasks closed, tickets resolved, milestones hit, the colour of the status field. Activity is attractive because it is visible and countable, which makes it feel like control. But activity is a proxy, and it is a proxy that has come loose from the thing it was meant to stand for.

Here is the mechanism. A team can be fully utilised, admirably busy, every timesheet full and every status green, and still be quietly making the decisions that will lose the engagement money. The status deck tells you delivery is amber. It does not tell you why, and the why is the only part a leader can actually act on. Activity metrics describe motion. Leadership is about direction, and you cannot steer by watching the speedometer.

Activity metrics describe motion. Leadership is about direction, and you cannot steer by watching the speedometer.

Picture the case that belongs framed on every operating wall. A delivery team runs a flagship account at ninety-eight per cent utilisation for a year. Every agent is busy, every timesheet full, every weekly status a calm green, and the team is quietly held up as the one the others should copy. What no dashboard shows is the string of small, sensible-looking decisions made under all that utilisation pressure: the senior lead moved off to a newer account to keep her billable, the scope creep absorbed rather than repriced, two escalations smoothed over rather than surfaced, the renewal conversation deferred because everyone was too busy delivering to have it. Then the client did not renew, and the lifetime value of a relationship the firm had spent three years building was gone in a single quarter. The firm lost the account at full utilisation, and not one instrument on the wall so much as flickered.

Managing decisions means asking a different question about the same work. Not "is it green or amber?" but "what did we choose, and why, and on what evidence?" Compare two accounts of the same troubled project. The first is a status update: "Delivery is amber, we are behind on milestone three, mitigation in progress." The second is a decision: "To protect margin on a job we priced too tight, we staffed two juniors where we had quoted a senior, betting the client would not feel the difference. The client felt the difference. That bet is why we are amber." The first is weather. It reports the conditions and gives you nothing to hold. The second is leadable. It names a choice, an owner, a piece of reasoning, and a bet that can be revisited. There is nothing to do with the first except wait and hope the mitigation works.

Multiply that difference across a firm and it decides everything. A firm that runs on weather reports meets every problem as a surprise and loses sight of its cause. A firm that captures its decisions spends the same hours reading the bets it is making while they can still be changed, which is the only moment at which leadership is actually possible.

Most firms run almost entirely on the first kind of sentence, having built elaborate machinery to collect weather reports and almost none to capture decisions, because activity is easy to record and decisions are not. We manage what our systems will hold. Hold that thought. It is the hinge the whole book turns on.

Setting the conditions instead of making the calls

Once you accept that you cannot make or watch most of the decisions, the leader's job moves upstream. You stop trying to be the person who makes the hundredth pricing call and become the person who makes sure the hundred pricing calls get made well by other people. That is not a retreat from responsibility. Done properly, it is a more demanding form of it.

What are the conditions for a good decision? Strip it back and there are only a few, but they are exacting. People need clarity about what good looks like, which means knowing what the firm is actually optimising for in this situation, because "protect the margin" and "protect the relationship" often point in opposite directions and someone has to have said which one wins here. They need access to the relevant evidence, the true state of the account and the engagement, not a rumour and a gut feel. They need clear ownership, so that a decision has a name attached and does not fall into the gap between two roles. They need a route to escalate without shame when a call is above their pay grade. And they need it to be safe to record a decision honestly, including its risks and the bet inside it, without that honesty being turned against them the moment the bet goes wrong.

Notice that every one of those conditions is something a leader builds, not something a leader decides. You cannot personally make the right pricing call on an account you have never heard of. Set those conditions and a thousand good decisions happen without you. Fail to set them and you can review every deck in the firm and still be surprised every month, because the surprises are being manufactured upstream of anything a report will ever show you.

There is a compounding quality to this that rewards the patient leader. A decision made in the right conditions does not merely come out better, it comes out legible: the person who made it can say afterwards what they were optimising for and what they stood on, and a legible decision is one the firm can learn from. Set the conditions habitually and you build a firm that gets better at deciding, because each good decision leaves a trail the next one can follow.

This is harder than making the calls yourself, not easier. When you make the decision, you can at least tell yourself you were in control. When you set the conditions, you are accountable for outcomes you did not directly choose, produced by people you will never fully supervise, and you have to resist the constant pull back towards doing it all yourself. That pull feels like leadership. It is actually just the highest-paid person in the building volunteering to become the bottleneck.

The leader as an editor of judgement

There is a useful picture for what this leader does day to day, and it comes from a craft that solved the problem long ago. A good editor does not write the book. The editor is very often not as gifted a writer as the best author on the list. What the editor does is different and irreplaceable. They hold a standard for what good looks like. They ask the question that exposes the weak paragraph. They see across many manuscripts what a single author cannot see from inside one. They spread what is working and cut what is not. The words remain the writer's. The judgement about the whole is the editor's.

That is the shape of the new leadership. You are not the source of the answers. You are the editor of your firm's judgement. You review the decisions that matter, and reviewing does not mean re-deciding. It means asking the questions that make weak reasoning visible: what were we optimising for, what did we give up, what would have changed our minds? You keep alive a standard of what a good decision looks like in this firm, so that "how we make calls here" is a real and shared thing rather than two hundred private methods. And you notice the pattern, the pricing mistake being made in three different teams for the same reason, that no individual decision-maker could ever see because each of them sees only their own work.

The craft of it is mostly in the questions. An editor of judgement resists the urge to supply the answer, because supplying the answer trains people to bring you problems instead of proposals and quietly recentres every decision on you. Instead you ask the questions a good decision should already have answered, and you ask them consistently enough that people begin to ask them of themselves before they reach your door. What were we optimising for. What did we trade away. What would have to be true for this to be the wrong call. The standard has moved out of your head and into the firm's habits, where it can work without you present.

This is the only way judgement scales. The alternative, the leader as sole source of answers, does not scale past the leader's own calendar. An editor does not need to be a better writer than everyone on the list to make the whole list better. A leader does not need to be the best decision-maker in the building to make the building decide well. They need to see the whole and hold the standard, and both of those depend on one thing the old model never provided: being able to see the decisions in the first place.

Transparency of decisions, not surveillance of activity

When control slips there are two directions a leader can go, and they lead to opposite firms. The first is surveillance. If I cannot see the work, I will watch the workers: monitor the hours, track the logins, count the keystrokes, tighten the status reporting until I have a minute-by-minute picture of who is doing what. The second is transparency. If I cannot see the work, I will make the decisions visible: build a firm in which what we chose, and why, is open to the people who need to see it.

These feel similar because both are about visibility, but they could hardly be more different. Surveillance watches activity, the wrong thing to watch, and worse, it corrodes the very thing a people business runs on. It says, in every keystroke it logs, "I do not trust you, so I am watching you." They perform busyness. They optimise for the metric. They stop telling you about the risky decision, because visibility has become a threat rather than a shared act. You end up with a firm that is fully monitored and less honest than it was before you started watching.

Transparency of decisions does the opposite. It assumes good faith and examines choices, where surveillance assumes bad faith and measures movement, and it is the move that actually builds trust, for the reason the first chapter set out: you cannot demonstrate good judgement if your judgement is invisible. A firm whose decisions are legible can show a client, a nervous CFO, or an account manager who has just inherited a relationship she has never seen, not merely what happened but why it was chosen. That "why" is what earns trust. A firm whose decisions live only in people's heads is asking to be trusted on faith, and faith does not survive contact with the first bad surprise.

There is a second reason transparency matters more now than it did a decade ago. Once machines are in the decision loop, drafting, ranking, pricing, and flagging, you need to see not just the decision but the evidence beneath it, because a model will hand you a confident answer whether or not it is grounded in anything true. Surveillance tells you the account manager approved the model's suggested price at four o'clock. Transparency of the decision tells you what the price was based on, what alternatives were weighed, and whether the machine's confidence had any foundation at all. In a world of confident machines, the leader who can see the reasoning holds an advantage over the one who can see only the activity, and it is not a small one.

It is worth being concrete about what legible means here. A legible decision is not a longer status report. It is a short, honest record of a choice: what we decided, who owns it, what we were trying to achieve, what the main alternative was, and what we are betting on that could turn out wrong. That last part is the one firms flinch at, because writing down the bet feels like writing down the blame in advance. It is the opposite. A risk named in the open can be revisited while there is still time to change course; a risk never spoken becomes visible only as a result, by which point it is no longer a decision but an outcome, and outcomes cannot be led, only survived.

"Someone still has to decide"

The honest objection to all of this is that it sounds soft. Setting conditions, editing judgement, choosing transparency over surveillance: it can read like a leader who has talked themselves out of leading, who has dispersed responsibility into culture and process so thoroughly that when something goes wrong there is no one left holding it. Someone, the objection goes, still has to make the call. Someone still has to be accountable. You cannot edit your way out of a decision that lands on your desk at ten at night with the client on the phone.

All of that is correct, and none of it counters the argument. Someone does have to decide, own the call, and carry the consequences. A firm in which "the conditions" decide and no human is accountable is not an enlightened firm, it is a fog, and fog is where bad decisions go to escape their owners. The new shape of leadership does not abolish the accountable individual. It depends on one.

What it changes is whether accountability is real or theatrical. In the old model, decisions live in heads, in meetings nobody minuted, in threads that scroll away, so when something goes wrong, accountability becomes retrospective and adversarial. Everyone reconstructs the history, and in the reconstruction it is remarkable how the decision turns out to have been someone else's, or the inevitable result of circumstances, or a thing that was never really decided at all. You cannot hold a person accountable for a decision no one can find. The old model did not deliver accountability at scale. It delivered the performance of accountability, punctuated by blameless post-mortems whose true cause is always "a breakdown in communication."

You cannot hold a person accountable for a decision no one can find.

Making decisions visible and reviewable is what makes accountability possible across hundreds of people rather than only inside the one room the leader happens to be sitting in. You can own a decision you recorded, review a decision you can see, and learn from a bet whose reasoning was written down before anyone knew how it turned out, because a decision judged solely by its outcome teaches you nothing about the judgement that produced it. Far from being soft, the new model makes the hardest job in the firm, being answerable for choices you did not personally make, actually doable. The old model just hid how undoable that job had quietly become.

What the leader cannot see

Everything in this chapter rests on a single assumption. Setting the conditions, editing your firm's judgement, choosing transparency, holding people accountable for decisions rather than activity: every one of these moves assumes you can actually see the decisions your firm is making, that you can find them, read them, and connect them to what happened next.

You cannot. Not in most firms, and not because the leaders are careless or the people are hiding things. You cannot see the decisions because nothing you bought was built to hold them. Every system in the firm records something, faithfully, and not one of them records the decision itself, the actual "we chose to staff it this way, price it that way, escalate here, walk away there," with its reasoning and its owner and its bet. That lives in a head, a thread, a meeting nobody minuted, and then it evaporates.

Leaders cannot lead what they cannot see. The software was supposed to help them see, and it does not, and understanding exactly why it does not is the whole work of the next part of this book. So that is where we turn: to the machinery itself, and to the strange fact that we have built the most heavily instrumented businesses in history and still cannot answer the one question that matters. Not what happened. What we decided, and why.