What the role actually does
An executive assistant manages the founder's time, attention and commitments. Calendar and inbox are the visible part; the substantive part is holding the state of your week — what is outstanding, who is waiting, what you promised — and deciding what is worth interrupting you for.
At its best the role is judgment-heavy rather than administrative. Handling the reply that needs to be diplomatic, negotiating scheduling with a difficult counterparty, preparing you for a meeting with the context you would not have gathered. That is skilled work and it is why good EAs are hard to find.
The role also provides something no system does, which is elastic coverage. Software does what it was built for; an assistant handles the thing nobody anticipated, and in a founder-led business that is a large share of every week.
What separates a great EA hire from an expensive one is not the person. It is what proportion of their week goes to judgment work rather than to compensating for information that does not move on its own.
The nature of the role has also shifted with tooling. Scheduling logistics and inbox triage are substantially cheaper to solve than they were, which raises rather than lowers the value of a strong EA — because what remains is disproportionately the judgment work that no tool touches.
When you genuinely should hire one
Hire when your week is genuinely dominated by human-judgment work. If the blocks that are eating you are relationship management, difficult conversations, preparation and decisions about what to prioritize, a good assistant will return more than any systems project would.
Hire when the volume of external coordination is real. Founders with heavy meeting loads, speaking commitments, travel, or a lot of counterparties spend hours a week on negotiation-shaped scheduling that no calendar tool solves, because the hard part is the judgment about priority.
Hire when you need someone to hold state. The cognitive load of tracking every outstanding commitment is substantial and it degrades every other decision you make. A person who holds it and tells you is different in kind from a system you have to go and check.
General market context: executive assistant compensation in the US ranges very widely — broadly from around fifty thousand for a junior role to well over a hundred for an experienced EA supporting a senior executive, and the offshore and part-time market is considerably lower. Loose orientation rather than a benchmark.
Hire when the cost of a dropped commitment is high. Founders who are the single point of contact for clients, partners and their own team drop things not through carelessness but through volume, and each one costs more than the hour it would have taken.
Where the hire fails
It fails when the assistant becomes the integration layer between systems that do not talk. Copying information between tools, chasing people for updates the system should surface, assembling reports by hand. Each of those is a design gap being paid for hourly, forever, and it scales linearly with your client count.
It fails on bus factor. An EA who has manually held the business together for two years is carrying an enormous amount of undocumented context, and their departure is an operational event rather than a staffing one.
It fails invisibly, which is the part that matters most. Work absorbed by a person stops being felt as a problem — the manual reports still get assembled, but nobody notices, so the design change never happens. You have purchased away the pain that would have driven the fix.
And there is an uncomfortable version worth naming: some of the work handed to an assistant exists because the founder is a bottleneck. The chasing, the re-explaining, the reconstructing of context for people who were waiting on an answer. Hiring someone to manage the consequences of a bottleneck leaves the bottleneck in place and makes it cheaper to live with.
It also fails when the founder does not actually delegate. An assistant who is not permitted to answer anything, decline anything, or decide what is worth an interruption is a very expensive scheduling interface, and this is the most common quiet version of the failure.
What has to exist first
An honest log of where your time actually goes. Thirty-minute blocks, five days, marked as judgment or mechanical. This is the entire decision and it takes a week of mild discipline to produce.
Removal of the obvious mechanical work first. Status that updates from the work rather than from someone asking, handoffs that route on a stage change, reminders that fire from a date in the system. Those are small builds and they change what job you are hiring for.
Written decision rights, in the same way every other role on this axis needs them. What the assistant answers without asking, what they decide about your calendar, what they escalate. Without it, you have added a person who asks you things.
And documentation as an expectation from day one — how they do what they do, written as they learn it. This is the cheapest possible insurance against the bus factor and it is nearly impossible to retrofit two years later.
A genuine willingness to be unavailable. The point of the role is that things get handled without you, and a founder who remains reachable through every side channel has ensured that nothing routes through the person they hired to route things.
How to tell which you need
Sort the logged week. If the human pile dominates, hire, and do not let a preference for building things delay it — a founder whose week is genuinely full of judgment work does not need better plumbing, they need a colleague.
If the mechanical pile dominates, building removes the work rather than relocating it, once rather than annually. That comparison is not close: a fixed build against a recurring salary, for work that grows with your client count.
If it is genuinely mixed, which is common, do both in this order — build first, hire six months later. That sequence means you hire against your real needs rather than your current chaos, and it is the difference between an assistant who runs your calendar and one who runs your operations.
One more check. Ask what a new assistant would spend their first month doing. If the honest answer is mostly moving information between places, you have found a design problem with a job title attached, and it will still be there in year three.
It is also worth checking what you would do with the returned hours. Founders who cannot answer that tend to refill the week with the same category of work within two months, which is a real outcome and not a criticism — it just means the constraint was never the hours.
An EA is the best answer to work that needs a human and the worst answer to work that should not exist. Log a real week in blocks — the estimate from memory is wrong in a predictable direction.