What 14 executive assistants taught us about the standard.

When I wrote the rubric I had a view about what would separate a Level 3 EA from a Level 2. I thought it would be the big moments: the journalist who already knows something, the launch the CEO wants to confirm against Engineering’s advice. Fourteen sittings in, that isn’t what the results show, and I think both employers and EAs will find the real pattern more useful.

A word on the numbers first. Fourteen people is a beta cohort, not a market survey. They are mostly UK-based, most came through my own network, and one of the nine scenarios was added partway through, so seven of the fourteen sat it. Every result below was scored by three independent reads with the median taken, then moderated by me, and all of them are verifiable. I’ll say where the sample is too small to lean on.

Where the fourteen landed

Level	People	Scores out of 100
Executive Business Partner	3	82, 82, 83
EA to Senior Leadership	6	74 to 81
Operational EA	5	69 to 73

Two things stand out before you get into the detail. The spread is narrow, fourteen points from the lowest score to the highest, which tells you these are all competent EAs and the standard is separating good from very good rather than good from bad. The second is that the Level 3 line at 82 is a real line. The three people above it are there by a point or two, and the three nearest below it are at 81, 79 and 78. The report tells each of them exactly which answers decided it.

The level is decided by breadth, not by one great answer

Every scenario is marked on two or three of the five pillars, so each person collects around twenty-one marks across the day, each a 1, 2 or 3. The three Executive Business Partners scored a 3 on six or seven of their marks. The Operational EAs scored a 3 on between none and three. The EA to Senior Leadership group sat in between, mostly on one to four, with one person at seven who missed the top level for other reasons the report sets out.

The bottom of the scale matters as much. Two of the three Level 3s didn’t score a single 1 all day. The Operational EAs scored between one and six of them. That’s the honest answer to the question I get asked most, which is whether someone can be a Level 3 and still make mistakes. Yes, and one of the three did, with three marks at the bottom. What they can’t do is make them often, and what carried them was the number of scenarios they handled at the top mark, not the absence of errors.

The same shows in what the reviewers flagged. Across the day the three Level 3s picked up 25 standout moments between them and one red flag. The five Operational EAs picked up six standouts and eight red flags. A red flag is a decision that would have cost the executive something real, and at Operational level there was one every sitting or so. That is the clearest single difference in the data.

The hard parts of the day are the ordinary ones

I expected the journalist and the launch decision to be where people lost marks. They were the two scenarios the cohort handled best, with an average mark of 2.2 out of 3 on each. The two they handled worst were the first two of the day: sorting a full inbox with the executive unreachable, at 1.86, and three requests arriving at once with a hard finish to protect, at 1.88.

I think that’s because the dramatic moments announce themselves. Everyone knows a reporter asking about a private conversation is a test, and everyone knows to be careful. A full inbox at ten to nine doesn’t announce itself, and the difference between the levels was whether someone sorted it by what would cost most if it waited or by which subject line sounded loudest. If you’re about to sit the assessment, that’s where to put your attention, and I’ve written separately about what the reviewer is looking for in the inbox. If you’re hiring, it’s the part of the job least visible at interview and most worth testing.

Judgement separates the levels. Execution doesn’t.

Across the eight people for whom I have full pillar averages, Execution came out almost identical at every level, around 2.0 out of 3. Operational EAs got things done as reliably as the Level 3s did. What moved with the level was Judgement, which averaged 1.7 at Operational, 2.2 at EA to Senior Leadership and 2.3 at Executive Business Partner, and Communication, which climbed from 1.8 to 2.3 across the same three groups.

That matches what I see in placements. The EAs who don’t progress are rarely the ones who drop things. They are the ones who do everything they are asked and don’t yet read what they weren’t asked. A job description that lists tasks will find you someone who executes. It won’t tell you which level of judgement you’re getting, which is why I’d write the description to the level rather than to the task list.

The pillar everyone found hardest

Partnership, the line between what an EA closes on their own and what they take to the executive, was the lowest-scoring pillar for the Level 3s and the Level 2s alike, at just under 2.0. Both ends lose marks. Escalating everything reads as a pair of hands. Deciding things that were never yours reads as a risk. The strong middle, acting within sensible limits and saying so in writing, is rarer than I expected even among the best people in the cohort, and in my experience it’s the thing most executives never explicitly agree with their EA. If you do one thing after reading this, agree that line with yours.

What the reflection questions showed

At the end of the day every candidate answers four short reflection questions, scored separately from their level. The highest-scoring of the four was AI readiness, at 2.4 out of 3, which surprised some people in the cohort more than it surprised me. The EAs sitting this assessment are already using AI carefully and know where it doesn’t belong. The lowest was commercial judgement, at 2.0, which is also the pillar that climbs most steeply from Level 2 to Level 3. That’s the development edge for most of the people in this group, and it can be developed.

What I’d take from it

If you’re hiring. Test the ordinary morning, not just the crisis, because that’s where the levels separate. Ask about the line between deciding and escalating, because nobody finds it easy. Don’t be reassured by a candidate who executes well, because every level in this cohort did.

If you’re sitting it. The level comes from how many scenarios you handle at the top mark, so don’t save yourself for the dramatic ones. Read every email twice in the inbox. Say in writing what you decided yourself and why. If you miss Level 3 by a few points, the report will name the answers that decided it, and the strongest result stands when you sit again.

I’ll update this as the cohort grows. At fourteen the pattern is clear enough to act on and small enough to be wrong in the detail, and I’d rather say so than pretend otherwise.

Common questions

---