Every few months, a new tool launches promising to do what a personal assistant does, only faster and without a salary attached. The list is endless – scheduling meetings, drafting emails, and reorganising a calendar in seconds. And every time one of these tools makes headlines, a familiar reaction follows, relief from finance teams, excitement from software vendors, and a quiet, weary sigh from every assistant who has actually done this job.
That sigh is the important part, because it is not defensiveness so much as recognition of a mistake that keeps repeating itself, dressed up each time as innovation.
Here is what nobody selling “AI assistant” software wants to say out loud: automating the visible tasks of a PA role does not replace the role. It replaces a small, easily observed slice of it, and treats that slice as though it were the whole job. That confusion is not a technology problem, it is a management problem, and it has been costing companies for years, long before AI entered the picture at all.
The job was never really about the calendar
Ask most senior leaders what their assistant does and you will get a list: scheduling, travel, inbox triage, expenses. It is all true, and it is also the smallest, most mechanical part of what a genuinely good assistant provides, since the actual value sits underneath those tasks rather than inside them. This is precisely why it so rarely gets counted.
A great PA reads a room before a meeting starts and adjusts the agenda accordingly, knows which introductions to make and which to avoid, and catches the political undercurrent in a reply that took four minutes to write, noticing it needs rewording before it ever goes out. They hold institutional memory that never appears in a handover document, simply because nobody thought to write it down. None of that shows up on a task list, which is exactly why it gets left off the list of things AI is supposedly ready to replace.
When a company swaps a PA for a scheduling tool and calls it modernisation, what it is actually doing is admitting that it never noticed the second, harder job underneath the first one; it saw the diary management and missed the judgement entirely.
Automation is not the threat, misdiagnosis is
There is nothing wrong with using software to handle repetitive administrative work, and good assistants have been doing exactly that for years, automating the parts of their own job that do not require a human brain so they can focus on the parts that do. Plenty of the best EAs in the world are, in fact, the ones who have most enthusiastically adopted these tools themselves.
The problem is not the technology so much as a leadership team looking at a role, seeing only its most visible and repeatable tasks, and concluding that the whole job can be reduced to a workflow. That is not really a forecast about artificial intelligence at all, it is a confession about how little attention was ever paid to what the assistant was actually managing on the company’s behalf.
Consider what tends to happen in the weeks after a company “replaces” a PA with a tool: meetings that used to run smoothly start double booking, travel gets arranged without accounting for the client dinner that always follows the flight, and sensitive information that used to be filtered through a trusted, discreet human now sits in a system with far less judgement attached to it. None of these failures show up in a task audit at all. Instead, they show up in the slow erosion of an executive’s time, reputation, and relationships, which is precisely the kind of cost that never lands on a balance sheet where anyone is looking for it.
The role was already being underpriced before AI arrived
This is the part that should concern the industry more than the tools themselves, because AI did not create the assumption that assistants are interchangeable with software, it simply gave that assumption a new and more convenient outlet.
Long before chatbots could draft an email, plenty of organisations were already treating the EA function as a cost centre to be trimmed rather than a capability to be developed, with job descriptions that stayed static for a decade, career paths for assistants that remained vague or nonexistent, and pay structures that rewarded task volume over judgement. The arrival of AI did not introduce this thinking, it simply gave it a more convenient justification for continuing.
That is the real story here, and it is a far more useful one than “robots are coming for admin jobs,” because the companies now piloting AI as a PA replacement are not discovering something new about automation, they are revealing something that was already true about how they valued the role in the first place.
What the smart companies are doing instead
The organisations getting this right are not avoiding AI, they are using it precisely where it belongs, on the mechanical layer, while investing more deliberately in the judgement layer that sits above it, giving assistants the tools to offload low value tasks alongside the mandate to spend that reclaimed time on the parts of the job that actually protect an executive’s time, reputation, and decision making.
That is a very different story to the one currently being told in press releases about AI powered scheduling, and it is also, frankly, a better one, because it treats the assistant function as something worth developing rather than something waiting to be phased out.
The tell is in what gets automated first
Look at which parts of the job actually get offered up for automation first, and the pattern says more than any survey could. It is always the calendar, the inbox, the travel booking, and never the judgement calls, the discretion, or the relationship management, because those were never things the company could see clearly enough to even consider automating in the first place. Companies do not reach for the visible layer simply because it is the easiest part to build software for, although it usually is, they reach for it because it is the only layer they ever really understood the assistant to be doing.
That is the real diagnostic value of this moment, more useful than any headline about robots taking admin jobs. Watching which parts of a role a company is willing to hand to a machine tells you, with more honesty than the company would probably offer if asked directly, exactly how much of that role it ever actually saw.
The real test
If a company can hand its scheduling to a tool and nothing important changes, that says less about the tool than it does about how narrowly the role was defined to begin with, and the organisations that will regret the swap are not the ones being dramatic about it, they are the ones who will not notice the cost until it shows up somewhere else, in a missed introduction, a mishandled piece of information, or an executive who is less protected than they used to be.
AI is not the disruption here, it is the mirror, and right now it is reflecting back exactly how little some companies have ever understood about the people who have been managing far more than their diaries all along.
If any of this landed, you are exactly who we built The Assistant Room for. It’s a space where assistants stop being underestimated and start being properly understood, properly resourced, and properly valued, by people who have actually done this job and know exactly what it takes. Come and see what we are building. Come and be part of a conversation this industry has needed for a long time. We would love to have you in the room. Join us here.
