Taking an assessment rather than buying one? This page is written for employers. Here is the page for candidates.

Administrative and business support · Mid level

How to assess a Executive Assistant

The CV screen measures proximity rather than judgment: four years beside a CEO is evidence of tenure and of the CEO's retention, not of a single decision the candidate made. The reference is the strongest instrument in this process and also the one most likely to describe a relationship rather than a capability, because a principal's assessment of their own EA is inseparable from how much they liked working with them. And the central act of the job — turning five things that all claim to be urgent into three handled, one escalated with a recommendation attached, and one declined — is never once observed before the offer is made. This is also the role where the interview is most actively misleading, because articulacy about prioritisation and skill at prioritising are close enough to be mistaken for each other by an interviewer who has never seen the candidate's inbox.

The executive assistant role is the most consequential job in this family and the least well understood by the people hiring for it. The job description will describe calendar management, travel, expenses and inbox support. Those are the mechanisms. The function is to make a scarce, expensive, over-committed person's attention go to the right things, which means the EA is making prioritisation decisions on behalf of someone more senior all day, mostly without asking.

Consider a specific morning, because the abstraction hides the difficulty. At 08:40 the principal has a board paper due at noon and has not started it. In the inbox: a client who was promised a call yesterday and is now writing to the principal's manager; a direct report asking for thirty minutes on something they have described only as "a bit sensitive"; a conference organiser needing a final answer on a keynote in six weeks; a flight cancellation that breaks tomorrow's two-city day; and a family commitment already in the calendar at four o'clock that the principal has moved twice. Every one of these is legitimate. Four of them are being escalated by someone who believes theirs is the urgent one, and one of them is being escalated by nobody at all — which, as it happens, is the one that will become the real problem.

A weak EA presents all five to the principal, correctly labelled, and calls that support. It is not support; it is transferring the work back to the person whose time the role exists to protect, and it is the most common failure by a wide margin. A different weak EA decides all five without reference to anyone, which looks like initiative until the day they decline something on the principal's behalf that the principal would have moved the board paper for. The strong behaviour sits between and is quite specific: handle what can be handled and say so briefly, escalate what genuinely needs a decision with a recommendation attached rather than a question, and be explicit about what has been declined so the principal is never surprised by something they did not know was refused in their name.

The written component is where the role is most exposed and most gradeable. Much of an EA's output leaves the building under someone else's name, to clients, board members, investors and journalists. Writing a "no" that preserves a relationship is a genuine craft: it has to be quick enough not to be a negotiation, warm enough not to be a snub, specific enough that the recipient does not simply ask again next week, and written in a voice that is recognisably the principal's rather than the assistant's. Very few candidates can do this and almost none are ever asked to demonstrate it before being hired. It is trivially easy to set as a task and it discriminates immediately.

The last differentiator is discretion, which is different from confidentiality. Confidentiality is not repeating things. Discretion is knowing which things are information the principal needs and which are noise that will merely worry them, and knowing that a direct report's "a bit sensitive" may be a grievance that must not be summarised in a calendar invitation title. An EA hears the organisation before the organisation is ready to be heard, and the judgment about what to do with that is not covered by any policy.

One honest limit belongs on this page rather than in a footnote. Executive assistant hiring has a legitimate component of personal fit with a specific principal — of working rhythm, tolerance for interruption, and how someone likes to be told bad news. No assessment should claim to measure that, and a strong result here does not settle it. What an assessment can do is rank prioritisation, written judgment and discretion honestly, so that the conversation about fit happens with the three or four candidates who can actually do the work rather than the three or four who interview well.

What the job actually needs

• triage under conflicting urgency
• delegated judgment
• writing in someone else's voice
• discretion
• declining on a principal's behalf without damage

How people fail in this seat

• forwards everything and decides nothing
• decides everything and becomes the bottleneck
• shields the principal from information they needed
• writes an external refusal that reads as a slight
• repeats what they overheard
• protects the calendar so well that nothing gets through

What most employers do instead

CV screen weighted to the seniority of executives previously supported, a conversational interview, a meeting with the principal for chemistry, and a reference from the previous principal.

The assessment

About 40 minutes end to end.

The systems it runs in

A mailbox and calendar the candidate holds delegate access to, not their own: Outlook with delegate permissions on the principal's inbox and diary is the reference build, and published advertisements for this seat routinely ask for exactly that — delegate access, shared mailboxes and complex calendar rules — because sending as somebody else is the mechanical fact the role turns on. That matters more here than in any other seat on this hub, because three of the four messages in t2 leave the building under the principal's name, and they are written and sent from the principal's account rather than described in a text box. A chat client carries the corridor traffic, and a travel and expense platform is the fourth surface: the cancelled flight that breaks tomorrow's two-city day is rebooked in the tool the organisation books in, with the itinerary, the cost and the approval that go with it. Concur and Navan are the named reference builds, and Concur experience is a stated requirement on a visible share of executive assistant postings. Naming the calendar changes what one card observes directly. The direct report's meeting is not answered as a sentence about discretion; it is created as an invitation, and the title, the attendee list and the visibility setting on it are the artefact. An assistant who titles it with the subject and leaves it visible to a shared diary has disclosed something, and no written answer would have shown that.

• Microsoft Outlook
• Microsoft 365
• Gmail
• Google Calendar
• Microsoft Teams
• Slack
• SAP Concur
• Navan
• TravelPerk

Google Workspace with delegated Gmail and Calendar access is the standing swap and the fixture is maintained in both. Any chat client substitutes for the corridor traffic, and any travel or expense platform substitutes for the rebooking, which is scored on what the itinerary and the note to the principal contain rather than on operating the booking flow. The fixture adds no HR system and no personnel record, deliberately: the direct report's sensitive request is routed, never looked up, and there is nowhere in this environment for a candidate to go and find out what it is about.

What the candidate actually does

The mark scheme

Each criterion is scored 1 to 5 against written anchors, and every score is reported with the excerpt that earned it. A criterion marked floored is reported as a finding rather than averaged into the total. The first is open; open any other to read its anchors in full.

How it is scored

Weighted mean of the five criteria, each scored 1 to 5 against the anchors and reported with the disposition, message or transcript line that earned it. Three raw counts are printed beside the score and annotated explicitly as not positive signals: how many of the six items were handled, escalated and declined. Buyers read five-handled-of-six as initiative, and in this fixture it is the profile most likely to produce the failure the role is hired to avoid. The report also shows the locked triage against what the four messages actually say and against what was said on the call, because the gap between what a candidate decided in private and what they defended out loud is frequently the most informative artefact the session produces. The call is scored from the transcript, which is the reviewer's default and only view; audio is retained for dispute and is not a scoring surface, and no criterion is named clarity, professionalism or polish.

Integrity

• one unbroken monitored sitting in the order t1, t2, t3, t4, with the triage timestamped and locked before the writing opens
• the four messages are graded against the candidate's own triage and the brief-back against a transcript the candidate cannot re-read, which is the strongest anti-coaching control available here: a prepared decline cannot match a triage it has not seen, and an off-camera coach cannot supply a summary of a conversation that went somewhere unplanned
• fixture variants rotated between sittings so the identity of the quiet item and the content of the client escalation differ
• one live follow-up question at the mid band, which changes a requirement rather than raising difficulty: the candidate is told the principal has just agreed to speak at the conference after all, and asked what they now write to the organiser and what they do about the four o'clock commitment
• session log records time distribution and paste-versus-typed provenance in the four messages; no automated integrity verdict is produced from it

The log describes what happened. It does not produce a cheating verdict — the follow-up conversation is the control, because a statistical accusation is not something we would ask a reviewer to defend.

What you receive

• the timestamped triage with a disposition, an action and an order for all six items
• the four messages as sent, including the message to the principal
• full call transcript with the principal's interruption time-marked
• the four card answers with their one-line reasons
• handled, escalated and declined counts, annotated as not positive signals
• per-criterion score with the disposition, excerpt or transcript line that earned it

Who decides

Required, and for a reason particular to this role rather than a general preference. Executive assistant hiring has a legitimate component of personal fit with a specific principal — working rhythm, tolerance for interruption, how someone likes to be told bad news, how much they want decided for them — and no assessment should claim to measure it. This one does not, and a strong result here does not settle it. What the design does is rank prioritisation, written judgment and discretion honestly, so that the conversation about fit happens with the three or four candidates who can do the work rather than the three or four who interview well. The reviewer has two specific jobs. The first is to set the delegation line before scoring: a chief executive who wants five of six items decided for them and one who wants to see everything are both real, and the same triage is correct for one and wrong for the other, so the first criterion is confirmed or overridden in writing against this principal's actual preference. The second is the voice criterion, which is a judgment about register that a reviewer who knows the principal can make and a rubric cannot.

What this does not measure

The sensitive item is scored on routing and discretion only. The candidate is never scored on any view of the direct report's circumstances; the fixture never states what the matter is, the candidate is not permitted to ask, and no anchor rewards or penalises an attitude to it. What is scored is whether the substance stayed out of the artefacts where it does not belong, whether the candidate considered that this might belong on a different route entirely, and whether the principal was given enough to prepare without being given more than they were owed. A reviewer who finds themselves reasoning about the underlying circumstance has left the rubric. The voice criterion is the second place this design could go wrong, and it is constrained deliberately: it is scored against two supplied messages the principal actually wrote, so the target is a specific document rather than an idea of how executives sound, and it is not satisfiable by formality, vocabulary or idiom. Scoring the call from the transcript rather than the audio is the structural control on accent and speech variation, and no criterion here is named clarity, professionalism or executive presence, which are the labels under which that filter conventionally re-enters a rubric for assistant roles. The rubric dimensions are published to candidates in advance, because this fork survives disclosure: knowing that the split is scored does not make handling all six less tempting when four senders are pressing, and knowing that discretion is scored does not tell a candidate which of the six items is the quiet one. What the design cannot reach is the accumulation. An executive assistant's real performance is a year of mornings, and one morning samples the judgment without observing whether it holds when the principal is tired, the relationship is three months old, and the last two escalations were received badly.

The central act of this job is turning several things that all claim to be urgent into a few that are handled, one that is escalated with a recommendation attached, and one that is declined — and it is never once observed before the offer is made. The CV screen measures proximity to seniority, which is evidence of tenure rather than of a single decision the candidate made. The reference describes a relationship, because a principal's view of their own assistant is inseparable from how much they enjoyed working with them. And the interview is actively misleading here, since articulacy about prioritisation and skill at prioritising are close enough to be mistaken for one another by anybody who has never seen the candidate's inbox. This design puts the inbox in front of them instead.

The six items are the morning from the role page, reproduced because the abstraction hides the difficulty. Every one of them is legitimate. Four are being pushed by somebody who believes theirs is the urgent one. One is being pushed by nobody at all, and nothing in the fixture points at it — no formatting, no ordering, no hint in the framing. If a candidate can tell from the presentation which item is the trap, the exercise has become reading comprehension of a scenario rather than judgment about a morning.

Both failure modes are available and both score 1. Presenting all six to the principal, correctly labelled, is the commonest weak answer by a wide margin, and it is not support: it transfers the work back to the person whose attention the role exists to protect. Deciding all six without reference to anybody is the other failure, and it looks like initiative until the day something is declined in the principal's name that they would have moved the board paper for. Scoring both at 1 and printing the three raw counts on the report is the only way to stop a buyer reading five-handled-of-six as a strong result.

The triage is locked and timestamped before anything else opens. That single mechanism carries most of the design's weight, because without it a candidate who says one thing in a message and another on a call cannot be shown to have moved — they can always say afterwards that this was the plan. With it, the report can put the private decision, the written artefact and the spoken brief-back side by side, and the gaps between them are usually the most informative thing in the session.

The written segment is sixteen minutes and it is the largest single allocation in the file, because it is where the role is most exposed and most gradeable. Much of an assistant's output leaves the building under somebody else's name. Writing a no that preserves a relationship is a genuine craft with four constraints that pull against each other: quick enough not to become a negotiation, warm enough not to be a snub, specific enough that the recipient does not simply ask again in a fortnight, and in a voice recognisably the principal's rather than the assistant's. To keep that last constraint from becoming a proxy for how executive somebody sounds, the fixture supplies two messages the principal actually wrote, and the anchor is agreement with those documents rather than with an idea of seniority.

The four-minute call exists because prioritisation on paper and prioritisation out loud are not the same skill, and because of one specific observation that only a live segment can produce. The principal asks what they need to know. Four minutes is enough for a competent summary of six items, or for a shorter summary plus one decision actually obtained — and the second is the job. Meanwhile the item the candidate declined is closed, was not asked about, and mentioning it invites a question. Volunteering it anyway is the behaviour that separates an assistant a principal can trust with delegated authority from one who merely exercises it, and it is worth its own criterion.

The sensitive item runs through the whole session and is handled under a constraint stated plainly in the file: routing and discretion are scored, and nothing else. The fixture never says what the matter is, the candidate cannot ask, and no anchor reads their attitude to it. What is graded is whether the substance stayed out of a calendar title and a summary, whether the principal was given enough to prepare and no more, and whether the candidate considered that this might belong somewhere other than with the principal at all. That criterion is floored, because discretion is the one thing in this role where a single failure is not recoverable by strength elsewhere — an assistant hears the organisation before the organisation is ready to be heard, and there is no policy that prevents them repeating it, only the person.

Forty minutes is defensible for a seat that is hired individually, sits closer to the centre of an organisation than its salary band suggests, and is currently screened by a conversation and a reference. Human review is required rather than recommended, and not as a formality: the correct delegation line differs between principals, so the reviewer sets it before scoring, and the fit question the buyer actually cares about is one this design deliberately refuses to answer.

Sources

Every figure on this page is traceable. Where a claim could not be sourced it is stated qualitatively instead.

1. US Bureau of Labor Statistics, Occupational Outlook Handbook, Secretaries and Administrative Assistants, 2025, executive secretaries and executive administrative assistants 489,300 jobs and 0 percent projected change to 2035 against minus 6 percent for the general segment, https://www.bls.gov/ooh/office-and-administrative-support/secretaries-and-administrative-assistants.htm