The Rise of the AI-Enabled Chief of Staff

A chief executive we know keeps a private metric. Every Friday she counts the number of decisions that were ready to be made on Monday and still had not been made. For most of her tenure the number tracked her calendar: when the week was busy, the backlog grew. Two years ago the relationship broke. The backlog began growing during quiet weeks too, and it kept growing after she cut her meeting load by a third.

What had changed was not her schedule. It was the volume of decision-ready material arriving at her desk. Her organization had gotten dramatically better at producing analysis. Market scans that used to take an analyst two weeks now took an afternoon. Competitive teardowns, customer research syntheses, scenario models, first-draft board memos — all of it had become cheap, fast, and plentiful. The supply of things worth deciding had increased by something like an order of magnitude. Her capacity to decide had not moved at all.

This is the condition that has made the Chief of Staff the fastest-emerging executive role of this cycle. Not because AI created a need for someone to run the AI. Because AI made analysis abundant, and abundance moved the bottleneck. When producing a good options memo was expensive, the scarce resource was analytical horsepower and organizations invested there. Now producing the memo is cheap, the memo arrives, and the scarce resource is the thing that converts a memo into a decision that someone owns, that survives contact with the organization, and that actually changes what people do on Tuesday.

That conversion has always been the Chief of Staff's real job. What changed is that it went from a nice-to-have for a certain kind of executive to the constraint on the whole enterprise.

The role is also, by a wide margin, the most consistently mis-hired senior position we see. Three failures repeat. Organizations write a job description that blends three incompatible archetypes and then wonder why no candidate fits. They grant the responsibilities of the role without the authority, producing a talented person with a title and no ability to move anything. And they recruit from one talent pool when the competencies they actually need are distributed across at least six.

That last failure is the most costly and the easiest to fix, so we will spend real time on it. The default assumption is that a Chief of Staff comes from the executive assistant track — that the role is what an exceptional EA grows into. Sometimes it is, and the best of those are formidable. But it is a partial view of the pool, and organizations that hold it exclusively are searching a narrow slice of a wide market. Business analysts, program and PMO leaders, revenue and business operations managers, strategy associates, military staff officers, and engineers who have carried cross-functional programs are all strong sources for this seat, and most of them have never been approached about it.

Here is where the argument goes.

We start with the lineage, because the role is two centuries old and the traditions that built it — the Prussian staff system, the White House — understood something about it that corporate adopters routinely miss. Then the four archetypes, which is the single most useful distinction in this paper: the gatekeeper, the strategic partner, the program driver, and the proxy. Most failed searches can be diagnosed by holding the job description against that list and counting how many it asks for.

Then what actually changed with AI, stated specifically enough to be useful, followed by the working method — the weekly mechanics of the role as it is now practised, including the verification discipline that has become its most distinctive new duty. Then the decision operating system, which is the seat's real product and the thing that persists after any individual holder leaves.

Then the talent argument, at length, because it is where the most value is being left on the table. Six pools, what each brings, and where each hits a ceiling. Then a competency model with assessment methods, because these competencies are unusually resistant to conventional interviewing and unusually easy to test directly.

Then how the seat fails — five modes, four of which are the employer's fault and none of which are fixed by finding a better candidate. Then the practical parameters: reporting line, compensation banding, tenure, succession. Then an argument that is rarely made and probably should decide the appointment: the role is the most efficient leadership development mechanism most organizations possess, and treating it that way changes who you should hire.

Then a section addressed to candidates, three composites from real engagements, and the serious objections. Two appendices close it: an interview guide with the exercises we actually use, and a structure for the first ninety days.

The through-line, if you want it in one sentence: this is a role whose scarcity was created by abundance elsewhere, and it is being recruited with a model of the job that is roughly a decade out of date.

Analysis became cheap. Deciding did not. The Chief of Staff sits exactly on that gap, which is why the seat suddenly matters.

Part I — The Role Is Old. The Scarcity Is New.

The Chief of Staff is not an invention of the technology industry, and understanding where it comes from clarifies what it is for.

The lineage runs through military staff systems. The Prussian General Staff, formalised in the nineteenth century, institutionalised something armies had learned expensively: a commander's capacity to direct is finite, and the binding constraint in a large operation is rarely the commander's judgment. It is the machinery that turns judgment into coordinated action across units that cannot all be in the room. The staff officer's function was never to command. It was to ensure that what the commander decided was understood, resourced, sequenced, and executed — and that what the commander needed to know arrived in time to be useful. Every serious military since has adopted some version of this, and the vocabulary of the role in business — staff work, the commander's intent, the running estimate — comes directly from it.

The civilian government version arrived formally in 1946, when the White House Chief of Staff position was created, and it inherited the same logic. A president's time is the scarcest resource in the executive branch, the volume of matters demanding attention vastly exceeds it, and someone must decide what reaches the desk, in what form, having been through what process — and then ensure the decision propagates.

Corporate adoption came later and unevenly. It became common in technology companies over the last fifteen years, where hypergrowth produced executives whose span of control expanded faster than their ability to be present in it, and it has since spread across sectors. For most of that period it was understood as a founder-scaling role: something a company acquired at Series C and possibly retired later.

Across all three traditions, the function is constant. Somebody manages the executive's information supply chain in one direction and their decision propagation in the other, and holds the loop between them closed.

What has changed is the ratio between the two directions.

Historically, the input side was the expensive part. Getting a good answer to "should we enter this market" required weeks of work by expensive people, and the Chief of Staff's contribution was substantially about commissioning that work, chasing it, and packaging it. The output side — making sure the decision propagated — mattered, but it competed for attention with the grind of getting the analysis produced at all.

That has inverted. The input side has collapsed in cost. A competent operator with modern tools can produce, in a day, a market scan, a competitor teardown, a synthesis of two hundred customer interviews, and three scenario models with sensitivity analysis. The quality varies and requires exactly the kind of scrutiny described in our companion paper on the ghosts in the machine — but the raw production is no longer the constraint.

So the bottleneck moved downstream, and it landed on three things that resist automation:

Judgment about what deserves the executive's attention. When ten things could plausibly reach the desk, someone must decide which three do. This is not filtering by importance in the abstract; it is filtering by what this executive, given what they are currently trying to accomplish and what they already know, needs to spend attention on this week. It requires holding a model of the executive's priorities that is accurate and current, and it is why the role does not transfer cleanly between principals.

Verification of what the machine produced. The memo is fluent. Is it right? Someone has to know enough about the domain to notice that the market-size figure double-counts a segment, or that the competitor analysis is describing a product that shipped eighteen months ago. Fluent and wrong is a far more dangerous input than obviously incomplete, and it now arrives constantly.

Propagation and follow-through. A decision made in a room does nothing until it reaches the people who will act on it, in a form they can act on, with the resourcing and sequencing worked out and the conflicts with prior commitments resolved. This has always been the largest part of the job, and it is entirely untouched by improvements in analysis. If anything it has gotten harder, because the decision rate has gone up.

The Chief of Staff role became urgent because these three things became the constraint simultaneously, in a period when most organizations were investing heavily in the part that was no longer scarce.

Part II — Four Archetypes, and Why Job Descriptions Blend Them

Before discussing who fits, we have to be precise about what is being hired, because the single most common failure in these searches is a job description that describes several incompatible roles at once.

There are four recognisable archetypes. Each is legitimate. They require different people, different backgrounds, different authority, and different compensation.

The Gatekeeper

Owns the executive's time, attention, and access. Runs the calendar as a strategic instrument rather than a scheduling problem: what gets an hour, what gets fifteen minutes, what gets declined, what gets delegated, and who gets to be in the room. Manages the flow of inbound requests and protects the executive's capacity for the work only they can do.

This archetype has the strongest overlap with the senior executive assistant track, and the best practitioners are often people who have done that work at a high level and understand that access control is a genuine strategic function rather than administration. The competencies are judgment about relative priority, political awareness, extraordinary discretion, and the confidence to say no on someone else's behalf and make it stick.

Where it hits a ceiling is content. A pure gatekeeper can tell you that a decision is stuck; they may not be able to tell you that the analysis underneath it is wrong.

The Strategic Partner

A thinking partner who engages with the substance. Pressure-tests the executive's reasoning, prepares the difficult questions before the board does, holds a view, and is expected to disagree. Frequently drafts the executive's most important communication — board letters, strategy memos, all-hands remarks — not as a ghostwriter transcribing, but as someone who can construct the argument.

This is the archetype most often sourced from consulting, corporate strategy, investment banking, or an operating role in the business. The competency is analytical judgment plus the standing to challenge a chief executive without damaging the relationship, which is a rarer combination than either component alone.

Its failure mode is a person who is excellent in the room and cannot make anything happen outside it.

The Program Driver

Owns execution of the executive's priorities across the organization. Runs the operating cadence — the weekly business review, the quarterly planning process, the goal system. Chases the cross-functional initiative that has stalled between two VPs who both think it is the other's problem. Is measured on whether things that were decided actually got done.

Sourced most naturally from program management, business operations, PMO leadership, and military staff backgrounds. The competency is systems thinking plus relentless follow-through plus the political skill to drive people who do not report to you.

Its failure mode is process for its own sake: an immaculate cadence that produces meetings rather than outcomes.

The Proxy

Represents the executive in rooms the executive cannot attend, with real delegated authority to commit. Makes calls in their name and is backed when they do. This is the deepest version of the role and the least common, because it requires a level of trust that is typically earned over years rather than hired for.

Sourced almost entirely internally, or from someone with a long prior relationship with the principal.

The problem is that most job descriptions ask for all four. They want the discretion and access control of the gatekeeper, the analytical firepower of the strategic partner, the operational grip of the program driver, and the standing of the proxy — at a compensation band supporting perhaps one and a half of those, reporting into a structure that grants the authority of none.

Such a search either fails outright or produces a strong candidate in one dimension who is then evaluated against expectations set by the other three. This is the most common way a good Chief of Staff hire becomes a bad Chief of Staff outcome, and the diagnosis is almost always available in the job description before anyone is interviewed.

The remedy is unglamorous. Rank the four. Decide which one is genuinely primary for the next eighteen months, which is secondary, and which two you are explicitly not buying. Write that down. Then recruit against it.

Part III — What AI Actually Changed About the Job

It is worth being specific here, because "AI-enabled Chief of Staff" is drifting toward meaning "a Chief of Staff who uses ChatGPT," which is both trivially true of everyone and not a role definition.

Four things changed materially.

The preparation ratio inverted

Preparing an executive for a significant meeting used to be dominated by gathering: pulling the numbers, finding the prior commitments, assembling the history, remembering what was said last quarter. That gathering is now largely mechanised, and what remains is the part that was always more valuable and always got squeezed — deciding what the executive most needs to understand, anticipating the two questions that will actually be difficult, and identifying the thing everyone in the room is avoiding.

The practical consequence is that the quality bar for preparation has risen sharply. When gathering was expensive, a well-gathered briefing was good work. Now it is table stakes, and the differentiator is judgment about salience. This has quietly raised the competency requirement for the seat while the market's mental model of the role has not caught up.

A new function appeared: owning the executive's information supply chain

Senior leaders are now receiving synthesised, machine-generated material from every direction — their own tools, their teams' tools, vendors, boards, advisors. Much of it is fluent. Some of it is wrong. Almost none of it arrives labeled with how it was produced or what it depends on.

Somebody has to own the integrity of that flow. Not by blocking it, which is impossible and would be foolish, but by knowing which streams have been verified and which have not, insisting that material reaching the desk carries its provenance, and catching the confident paragraph that is confidently incorrect before it becomes the basis of a commitment.

This is a genuinely new duty and it maps directly onto the human-in-the-loop discipline. For a chief executive's decision inputs, the appropriate posture is almost always human-in-the-loop, and the Chief of Staff is frequently the human. Making that explicit — this is your job, you have the authority to send things back, we will measure whether you are catching things — converts an informal habit into a control.

The synthesis load moved up, not away

A common prediction was that AI would eliminate the Chief of Staff's synthesis work. What happened instead is that first-pass synthesis became cheap while the volume requiring synthesis exploded, so the net effect was to move the work up a level of abstraction. The Chief of Staff is now less often synthesising a document and more often synthesising across twelve documents that were each themselves synthesised, with varying and unrecorded reliability.

This is harder, not easier. It requires holding a model of what each source is good for and where each is likely to mislead.

Decision velocity became a governable variable

When analysis was the constraint, decision velocity was capped by something outside anyone's control. Now it is capped by organizational design, which means it can be managed — how decisions are queued, what quality of input is required before something reaches the desk, what gets decided by whom at what threshold, how decisions are recorded and propagated.

That is a real operating system, and someone has to run it. Increasingly that someone is the Chief of Staff, which pushes the role toward the program-driver archetype even in organizations that hired for the strategic-partner one.

What did not change: the political work, the trust, the discretion, the ability to walk into a room where two executives are dug in and get them unstuck. Nobody has automated any of that, and it remains a large fraction of why the role works when it works.

Part IV — The Working Method

If the role is genuinely AI-enabled rather than merely AI-adjacent, the difference shows up in the weekly mechanics. It is worth describing them concretely, because this is what separates a job description from a job.

The pre-read, inverted

The traditional pre-read was assembled from whatever the contributing teams sent, lightly edited, and circulated late. The modern version inverts the effort. Gathering and first-pass summarisation are compressed into a fraction of the time, and the freed capacity goes into three things that used to get cut.

The first is the disagreement map. Where do the contributing functions actually differ, and on what? Most executive meetings waste their first twenty minutes discovering a conflict that was visible in the source material. Surfacing it in advance — naming who believes what and why — is high-value work that nobody had time for when assembling the deck was the job.

The second is the question set. Not discussion topics; the two or three questions that, if answered, would resolve the decision. Executives are good at answering sharp questions and poor at responding to comprehensive briefings, and the difference between the two is entirely the preparer's work.

The third is the provenance line. Every material claim carries where it came from and how confident anyone should be in it. This takes a paragraph and prevents a category of expensive error.

Verification as a standing discipline

The Chief of Staff who does this well runs a checking routine, and it is specific rather than a general posture of scepticism, because general scepticism does not survive a busy week.

Numbers get traced to a source system, not to a document that cites one. Comparisons get checked for base consistency — the most common serious error in machine-generated analysis is a comparison against a subtly non-comparable denominator, and it is invisible unless someone looks. Claims about the external world get spot-checked for currency, because models describe a world that is some months stale. Anything that will be repeated externally gets checked twice.

The discipline that matters most is the one that is hardest to maintain: sampling rather than reviewing everything. A Chief of Staff who attempts to verify every input will verify none of them properly by Thursday. Deep checks on a rotating sample, plus complete checks on anything irreversible or externally visible, is the design that survives contact with a real calendar. The reasoning here is the same as in human-in-the-loop system design, and it fails the same way when ignored.

The follow-through ledger

The most valuable artifact a Chief of Staff maintains is also the least sophisticated: a live record of what was decided, by whom, when, what was supposed to happen next, and whether it did.

Most organizations do not have this. Decisions are recorded in meeting notes that nobody re-reads, and the question "what did we decide about pricing in March, and what happened" takes two days to answer badly. A maintained ledger answers it in a minute, and its second-order effect is more valuable than its first: when people know decisions are tracked, decisions get made more carefully and followed more reliably.

The tooling barely matters. The discipline does.

What does not get delegated to a model

The good practitioners we know all maintain a short list of work they have stopped routing through AI tools, and the list is instructive because it is remarkably consistent.

Anything requiring knowledge of what a specific person will do — how a board member will react, whether a leader is genuinely bought in or is complying — because the model has no access to the relevant data and will confabulate plausibly. Anything where the political framing is the substance, such as how to present a reversal without anyone losing face. Anything touching a personnel matter, for reasons of both judgment and confidentiality. And the final read of anything going to a board, not because a model cannot draft it, but because the accountability is personal and delegating the last look is how errors survive.

A candidate who can produce this list from experience is describing real use. A candidate who cannot has probably not been doing the job.

Part V — The Decision Operating System

The most durable contribution of a strong Chief of Staff is not any individual brief. It is a functioning system for how decisions get made, and this is the part of the role that has become most valuable as decision volume has risen.

Five components, none of them complicated, most of them absent in organizations that feel slow.

An intake standard. A defined bar that something must clear before it reaches the executive: what question is being asked, what the options are, what the recommendation is, what it costs to be wrong, and who has already been consulted. Material that does not clear the bar goes back. This feels bureaucratic for about six weeks and then becomes the single largest source of reclaimed executive time, because the majority of what reaches a chief executive's desk is not a decision request at all — it is an unresolved disagreement between two functions being escalated in the hope that someone will adjudicate it. Making the intake standard explicit forces that resolution to happen at the right level.

A decision queue with explicit prioritisation. Not everything is urgent, and the executive is a shared resource being allocated. Someone has to hold the queue, sequence it against what the organization is actually trying to accomplish this quarter, and be willing to tell a senior leader that their item is not this week's item. This requires the authority discussed above; without it, the queue is FIFO by whoever is most persistent, which is the default state of most organizations.

A stated decision-rights model. Who decides what, at what threshold, and who must be consulted versus informed. The specific framework matters far less than that one exists and is written down. A large share of the decisions escalated to a chief executive are escalated because nobody is confident they have the right to make them, which is a documentation failure rather than a judgment failure.

A decision record. What was decided, the reasoning, what was explicitly rejected and why, and what would cause it to be revisited. The rejected-alternatives section is the one people skip and the one that pays. Six months later, when someone proposes the thing that was considered and rejected, the record turns a forty-minute rediscussion into a two-minute reference.

A propagation mechanism. Decisions do not travel by osmosis. Someone must determine who needs to know, in what form, with what context about the reasoning — because people who understand why adapt intelligently when circumstances change, and people who received only the what comply literally and stop at the first unanticipated obstacle. This is the commercial version of commander's intent and it is the difference between a decision that survives contact with the organization and one that quietly dissolves.

Run these five well and organizational velocity improves measurably without anyone working longer hours. That is the Chief of Staff's actual product.

Part VI — The Talent Pool Is Six Pools

Here is the thesis this paper most wants to land.

Organizations searching for a Chief of Staff overwhelmingly begin in one of two places: the executive assistant track, or the consulting/strategy track. Both are legitimate. Neither is sufficient, and searching only these two means competing hardest in the most crowded segments of the market while ignoring several sources of well-qualified candidates who would be delighted to be asked.

Let us take the pools in turn, honestly, including where each hits a ceiling.

Executive assistants and business support leaders

The traditional path, and the source of some of the finest Chiefs of Staff working today. Their advantages are real and frequently underrated by people who have never worked closely with a great one.

They already possess the thing that takes everyone else the longest to acquire: an accurate, current model of how the principal actually operates — what they care about, how they read, what they will not say in a group, who they trust, what makes them impatient. They have practised discretion under pressure. They understand access as a strategic instrument. They have relationships across the organization that cross hierarchical lines, and they are frequently trusted by people who would not confide in a strategy director.

Where the ceiling appears is content authority and analytical depth — not capability, but credential and practice. A Chief of Staff drawn from this track may be excellent at knowing that the gross-margin discussion keeps stalling and less equipped to determine that the underlying model is wrong. In organizations where the role is primarily gatekeeper and proxy, this is not a limitation at all. In organizations where the role must engage with substance, it requires deliberate development, and the good news is that this is a far more tractable development problem than the reverse. Teaching a trusted, politically fluent operator to interrogate a financial model is easier than teaching a brilliant analyst to be trusted.

The mistake organizations make is treating this as the only path, or as a promotion rather than a role change. It is a different job with different success criteria and it needs to be scoped as one.

Business analysts and FP&A

This is the pool we most consistently find undervalued, and it is the one the market is slowest to consider.

Business analysts spend their careers doing something unusually well-matched to the modern version of this role: taking messy, partial, contradictory information and producing something a decision-maker can act on. They are trained to be suspicious of numbers. They know where data comes from, what it excludes, and how it can be made to say things it should not. In an environment where fluent machine-generated analysis arrives constantly and requires verification, that reflex is precisely the scarce skill.

They also tend to have unusually broad organizational visibility. An FP&A analyst supporting three functions has seen more of how the company actually works than most directors, because they have had to reconcile what each function claims with what the numbers show.

What they typically lack is executive presence and political fluency — the practised ability to be in a room with a chief executive and a board member and know when to speak. This is learnable and is usually a matter of exposure rather than aptitude. What they should be assessed on is whether they have ever changed a senior person's mind, and how they did it.

We have placed business analysts into Chief of Staff roles who became the strongest performers in a leadership team within two years. The market prices them as back-office. That is an arbitrage.

Program, PMO, and business operations leaders

The natural source for the program-driver archetype and, in our experience, the most reliable pool when the primary problem is that decisions get made and nothing happens.

They arrive with the core competency already built: driving outcomes across functions where they have no authority, which is the defining condition of the Chief of Staff role. They understand dependencies, sequencing, and the difference between a plan and a commitment. They have usually developed a high tolerance for the specific frustration of chasing people more senior than themselves, which is a genuine and underappreciated qualification.

The ceiling is that some are process-oriented in a way that does not translate: excellent at running a cadence, less comfortable with the ambiguity and improvisation the role demands when something is genuinely on fire. Assess for judgment about when to abandon the process.

Management consultants and strategy associates

The most commonly recruited pool after EAs, and a good one, with a well-understood failure mode.

They bring structured problem-solving, comfort with ambiguity, speed, and an ability to construct an argument for a senior audience. Two to four years in a top firm produces a person who can be dropped into an unfamiliar domain and be useful within a fortnight.

The failure mode is ownership. Consulting trains people to produce recommendations and move on. The Chief of Staff role is about what happens after the recommendation, in an organization where you will still be present in six months and where the people you pushed have long memories. Consultants who succeed in the role are the ones who have already made this transition, usually visible in whether they have ever owned an outcome end to end rather than advised on one.

A second, subtler issue: a consultant hired as Chief of Staff frequently treats the role as an eighteen-month waypoint to a P&L. That is a legitimate deal and it can work extremely well — but it should be an explicit deal, not a private plan, or you will be re-running the search unexpectedly.

Military staff officers and veterans

Structurally the closest analogue that exists, and the most under-recruited by a considerable margin.

A field-grade staff officer has done this job, in an environment with higher stakes and worse information. They understand commander's intent — the practice of communicating the purpose behind a decision so subordinates can adapt intelligently when circumstances change, which is exactly what propagating an executive decision through a large organization requires. They are practised at synthesising incomplete information under time pressure, running a battle rhythm, and briefing a senior decision-maker crisply.

The translation friction is vocabulary and commercial context, not capability, and it is usually resolved within a quarter. Organizations that have built a habit of recruiting here tend to keep doing it.

Internal operators from the line

Finally, the pool most likely to be sitting in your building: someone who has run a function or a large team, knows the business substantively, has credibility with peers, and is ready for a broader view.

Their advantage is that they arrive with trust and context already built, which are the two things that take longest. Their disadvantage is that they carry history — existing alliances, prior conflicts, a reputation formed in a narrower role — and they may struggle to be seen as neutral.

This is also the pool where the role's value as a development mechanism is highest, which we will return to, because it is the strongest argument for the seat that has nothing to do with the executive it serves.

The practical implication. Rank your archetypes first, then map to pools deliberately. A gatekeeper-primary role should be searched heavily in the EA and internal-operator pools. A program-driver-primary role should start with PMO, business operations, and military staff. A strategic-partner-primary role should look at consulting, FP&A, and internal operators. Almost no organization does this mapping, which is why almost every Chief of Staff search feels harder than it should.

The role is hard to fill because the search is aimed at one or two pools. The competency is distributed across six, and four of them are not being asked.

Part VII — The Competency Model

Strip away background and the seat requires eight things. This is what we assess against, and the assessment method matters as much as the list.

Judgment about salience. Can they look at twelve things and correctly identify the two that matter to this principal this week? Assess with a live exercise: give them a realistic pile of material and a stated executive priority, and ask what reaches the desk and what does not. Then ask why. The reasoning is more diagnostic than the answer.

Verification instinct. Do they check? Give them a fluent, well-formatted analysis containing two embedded errors — one arithmetic, one a subtler framing problem such as a comparison against a non-comparable base. Strong candidates find at least one and, more tellingly, express calibrated uncertainty about the rest. Weak candidates praise the structure.

Influence without authority. The core political competency. Ask for a specific instance of getting something done through someone senior who did not report to them and did not initially want to. Push for mechanism: what exactly did they say, what did they trade, what did they do when the first attempt failed. Generic answers here are disqualifying.

Written clarity under compression. The role runs on writing. A one-page decision memo, written live from a messy brief, is the single highest-signal exercise available. Look for whether the recommendation appears in the first three lines, whether the alternatives are represented fairly, and whether the uncertainty is stated rather than smoothed over.

Discretion and trust. Hardest to assess and most important. Reference checks carry more weight here than interviews. The useful question to a referee is not "were they discreet" but "did people tell them things they did not have to."

Follow-through. Ask about a decision they were responsible for propagating. What was the mechanism? How did they know it had landed? What did they do about the team that quietly ignored it? Candidates who have really done this have unglamorous, specific systems, and they will describe them without prompting.

Systems fluency with AI tools. Not model expertise — fluency. Can they use these tools to compress the gathering work, and do they have a working sense of where the output fails? The good answer is specific and slightly jaded: they will describe a category of task they have stopped delegating to a model because it was reliably wrong in a way that was hard to spot. Enthusiasm without that texture indicates limited real use.

Comfort with subordinated ego. The role is powerful and largely uncredited. The wins are the executive's. Candidates who need visible attribution will be unhappy within a year, and it is kinder to establish this in the interview than in month nine. A useful question: describe something you are proud of that nobody outside a small group knows you did.

Notice how little of this is background-specific. Every one of the six pools produces candidates who score well on some of these and poorly on others, and the specific profile varies more within pools than between them. This is the strongest practical argument for widening the search.

Part VIII — The Adjacent Seats

A recurring source of confusion, and a frequent cause of an expensive duplicate hire, is that several roles sit near this one. Distinguishing them properly is usually the fastest way to work out whether you need a Chief of Staff at all.

Chief Operating Officer. The COO owns operations with line authority and a P&L relationship. The Chief of Staff owns coordination with borrowed authority and no line. The distinction is not seniority; it is whether the person has an organization. A COO who is functioning as a Chief of Staff has been mis-titled, and a Chief of Staff being asked to run operations without reporting lines has been mis-scoped. If your actual problem is that nobody is running the business day to day, you need a COO and a Chief of Staff will not substitute.

Deputy or second-in-command. A deputy acts for the executive with full substitutability and is usually a successor candidate. A Chief of Staff acts on behalf of the executive within a defined remit. The proxy archetype approaches deputy territory, which is why it requires the most trust and is the least frequently hired externally. Confusing the two creates a succession expectation nobody intended.

Head of Strategy. Owns the analytical agenda and produces the thinking. The Chief of Staff owns the decision process and what happens after the thinking. These are complementary and frequently both exist; where organizations get into trouble is hiring a Head of Strategy to solve a follow-through problem, which produces excellent documents and no change in velocity.

Business Manager. In some large technology organizations this title covers the operational cadence, planning, and budget mechanics for a business unit. It overlaps heavily with the program-driver archetype and is often the same job under a different name. If your organization already has these, you may not need a separate seat — you may need to give one of them access and authority.

Technical Advisor. A pattern common in engineering-led companies: a strong technical individual rotating through a period shadowing a senior leader, providing technical judgment and translation. This is essentially the strategic-partner archetype with a technical rather than commercial centre of gravity, and it is one of the better-designed versions of the role because the rotation and exit are explicit from the outset.

Executive Assistant. Distinct profession, distinct accountabilities. The EA is accountable for the executive's operating logistics and is frequently the person with the most accurate real-time picture of the executive's world. The Chief of Staff is accountable for outcomes across the organization. Strong pairs work closely and the boundary between them is negotiated rather than fixed; what does not work is collapsing the two roles into one job with one salary, which is a reliable way to lose a good EA and fail to acquire a Chief of Staff.

A short diagnostic

Four questions, honestly answered, will tell you which seat you need.

Do decisions get made and then not happen? That is a Chief of Staff problem, program-driver archetype.

Do decisions not get made at all, because the executive is saturated? Also a Chief of Staff problem, gatekeeper-primary.

Is the thinking weak — are you deciding well on bad analysis? That is a strategy or analytics problem first. A Chief of Staff will not fix it and may mask it.

Is nobody running the day-to-day business? That is a COO problem, and hiring a Chief of Staff instead is one of the more expensive substitutions available, because it puts a person with no authority in front of a structural gap.

If the answer to the first two is yes and to the second two is no, you have a clear case. If everything is yes, you have a scale problem that no single hire resolves, and the sequencing question is more important than the search.

Part IX — How the Seat Fails

Five failure modes account for nearly all of the bad outcomes we see, and four of them are structural rather than about the person.

Responsibility without authority. The role is defined as driving the executive's priorities, and then the Chief of Staff has no standing to convene, no ability to require a response, and no visible backing when a senior leader declines to cooperate. The result is a talented person spending their days asking politely. This is the single most common failure and it is entirely the principal's fault. The fix costs one sentence, said publicly by the executive, and repeated: when they ask, it is because I am asking.

The blended job description. Covered above. Hiring against four archetypes produces disappointment measured against whichever three the candidate is not.

The wrong reporting line. A Chief of Staff who reports to anyone other than the executive they serve is structurally compromised, because their access and their independence both flow from that relationship. Reporting into a COO or a Chief of Staff office might look tidier on an org chart and it removes the basis of the role.

No exit path. The role has a natural life of two to four years. It burns out beyond that and, more importantly, its development value depends on there being a next thing. Organizations that do not plan the exit lose the person entirely, usually to a competitor who offers them the operating role their own employer never discussed. The exit should be a topic in the first month, not the last.

The principal will not actually delegate. Some executives hire a Chief of Staff as a symbol of scale and then continue to run everything through themselves. No candidate survives this and no search fixes it. It is worth testing directly before accepting an engagement: ask the executive what they will stop doing. If the answer is vague, the role will fail regardless of who fills it.

Part X — Structure, Compensation, and Tenure

Practical parameters, stated plainly because they are frequently gotten wrong.

Reporting. Direct to the principal. No exceptions that survive contact with the work.

Leadership team access. The Chief of Staff should attend the executive team meeting. Whether they speak is a matter of style and seniority; whether they are present is not optional, because a person expected to propagate decisions who did not hear them being made is running on a copy.

Compensation. This is where organizations most often misprice, because they band the role against its source pool rather than its scope. A Chief of Staff to a chief executive of a substantial company is doing work with organizational consequence comparable to a senior director or vice president, and should be banded accordingly, with the specific band varying by which archetype dominates. Banding against an executive-assistant scale when the role is program-driver-primary is a reliable way to lose the person in eighteen months, and the replacement cost dwarfs the saving.

Equity, where it exists, should be at the level implied by the band and not below it. This is a retention-critical role with unusually high context loss on departure.

Tenure and succession. Treat it as a two-to-four year appointment with an explicit next step. The best structures we see name the successor a full two quarters before the transition and run an overlap, because the handover involves a great deal of tacit context — exactly the kind of undocumented judgment described in the ghosts paper. A Chief of Staff transition without overlap loses more institutional knowledge than almost any other departure at that level.

Team. At sufficient scale the role acquires one or two people. Resist this longer than feels comfortable. The value of the seat comes substantially from its proximity and its lack of bureaucratic weight; a Chief of Staff office develops its own interests remarkably quickly.

Part XI — The Argument Nobody Makes: It Is a Leadership Pipeline

The strongest case for this role has little to do with the executive it serves.

A Chief of Staff spends two to four years with a complete view of how an enterprise actually operates — where decisions are really made, which relationships carry weight, how capital gets allocated, why the thing everyone agreed to in March still has not happened. There is no other seat in the organization that provides this, and no training program that approximates it. It is the single most efficient leadership development mechanism available, and most organizations use it accidentally.

Used deliberately, it changes the character of the appointment. You are not only solving an executive-capacity problem; you are running a two-year immersion for a future senior leader, and the second outcome may be worth more than the first. This argues for weighting internal candidates more heavily than most searches do, for treating the exit as the point rather than the risk, and for being willing to appoint someone whose current gaps are real but developable — because development is part of the return.

It also changes who you should be looking at. If the seat is a pipeline, the business analyst with exceptional judgment and no executive exposure is not a compromise candidate. They are the highest-return one.

Organizations that have run this deliberately for a decade have a recognisable characteristic: an unusual number of their senior operators came through the role, and they share a common understanding of how the company works that is difficult to build any other way.

Part XII — For the Candidate

A note for people reading this from the other side, because the market is moving faster than the career advice has.

If you are an EA, an analyst, a program manager, or an operator wondering whether this seat is available to you: it is, and the constraint is usually that nobody has told you to consider it.

Three things make a candidacy credible.

Evidence of moving something you did not control. One well-documented instance of driving an outcome across functions, against resistance, where you can describe the mechanism specifically, is worth more than a title. Collect these deliberately. Most people have one or two and have never framed them this way.

Written work you can show. The role runs on compressed writing. A decision memo you actually wrote — recommendation first, alternatives fairly represented, uncertainty stated — is the strongest artifact you can bring. If you do not have one, write one this month about a real decision in your current organization.

A worked view on verification. Be able to describe, concretely, a category of work where you have learned not to trust machine output, and how you check. This single answer distinguishes candidates more sharply than any other in current interviews, because it demonstrates real use rather than enthusiasm.

What to be clear-eyed about: the role is uncredited by design, the hours are genuinely demanding, your success is someone else's success, and it ends. Take it for the view and the compounding of judgment, and negotiate the exit at the front.

Part XIII — Three Composites

Details changed, patterns intact.

The analyst nobody had considered

A healthcare services company had been searching for a Chief of Staff to the CEO for five months. The specification asked for consulting background, healthcare domain depth, and executive presence. Three finalists had been through the process; two declined at offer over compensation, and the third was rejected by the CEO after a final conversation he described as "impressive and somehow not right."

We reframed the search by first asking what was actually broken. The answer, once we pushed, had nothing to do with strategy: decisions were being made in the executive meeting and dying afterwards. The CEO was re-litigating the same three questions monthly. This was a program-driver problem being recruited as a strategic-partner role.

The person hired was an FP&A manager two levels down inside the company. She had never held a strategy title. What she had was four years of reconciling what each function claimed against what the numbers showed, relationships with every operating leader from budget cycles, and a reputation for being unwelcomely accurate. In the assessment she found both planted errors in the analysis exercise in under six minutes, including the base-comparison one that four external finalists had missed.

Within two quarters the monthly re-litigation had stopped, largely because she had instituted a decision record with a rejected-alternatives section. She is now a divisional COO. The search that nearly failed for five months resolved in three weeks once the archetype was named correctly and the pool widened.

The proxy who was given a title and no cover

An industrial manufacturer hired an experienced Chief of Staff from a larger competitor — genuinely strong, well-referenced, with a track record. Eleven months later he resigned.

The exit interview was unambiguous. He had been asked to drive the CEO's priorities across a leadership team that had not been told what his role was. When he convened meetings, attendance was optional in practice. When he asked a business unit head for a commitment date, he was told it would be discussed directly with the CEO. He had spent a year as a highly paid person sending follow-up emails.

The CEO's view, offered honestly, was that he had not wanted to appear to be creating a layer, so he had introduced the hire lightly and left the role to establish itself.

There was no candidate problem here. The role failed at the announcement. The replacement search succeeded on a nearly identical profile after the CEO opened three consecutive leadership meetings by routing his own requests explicitly through the seat.

The Chief of Staff who became an information control

A financial services firm's Chief of Staff, an ex-Army major, noticed that board materials were increasingly assembled from AI-generated syntheses produced across several functions, with no consistent verification and no record of what had been checked.

Rather than escalating it as a risk issue, she made it part of her role: nothing reached the board pack without a provenance line, material claims were traced to source systems, and she maintained a rotating deep-check sample. In the first quarter she caught a market-sizing figure that double-counted a customer segment and would have appeared in a board discussion about capital allocation.

What is instructive is what happened next. Internal audit, reviewing the board reporting process a year later, treated her routine as a documented control — the only one of its kind in the firm — and recommended it be extended to two other reporting streams. The role had, without anyone planning it, become the organization's first functioning human-in-the-loop design for executive information.

Part XIV — Objections

"This is just an executive assistant with a better title."

It is not, and the conflation does a disservice in both directions. The senior EA function is a real profession with its own depth, and the Chief of Staff role has different accountabilities — outcomes across the organization rather than support of a person. Some of the best Chiefs of Staff came through that track, which is an argument for the pool's quality, not for the roles being identical.

"We are too small."

Probably true below a few hundred people, where the executive can hold the whole picture and the coordination cost is low. The signal to watch is the one in the opening paragraph: a growing backlog of decisions that were ready and did not get made, which persists after the calendar has been cut. When that appears, the constraint has moved.

"AI will absorb this role in three years."

The prediction has the direction backwards. AI absorbed the inputs to this role, which is why it became scarce. What remains is judgment about salience, verification, political propagation, and trust — the least automatable cluster in the enterprise. If anything, improvements in analytical capability increase the value of the seat, because they increase the volume of material requiring exactly the scrutiny the role provides.

"Great ones are impossible to find."

They are impossible to find in the two pools everyone searches. They are findable in six. That is most of what this paper is arguing.

"Creating this role adds a layer and slows us down."

It can, and the failure mode is real enough to name precisely: it happens when the seat becomes a routing point rather than a resolution point. If every request now stops at the Chief of Staff and waits, you have added latency and called it governance.

The distinguishing test is what the role does with an item it receives. A resolution point disposes of most things — decides them at its own level, sends them back for rework, or routes them to the right owner — and passes upward only what genuinely requires the executive. A routing point queues everything and forwards it with a summary. The first reduces load; the second adds a step.

This is largely determined by authority. A Chief of Staff who is permitted to say no on the executive's behalf resolves. One who is not can only forward. If you are worried about adding a layer, the mitigation is not to weaken the role — it is to strengthen it, which is counterintuitive and correct.

"Our executive already has a great EA. Isn't this redundant?"

No, and the pairing is usually the strongest configuration available. The EA holds the executive's operating reality — the calendar, the relationships, the real-time picture of what is happening today. The Chief of Staff holds the organizational agenda. They solve different constraints and the best pairs are in constant contact.

What is genuinely redundant is hiring a Chief of Staff to do gatekeeping that an excellent EA is already doing well. If your EA is effectively running access and priority, then your gatekeeper archetype is filled, and the question becomes whether you have a strategic-partner or program-driver gap. Frequently you do, and recognising that the gatekeeper seat is already occupied makes the remaining search substantially easier to specify.

The version of this that goes badly is promoting an excellent EA into a Chief of Staff role to solve a program-driver problem, and thereby losing an excellent EA and acquiring a struggling Chief of Staff. If the EA wants that path — and some do, and are outstanding at it — then backfill the EA seat properly and support the transition. Do not treat one person as the solution to two vacancies.

Conclusion — The Seat on the Gap

The condition that made this role urgent is not going to reverse. Analysis will keep getting cheaper and more plentiful. The volume of decision-ready material arriving at senior desks will keep rising. And the human capacity to absorb, verify, decide, and propagate will remain roughly what it has always been.

That gap is now the primary constraint on organizational velocity in a large number of enterprises, and it is not a technology problem. No model closes it, because every improvement in the model widens the input side of the gap rather than the output side.

What closes it is a person: someone with the judgment to know what matters, the rigour to check what arrives, the standing to be believed, and the persistence to make a decision real once it is made. That person has existed in military staffs for two centuries and in the West Wing since 1946, and the reason the role is having a moment in the enterprise is not that it is new. It is that the thing it was always for has become the thing that matters most.

Define which archetype you actually need. Grant the authority along with the responsibility. Search all six pools, not two. Price it against the scope rather than the source. Plan the exit at the start.

And when you build the shortlist, look carefully at the analyst who has spent four years being right about numbers other people acted on. The market has not priced them yet.

Appendix A — The Interview Guide

Structured around the eight competencies. Use the exercises; the conversation alone will not separate candidates at this level.

Exercise one: the inbox (salience)

Provide twelve realistic items — a board request, two escalations, a partnership inbound, a resignation, a customer issue, a budget variance, three status reports, a press inquiry, a legal notice — plus a one-paragraph statement of the executive's current top three priorities.

Which three reach the desk this week? Which do you handle yourself? Which do you send back, and what do you say?

Listen for reasoning tied to the stated priorities rather than to intrinsic urgency. Strong candidates will send at least two items back and will articulate what those items were missing.

Exercise two: the flawed analysis (verification)

Provide a two-page, well-formatted, confident analysis containing one arithmetic error and one base-comparison error.

You have fifteen minutes. Would you put this in front of the CEO?

Strong candidates find at least one error, flag uncertainty about what they have not checked, and say what they would verify before it went further. Weak candidates comment on the structure and the writing.

Exercise three: the decision memo (written clarity)

Provide a messy brief — conflicting inputs from three functions on a real-shaped decision.

One page. Thirty minutes.

Look for: the recommendation in the first three lines, alternatives represented fairly enough that their advocates would recognise them, uncertainty stated rather than smoothed, and a clear statement of what would change the recommendation.

Conversation questions that carry weight

• Tell me about getting something done through someone senior who did not report to you and did not want to. What did you actually say? What did you do when the first attempt failed?
• Describe a decision you were responsible for propagating. How did you know it had landed? What did you do about the team that ignored it?
• What kind of work have you stopped giving to AI tools, and what went wrong that taught you that?
• Tell me about something you are proud of that almost nobody knows you did.
• What would you need from the executive in your first month for this to work?
• What do you want to be doing in three years?

That last question is not a formality. If the answer is "this role," probe carefully — the seat has a natural life and a candidate with no forward view often becomes a retention problem in year three. If the answer is an operating role, you have a development deal to structure explicitly.

Reference questions

• Did people tell them things they did not have to?
• When they disagreed with the principal, how did that go?
• What happened after they left?

Appendix B — The First Ninety Days

A structure that works, adapted to archetype.

Days 1–30: listen and map. Thirty conversations minimum, across every function and at least two levels down. The questions are consistent: what is stuck, what decision are you waiting on, what do you wish the executive understood, and what happens here that would surprise them.

Produce one artifact at the end of the month: a written map of where decisions are actually made, where they are stalling, and the three things you believe are most costly. Deliver it to the executive privately. This document establishes whether the appointment will work.

Change nothing in this period beyond the obviously broken.

Days 31–60: install one thing. Pick the single highest-leverage component of the decision operating system — usually the intake standard or the decision record — and implement it properly rather than implementing three things partially.

This is also when the authority question gets settled in practice. The first time a senior leader tests the seat, the executive's response is watched by everyone. Make sure it has been discussed in advance.

Days 61–90: establish the cadence. The weekly rhythm should now be running: the pre-read discipline, the follow-through ledger, the standing checks. Deliver a second written assessment covering what changed, what did not, and what you need.

The single most important thing in the first month is not a deliverable. It is the standing agreement with the principal about how disagreement works — whether you are expected to push back, in what setting, and how hard. Chiefs of Staff who never establish this become expensive relays. Those who establish it early become the person the executive checks their thinking against, which is the whole point of the seat.

---

Search & Strategy runs the Orchestration Practice: retained search for the leaders who own applied AI, data, and platform — and the human-in-the-loop operators who turn decisions into execution. If you are defining a Chief of Staff seat, or wondering whether you are ready for one, start a conversation.