AI is already automating the administrative layer of the assistant profession. What it cannot touch is the human layer - judgement, relationships, and proximity to power. This piece explains exactly where the divide falls.
There are two stories being told about AI and the assistant profession right now. The first says every EA will be replaced within five years. The second says nothing meaningful will change. Both are wrong, and the people repeating them either have not read the data or are choosing to ignore it because the truth is more complicated and less comfortable.
The truth is this: AI is not coming for the assistant role. It is coming for a specific version of it. The version built on task execution, on throughput, on being the person who keeps the trains running but could be replaced by anyone else who could also keep the trains running. That version of the role is in serious trouble. But there is another version, one defined by judgement, relationships, and genuine proximity to power, that is about to become significantly more valuable than it has ever been.
This is not opinion. The data is now clear enough to say it with confidence.
McKinsey Global Institute published what may be the most consequential report on the future of work in November 2025. It is worth reading in full. But here are the numbers that should change how every assistant thinks about the next three to five years of their career.¹
57% of all U.S. work hours could be automated with technologies that exist right now. Not “by 2035.” Not “if AI continues to advance.” Right now, with tools that are already built and already deployed. Two years ago, McKinsey’s estimate was 30%. That number almost doubled in twenty-four months.¹
44% of U.S. work hours are nonphysical tasks that could be handled by AI agents. Scheduling. Correspondence. Document preparation. Travel coordination. Data entry. Inbox management. If you looked at many entry-level assistant job descriptions, you would struggle to find a single responsibility that does not fall into this 44%.¹
Now here is the number that changes the narrative entirely:
70%+ of the skills employers value today are used in both automatable and non-automatable work. AI does not make human skills obsolete. It redirects them. People will spend less time preparing documents and doing basic research, and more time framing questions, interpreting results, managing relationships, and making decisions that require human judgement.¹
Read that again. The same report that says 57% of work hours are automatable also says that the vast majority of the skills behind that work will still be needed. They will just be applied differently. This is the nuance that gets lost every time someone writes a headline about robots taking jobs. The work is changing. The skills are not disappearing. But the people who refuse to change how they apply those skills will find themselves competing against people who did.
Let’s stop being polite about it. If your role can be fully described by a task list, if the sum total of what you do could be written on an index card, you are in the part of this profession that AI is compressing. Not because you are bad at your job. Because the tasks that define your job are exactly the kind of structured, repetitive, language-based work that AI was literally designed to do.
40% of enterprise applications will include task-specific AI agents by the end of 2026, up from less than 5% in 2025.² These are not theoretical pilots. They are being built into Outlook, Google Workspace, Slack, and every other tool that assistants use daily. The infrastructure of administrative work is being automated from the inside.
A firm that used to need three assistants to handle scheduling, expenses, and inbox management across a leadership team is discovering that one assistant with the right tools can cover the same ground. This is already happening. We see it in the firms where Blackbook places candidates. The headcount is not always shrinking, but it is consolidating, and the bar for who survives that consolidation is rising fast.
But here is what almost nobody is talking about: at the same time, demand for high-level executive support is growing. Not staying flat. Growing. And the reason is simple. As AI strips away the transactional noise, the principals at the top of these firms can see more clearly than ever what they actually need from the person sitting closest to them. And it is not inbox management.
7x growth in demand for AI fluency in U.S. job postings in two years, the fastest-growing skill in the American labour market.¹ But McKinsey’s own Skill Change Index found that interpersonal skills, negotiation, relationship management, coaching, face the least disruption from AI of any skill category. The skills being automated are data entry and document processing. The skills that are becoming more valuable are the ones that require you to be a human being in a room with other human beings.
This is the split. The bottom of the profession, built on throughput, is being compressed by machines. The top of the profession, built on trust and judgement and relationships, is expanding because machines have made it unmistakably clear that those things cannot be replicated. The middle is where the discomfort lives, because the middle is where most people are, and the middle is being forced to choose a direction.
The assistants earning $200,000 to $300,000 and above are not being paid to manage calendars. They are being paid because their principal cannot afford to operate without them. Because they understand who matters in the room before the meeting starts. Because they know which deal is losing momentum based on how a counterparty’s assistant is responding to emails. Because they can pick up the phone and get information that is not available through any formal channel. Because they represent their principal to investors, board members, and clients in a way that strengthens the relationship rather than just completing the task.
None of that can be automated. Not because AI is not good enough yet. Because these are capabilities that depend on years of accumulated trust, on a personal network that took a career to build, on commercial instinct that comes from genuine engagement with the business, and on the kind of social intelligence that only exists between real people in real relationships.
McKinsey found that C-suite leaders are already the most active AI users inside their organisations, well ahead of individual contributors.³ These principals are not confused about what AI can do. They use it themselves. They know exactly where it helps and where it falls short. And they are more certain than ever that what they need from their assistant is not faster task execution. It is the human layer that sits above it. The reading of a room. The quiet management of a relationship. The instinct for when something is wrong before anyone has said it out loud.
AI has not made the exceptional assistant less necessary. It has made it painfully obvious how much of the profession was operating below that standard, and how much the people at the top were always worth.
If AI can handle 44% of the nonphysical work in your organisation, what exactly are you asking your assistant to do with the other 56%? Because if the answer is “more of the same but faster,” you have not understood the shift.
The firms that are getting this right are not cutting assistant headcount to save money. They are raising their standards for who they hire and what they expect that person to deliver. They want someone who attends certain meetings and comes back with a read on the dynamics, not just the minutes. Someone who flags a relationship that is deteriorating before it becomes a formal problem. Someone who briefs them on a person in a way that draws on context from the assistant’s own network. Someone who manages a project end to end without being walked through every step.
$2.9 trillion in annual U.S. economic value could be generated by AI-powered agents and automation by 2030.¹ McKinsey is clear that this value is not captured by removing people. It is captured by, in their words, “redesigning processes, roles, skills, culture, and metrics so people, agents, and robots create more value together.”
The assistant role is being redesigned whether assistants like it or not. The question is who gets to define what the redesigned version looks like. The firms that invest in senior, high-judgement, relationship-driven executive support will get more from their people than they ever did when those people were buried in admin. The firms that treat this as a cost-cutting exercise will lose their best people to the ones who understood what was actually happening.
If you are an assistant reading this and you recognise yourself in the top end of this profession, the people-driven, relationship-heavy, commercially aware end, AI is not your enemy. It is the single biggest tailwind your career has ever had. It is stripping away the noise that used to consume half your day and leaving behind the work that only you can do. Your principal is going to need you more, not less. And the market is going to reward you more, not less. That process is already underway.
If you are an assistant whose role is primarily administrative, this is not the time to pretend the ground is not moving. It is. The data says so. But it is also not the time to panic. It is the time to decide what direction you want to move in, and to start building the skills, the relationships, and the commercial understanding that will put you on the right side of the divide. The window to make that transition is open now. It will not stay open indefinitely.
And if you are a firm that hires assistants, this is the moment to stop thinking about the role as a cost line and start thinking about it as a leverage point. The best assistant you have ever worked with is about to become significantly more effective. The question is whether you are building the role to let them be.