What AI Can't Do for Executive Assistants: Context, Taste, and Judgment

You've probably sensed it, even if nobody has said it out loud. Someone in leadership runs a quiet test: what if AI handled the scheduling, the first-draft emails, the meeting notes, the research? The numbers come back cheap. Embarrassingly cheap. And somewhere in a spreadsheet, your role ends up in a column next to a monthly subscription fee.

I'm not saying this to scare you. I'm saying it because pretending it isn't happening is the thing that actually gets you. The better question is what AI can't do for executive assistants, and how to build your work around that answer.

How much of an executive assistant's job can AI do?

A large share of the task list, and that's worth saying plainly. First-draft emails, scheduling across time zones, meeting notes, vendor research, travel itineraries, and agenda prep can all be started with a prompt. Goldman Sachs estimated in 2023 that office and administrative support jobs have the highest share of automatable tasks in the US, at 46%.

The squeeze shows up in hiring, not headlines. Revelio Labs found that US entry-level job postings have fallen about 35% since early 2023. And a study of US résumé and job posting data by Hosseini and Lichtinger found that junior employment at companies adopting generative AI fell roughly 9% relative to non-adopters within six quarters, driven mainly by slower hiring rather than layoffs.

Nobody gets walked out. The role just doesn't get refilled.

Sit with that for a second. Then look at what's left over, because that's where your value lives.

What can't AI do for executive assistants?

AI can't access the context you've built by paying attention to one person over time. It has read about professional communication in general. It has never worked for your exec.

It doesn't know that your exec goes quiet in board meetings when they've skipped lunch. It doesn't know the investor on Friday's call had a tense exchange with your CEO months ago, so the follow-up has to land a very particular way. It doesn't know that "I'll look into it" from your exec means they've forgotten and need a nudge by Wednesday morning. It doesn't know which family member goes by a nickname that exists nowhere online.

Sound familiar? That isn't trivia. It's years of attention, and it compounds. The longer you're in the role, the more of it you hold, and the harder it is to replace.

Why taste is the part that stays expensive

Taste is knowing what good looks like for this exec, in this company, with these people. It's knowing which subject lines get opened and which get skipped. How your exec sounds with investors versus the team versus a difficult vendor. What "keep it brief" means to them, because brief for one leader is three sentences and for another it's three paragraphs.

AI can produce a polished draft. It can't tell you whether the draft is right for the person reading it. That call is judgment: asking the right question, catching the wrong answer, and noticing when something sounds right but isn't. As AI takes more of the routine, that judgment becomes the scarce part of the job.

Your expertise isn't the work you produce. It's the context you carry.

How to turn your context into files AI can use

Here's the trap. Most of us try to protect our hours, keeping AI away from the work so the work still looks like ours. The better move is to protect your context, and then put it to work.

If that context lives only in your head, it can't travel, it can't scale, and it can't make AI useful. So write it down. A starter set looks like this:

1. An about-me file. Your role, your standards, how you like to work, and what you never let leave the drafts folder.
2. A how-we-work file. Your exec's communication style, priorities, preferences, and pet peeves. Tone, sign-offs, phrases they'd never use.
3. Examples you love. Five to ten emails or briefs you'd be proud to send. These teach tone faster than any instruction.
4. Examples you'd never send. The generic, the too-formal, the off-brand. Showing what bad looks like is just as useful.
5. A relationship map. Key stakeholders, what they care about, and how your exec handles each one.

Feed the relevant files to your AI tool before a task, and the output stops being generic. It gets shaped by your standards and your read on the situation. It won't be perfect. It will be a draft you can work with instead of one you have to undo.

That's the idea behind the 25+ custom Claude skills I use daily as EA to the founder and CEO of an Inc. 5000 SaaS company. Each one is mostly context: who's involved, what good looks like, what to avoid. The prompt is the small part.

Done well, these files make you faster and make your judgment visible. Done badly, they turn into a dumping ground of PDFs nobody updates, or they put sensitive details somewhere they shouldn't be. Keep them current, keep them lean, follow your company's AI and data policies, and keep access, money, and people decisions out of AI entirely.

The 20% that was always yours

You're not racing AI. You're moving to a different lane. AI gets faster at producing. You get better at knowing what should be produced, what shouldn't, and what someone needed before they knew how to ask.

You've done it a hundred times. Your exec walks into a call looking off, so you quietly move the next thing on the calendar. You draft the email but hold it. You loop in the right person before anyone asks. You know today isn't the day for the difficult conversation, you know why, and you handle it without making it a thing.

That isn't administrative work. It's relational. You're the human layer between a very busy person and everyone who needs something from them. You manage energy, not just time.

The assistants who stay ahead use AI to clear the routine faster, with less drain, so more of them is left for the work that needs a person: reading rooms, making judgment calls, and making people feel that someone who knows them is on it.

FAQ

What can't AI do for executive assistants?

AI can't hold the personal context of working with one executive: their preferences, relationships, moods, and unwritten rules. It also can't exercise taste or judgment about what's right for a specific person in a specific moment.

How much of an executive assistant's work can AI automate?

A lot of the task list. Goldman Sachs estimated in 2023 that 46% of tasks in US office and administrative support jobs could be automated. The judgment and relationship work around those tasks remains human.

What is a context file for AI?

A context file is a document you give an AI tool before a task, describing who you are, who you support, how they communicate, and examples of good and bad work. It turns generic output into drafts that sound like your exec.

How do executive assistants stay valuable as AI improves?

By using AI for drafting and routine work, then investing the time saved in judgment, relationships, and anticipating needs. Writing your knowledge into reusable context files makes that value visible and portable.

AI can clear the work. Only you can make someone feel looked after.

Clarisse