AI Is Splitting the EA Profession in Two. Here's Which Side to Land On.

Two years ago, Cortney Hickey blocked off every Friday afternoon for the same two-hour ritual: manually pulling meeting prep for her CEO's upcoming week — CRM records, LinkedIn profiles, recent emails, all assembled by hand. She was good at it. She was also quietly terrified of it. Not because AI might take over the task. Because if that was her value, she was already running out of time.

She's not doing it anymore. A Zapier agent pulls all of it automatically now. Her Fridays go toward things no agent can do — culture initiatives, cross-team alignment, problems that don't have a form to fill out. Her CEO, Wade Foster, said publicly: "Every CEO should have someone like Cortney on their team."

Here's the other scene. The Brookings Institution estimates roughly six million US clerical and administrative workers are both highly exposed to AI displacement and least equipped to adapt. More than 85% are women.

The difference between those two futures isn't luck. It's one decision, made now.

But before you can make that decision, you need an honest picture of what AI is actually doing to this role right now — which tasks are already gone, which are going, and what remains stubbornly, irreducibly human.

What AI Is Already Automating

Start with the most concrete data point in the profession: nearly 30% of EAs name scheduling and calendar management as their single most time-consuming activity, according to a 2026 survey of 600 executive assistants by Vimcal. That's also the category AI tools handle most reliably. Scheduling assistants, inbox drafters, and meeting recorders have reached a level of competence that makes the old manual approach look like carbon paper.

The labor market is already registering the shift. After ChatGPT's public launch, job postings for occupations involving structured, repetitive tasks fell 13%, according to Harvard Business School research. Meeting capture, inbox drafting, and calendar coordination fit that profile precisely. These aren't predictions — they're documented changes in what employers are actually willing to pay someone to do.

Monique Helstrom, former executive assistant to Simon Sinek and now a leadership coach, frames the real danger more precisely than most: "Why AI won't replace executive assistants — but ignorance might."

That's the reframe worth sitting with. The threat isn't the technology. It's the gap between what AI can now do and what some EAs are still presenting as their primary value. If your job description reads like a list of tasks that a well-configured agent could complete overnight, the problem isn't AI. It's how narrowly your role has been defined — or how narrowly you've been defining it yourself.

Why AI won't replace executive assistants — but ignorance might.
— Monique Helstrom, leadership coach and former EA to Simon Sinek

The 30% of time currently consumed by scheduling, inbox management, and meeting prep isn't disappearing — it's transferring. The question worth asking is what fills that transfer.

This task category pressure isn't unique to EAs at tech companies. Office managers, legal administrative coordinators, healthcare practice managers — if your week contains repeating rituals that produce a calendar invite or a formatted document, you're looking at the same automation surface. The tools doing this work are already licensed inside Microsoft 365 and Google Workspace. The automation is already there. The question is who's directing it.

The Human Edge — And the Decision That Creates It

Here's what the same 600-EA survey found when researchers asked about the other side: 86% of respondents believe AI will enhance their role, not replace it. Only 3% foresee full automation. When asked what gives them their edge, EAs pointed overwhelmingly to things no transcript can capture — reading body language, hearing what's unsaid, knowing which executive is about to have a terrible Tuesday before Tuesday arrives.

Those aren't soft skills in the dismissive sense. They're high-precision judgment skills that require years of proximity to develop. AI is genuinely bad at them, and likely to remain bad at them for reasons that aren't purely technical: they depend on trust, relationship history, and contextual awareness that exists nowhere in a database.

This is where Cortney's story becomes a framework rather than an inspiration. When people ask her how to get started with AI, she gives them a question, not a tool recommendation: "What's the thing you like least about your job?" For her, the answer was email. Zapier's internal culture runs on Slack, which meant email felt alien and separate — a constant drain. So she built an agent trained on her writing style that scans her inbox, drafts personal replies based on sender and content, and saves them to her drafts folder. She still reviews. She just no longer generates.

The freed attention went somewhere no agent can follow.

That first question — what do I like least — is the entry point to the transition she made. It requires no budget, no IT ticket, and no prior technical knowledge. It just requires honesty about where your time is going versus where your judgment actually lives.

The distinction Dawn Stallwood, a leadership advisor and former M&A lawyer, identifies as the real value of senior EA work is "space" — the ability to protect an executive's strategic attention from the constant pull of logistics, relationships, and noise. That's not automatable. An agent can schedule the meeting. It cannot notice that two attendees have unresolved tension and quietly adjust the agenda.

You don't need to work at Zapier to ask what you like least about your job. The question works regardless of title, industry, or whether your employer has a formal AI initiative. Answer it honestly, and you have your automation roadmap — the column of your work that's costing you time without requiring your judgment.

Where AI Implementations Go Wrong

There's a reason so many people feel more overwhelmed after their company launches an AI initiative, not less. Dawn Stallwood points to an MIT finding that more than 70% of AI implementation projects fail — not because the technology doesn't work, but because organizations automate broken processes instead of fixing them first. "Rather than actually solving a pure process that's currently human," she notes, "they just put an AI process in instead — and it doesn't actually solve the problem."

This is the organizational chaos many EAs are living through right now. The tool gets deployed. The workflow underneath it was already dysfunctional. Now the dysfunction is faster.

Leaders and boards, please protect the roles of executive assistants and chiefs of staff. Don't think that AI or other automation can entirely replace their effectiveness and stewardship of your space.
— Dawn Stallwood, leadership advisor and former M&A lawyer

BCG research from April 2026 offers a more grounding perspective: over the next two to three years, 50 to 55% of US jobs will be reshaped by AI — not eliminated. The distinction between "reshaped" and "replaced" is doing real work there. Reshaped means the task mix changes; the role persists. EA roles are overwhelmingly in the reshaped category. The job doesn't disappear. The job description does — and that's actually an opportunity if you're the one writing the new one.

The reader who can look at a chaotic AI rollout and name what went wrong — "we automated the symptom, not the problem" — is providing strategic value that no implementation consultant can replicate from the outside. EAs are often the people with the most granular view of where organizational processes actually break. That diagnostic skill is worth making explicit.

If your company's AI rollout feels like it created more work, you're not behind. You may be in the best position to identify what's broken.

The Risk Is Real — And It Falls Disproportionately Here

There's data that this article can't responsibly skip.

The International Labour Organization found in March 2026 that female-dominated occupations are almost twice as likely to be exposed to generative AI as male-dominated ones — 29% versus 16%. The disparity is starker at the highest automation risk tier: 16% of female-dominated occupations fall into the most exposed categories, compared to just 3% of male-dominated ones. Administrative and executive support roles are the epicenter of this exposure. Women are heavily concentrated in clerical roles where tasks are routine and codifiable, which is precisely the profile that AI handles most efficiently.

That's the honest reckoning. Naming it plainly is the precondition for doing something about it.

Cortney's stated fear was never that she'd lose her job to AI. It was something more precise: "The more your value is tied to manual work, the more vulnerable your role becomes." That's the fear worth taking seriously — not replacement, but being caught still doing the automatable work when everyone else has moved on.

The salary data shows the floor and ceiling moving in opposite directions. In the UK, EAs with advanced AI tool proficiency command an 8 to 15% salary premium over those without it, according to Morgan Spencer's 2026 compensation data. In the US, senior EA roles with AI fluency are reaching $100,000 to $180,000 and above, according to C-Suite Assistants. The gap between an AI-literate and AI-passive EA is already visible in compensation. This isn't a projection — it's 2026 salary guide data.

For the reader who is a woman in an administrative role — which describes most of this audience statistically — this isn't background context. It's a direct address. The risk is real, the path forward is specific, and both of those things being true at the same time is the entire argument.

Which Side to Land On

Cortney's fear wasn't losing her job. It was being trapped in it — doing work that no longer required her judgment, while the work that did went somewhere else. That's a fear most EAs recognize immediately.

The EAs who will matter most in two years aren't the ones who learned every AI tool. They're the ones who decided, early, that their value was never the task in the first place. AI fluency matters — but it's downstream of that decision.

This week's exercise: spend 20 minutes listing every repeating task in your role. Mark each one with either "requires my judgment" or "follows a pattern." The second column is your automation roadmap. The first column is your job. Once you can see the difference clearly, you can start moving time from one to the other — no IT ticket required, no new software license needed to begin.

The question isn't whether AI is coming for the tasks. It is. The question is whether you're still defined by them when it arrives.

Recommended Tools & Resources