An AI executive assistant is worth paying for when it owns your calendar, your inbox triage and your follow-through end to end — the week that gets built without you touching the scheduler, the meetings that confirm themselves, the commitments that get chased without you remembering to chase them. It stops being worth paying for where the relationship is the product: the call that only you can take, the negotiation, the judgment call. Buy the seat that protects your time. Do not buy software that drafts replies and leaves you running your own week.
If you have hired an executive assistant — or burned out trying to be your own — you already know the pattern. The first hire takes six months to train and quits. The second one is great at scheduling and drowns the moment you add anything ambiguous. Or you never hired one, and your evenings are the follow-up queue: confirming, rescheduling, chasing, thanking, filing. That is the seat. Here is what to buy, what it should own on its own, what it costs, and where the human line sits.
What an AI executive assistant actually is
Some people call them agents. We call them digital employees. The difference is not the technology — it is where you stand at the end of the week. An agent does a task and hands it back to you. A digital employee holds a job: a role, a goal with a number, a trigger that starts it, and a person it reports to. (The full definition and the four tests are in What Is an AI Employee?, the hub for this subject.)
The job description for the seat looks like any hire you have ever made:
- Role: Executive assistant. Calendar, inbox triage, meeting logistics, follow-through, the briefing packet before every call.
- Goal: your week runs clean — zero double-bookings, zero unconfirmed meetings, zero commitments that fell through because nobody chased them. Not “emails drafted.”
- Backstory: a chief-of-staff-grade assistant who knows your priorities, your no-go times, who gets a yes and who gets a “pass it to me” — the operating identity it acts like, never invented credentials.
- Trigger: the new meeting request, the unanswered thread, the morning it builds your week. It starts itself. You do not open a tab and prompt it.
One seat, one job. Most of what is sold as an “AI executive assistant” today is a reply-drafter with a calendar sidebar: it writes the email, you check it, you send it, you reschedule, you chase. Faster assistance, every step of it still on your desk. If your week only gets managed when somebody opens the software, you have not bought an assistant — you have bought a subscription that still needs one. If you want to see how this seat sits next to the others — bookkeeper, SDR, service, marketing — the seat map is in AI Agent Examples.
What it owns: your week, start to finish
An executive-assistant digital employee should own, without a human in the loop:
- The calendar. Meeting requests answered inside your rules — priority, buffers, no-go times — with the back-and-forth to find the slot done by the employee, not by you.
- The inbox triage. Every inbound sorted: answer it, file it, escalate it, or bin it, against a written standard. You see the twenty that matter, not the two hundred that don’t.
- The follow-through. Sent proposals acknowledged, unanswered threads chased at set intervals, promises you made in calls captured and tracked until they are closed. This one alone is the reason most owners buy the seat: nothing falls through because nobody was chasing it.
- The briefing packet. Before every call: who they are, the history, the last thread, the open question. You walk in prepared because it assembled it, not because you skimmed the thread at 11:58.
- The reporting. On a Friday, one sentence: your week ran clean, here are the three threads it could not close and why.
If the answer on a Friday is “it drafted your replies,” that is activity, not a job.
What to buy: run the four tests on anything you are sold
Vendors will sell you “AI executive assistant” that is a reply-suggester with scheduling links. Run the four tests before you sign anything:
- Does it start itself? If a human has to open it, paste the thread and ask for the reply, it fails. That is a task with you at the front of it.
- Does it finish? If the meeting still needs you to confirm the slot and chase the no-show, it fails. That is a task with you at the back of it.
- Does it have a number? One outcome it is accountable for — zero dropped commitments, zero unconfirmed meetings, inbox at zero by six — that you can look at on a Friday and say yes or no. “It does lots of things” is what a tool does.
- Does it report to someone? When a thread goes unanswered or a meeting cannot be placed, a person hears about it, from the thing itself, with a reason and a proposal. If it fails quietly, it fails.
Most of what is marketed as an AI executive assistant today fails the first test. That does not make it bad software. It makes it a tool, and it should be priced and staffed like one.
Onboard whatever you buy the way you onboard a hire: a written job description, a narrow first task — the calendar alone, nothing else — a probation period where nothing leaves the building without a human yes. Two weeks on the full job with a human checking every outbound message is how you get to walk away without finding out in a month that it double-booked your best client.
The line you do not skip
This seat touches your relationships, so decide what it may send before it sends anything. Instructions are requests; permissions are guardrails — give it the smallest set that lets it do the job, and keep the send key for anything that carries your name into a negotiation. Anything that reads as cold or wrong in writing is a cost you pay in trust, and trust is the thing this seat is meant to buy back, not spend. A vendor who cannot show you the review queue and the audit trail is selling a liability with a calendar attached.
What it costs
The pricing model that holds up is salary-anchored: what you would pay a human for the month, you pay the digital employee for the year. Take the fully loaded monthly cost of the role — salary, taxes, benefits, overhead — and that monthly number is the annual fee for the seat, all in: design, build, deployment, support and a year of running. The minimum for any seat is $3,600 a year. Model and tool bills sit on top at cost, paid on your own accounts — on the recommended open-source frontier models, about two dollars a day. Year two is the same fee.
Compare that to the hire it sits next to and the arithmetic is the point: a year of the digital seat costs one month of the human. A human seat is a bundle of five to seven roles; a digital employee typically takes five of the seven — the scheduling and the chasing and the filing go, the judgment and the relationships stay — so price what is covered and staff the rest. (The hidden costs that never appear on the quote are covered in AI Agent Cost.)
What stays human
- The calls that carry the relationship. Negotiations, investor conversations, the sensitive reassignment, the firing. The employee books it, briefs it and chases the notes afterward. You take it.
- Judgment on priorities. It executes against your rules; when two things genuinely conflict — the client and the family, the board and the launch — a human decides. It should surface the conflict, not resolve it.
- The message that matters. Anything where the wording is the relationship — the apology to a good client, the offer, the no. The employee drafts and flags; you send.
- Your network’s trust. The strongest result we have published from a fleet of this kind is a media buyer who got 96 hours back in a year, with 80% less hands-on per report — as published on our results page. Their result, not a forecast of yours. What it bought was not speed. It was a human with her evenings back, doing the parts of the business only she can do.
That boundary is not a limitation that gets engineered away next quarter. It is the design.
The seat, not the software
If your evenings are a follow-up queue and your best hours go to scheduling, that is an unfilled seat on your org chart, and it is the cheapest administrative capacity you will ever buy. The buyers who regret it bought a reply-drafter and called it an executive assistant. The ones who do not regret it bought a job: a number, a trigger, a reporting line, and a human who takes the calls that matter.
One scoping note: this seat owns your week’s machinery — not the full map of where digital employees go in a business, which is in AI Agents for Business.
Want the seat filled for you? We diagnose the jobs, scope the seat, build the digital employee and hand it over with the reporting line attached. See how deployment works →