The decision, stated properly
The trigger is always the same: the week is full of things that should not require you, and the obvious response is to get someone to take them. That instinct is sound, and an executive assistant is one of the highest-leverage hires available to a founder.
The complication is that the pile on your desk is not homogeneous. Some of it is genuine work that needs a capable person. Some of it exists purely because information does not move on its own in your business — and that portion does not have a labour cost, it has a design cause.
Handing the second kind to a person does remove it from your desk, which is a real gain and should not be dismissed. But it converts a design problem into a permanent salary, and it does it invisibly, because nobody ever audits what an assistant spends their day on once the founder stops feeling the pressure.
What an executive assistant does well
A good EA does things no system will ever do. They handle the judgment-heavy edge of your week: the reply that needs to be diplomatic, the scheduling negotiation with a difficult counterparty, the decision about what to interrupt you for. That is skilled work and it is genuinely valuable.
They also provide something structurally unavailable from software, which is elastic coverage. A system does what it was built to do; an EA handles the thing nobody anticipated, which in a founder-led business is a large fraction of everything.
And they change how a founder's attention works. Having somebody who holds the state of your week — what is outstanding, who is waiting, what you promised — removes a cognitive load that no tool replicates, because tools require you to go and look and a person tells you.
For founders whose week is genuinely dominated by human-judgment tasks, this is one of the best hires available and it should not be deferred in favour of building systems. That founder exists and may well be reading this.
Where it structurally breaks
The failure is when an EA becomes the integration layer between systems that do not talk. Copying information from one tool to another, chasing people for updates the system should surface, assembling reports by hand, moving files between places. Every one of those tasks is a design gap being paid for at an hourly rate, forever.
That arrangement also has a bus factor of one. The EA who has been manually holding your business together for two years is carrying an enormous amount of undocumented context, and their departure is a genuine operational event rather than a staffing inconvenience.
It gets more expensive as the business grows, which is the part founders do not model. Manual integration scales linearly with volume — twice the clients means twice the copying — so the relief is temporary and the request for a second assistant arrives about eighteen months later.
And there is a quieter cost. Work absorbed by a person stops being visible as a problem. The status updates still get written by hand, the reports are still assembled manually, but nobody feels it anymore, so it never gets fixed. The pain that would have driven the design change has been purchased away.
There is a version of this that founders find genuinely uncomfortable to look at. Some of the work handed to an assistant is work the founder created by being unavailable — the chasing, the re-explaining, the reconstruction of context for people who were waiting on an answer. Adding a person to manage the consequences of a bottleneck leaves the bottleneck exactly where it was, and makes it cheaper to tolerate.
What we build instead
We take the second pile and remove it. In practice that means the things a founder does because information does not move: status that updates itself from the work rather than from someone asking, so nobody writes the update. Handoffs that route when a stage changes rather than when a person remembers. Reminders that fire from a date in the system rather than from a note in someone's head.
Then the reporting layer, which is where an enormous amount of assistant time quietly goes. If assembling a weekly view means opening four tools and a spreadsheet, that is a build, and it is a small one. The output is a number that exists without anybody producing it.
Then the intake surfaces that generate most manual coordination: forms that create the record, bookings that create the task, files that land where they belong. Individually trivial, collectively most of a day a week in a lot of businesses.
The honest framing is that this does not replace an EA and is not intended to. It changes what the EA is for. A business that has removed the mechanical pile and then hires an assistant gets a person doing judgment work at the top of their capability, rather than an expensive router. That is a much better job and a much better hire, and it is the same salary.
How to tell which one you need
Sort a real week. Log what you actually did in thirty-minute blocks for five days, then mark each block as needs-a-human or exists-because-nothing-else-does. Do not estimate this from memory; the estimate is wrong in a consistent direction, because mechanical work is forgettable and judgment work is memorable.
If the human pile dominates, hire the EA. You have a load problem, not a design problem, and a good assistant will give you back more than a systems project would. Do not let a preference for building things stop you from making the obvious hire — a founder whose week is genuinely full of relationship and judgment work does not need better plumbing.
If the mechanical pile dominates, building removes the work rather than relocating it, and it does so once rather than annually. The comparison is not a fair fight: a fixed build against a recurring salary, for work that will otherwise grow with your client count.
If it is close, do both in a specific order — build first, hire second, six months apart. 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.
An EA is the best possible answer to work that needs a human and the most expensive answer to work that should not exist. Log a real week in thirty-minute blocks before deciding which you have.