The executive-assistant role has spent the last decade quietly absorbing every productivity trend the rest of the office talks about. Calendar links replaced phone tag. Slack replaced the runner with a printed memo. AI is the next layer, and unlike most office software it actually changes what an EA spends their time on, not just what tool they use to do it.
Two data points sit on opposite sides of the story. The US Bureau of Labor Statistics projects the broader category of secretaries and administrative assistants to decline 7% from 2024 to 2034, one of the few white-collar occupations with negative projected growth. Inside that decline, executive secretaries and executive administrative assistants are projected to fall faster, hit hardest by routine-task automation. At the same time, the executives those EAs support are increasingly asking for AI-fluent partners, not headcount cuts.
The job isn’t disappearing. It’s reshaping around the work AI can’t yet do.
What’s getting automated, with evidence
The clearest data on what AI is actually being used for comes from the Anthropic Economic Index. For office and administrative tasks, the highest-frequency Claude conversations cluster around: drafting and editing email, summarising long documents, prepping briefing notes from a stack of inputs, and converting unstructured notes into structured records. None of those are surprises. They’re the parts of EA work that have always been “more time than skill.”
Microsoft’s Work Trend Index 2024 survey of 31,000 knowledge workers found that the most common AI use case at work was “writing and editing.” For EAs specifically, the same pattern shows up in Zapier’s 2024 admin survey: scheduling, email triage, and meeting summarisation are the three top tasks where assistants report using AI weekly.
What’s notably missing from those lists: the parts of the role that involve judgement about people. Knowing which meeting to push back on, which executive needs a heads-up before a difficult conversation, which travel route minimises jet lag for a CEO who has to be sharp on landing. The specific knowledge of one principal’s preferences, history, and quirks is exactly the data AI doesn’t have access to and can’t easily be given.
The hard data on EA productivity gains
The widely circulated “saves seven to ten hours a week” figure that floats around AI-for-EA marketing isn’t from a controlled study, and we don’t have a peer-reviewed equivalent for the EA role specifically. What we do have is close-adjacent evidence:
- A Harvard Business School and BCG study on consultants given GPT-4 found a 25% reduction in time-to-completion on knowledge-work tasks within scope of the model, with 40% higher quality scores.
- The GitHub Copilot RCT showed a 55.8% speedup on a benchmark task, but in a different cognitive domain.
- Salesforce’s State of Sales 2025 reports 64% of reps save 1 to 5 hours per week from AI automation of routine tasks. Email drafting and meeting summarisation overlap with EA work directly.
Collectively, these point at a real but variable productivity gain in the 20 to 40% range on the tasks AI is good at, which for an EA still leaves the majority of the week on tasks AI isn’t good at.
The career impact, sourced
Three pieces of data describe where the role is going:
The base rate is shrinking. BLS projects roughly 271,300 fewer admin and secretary jobs in the US by 2034. That decline is concentrated in the lower-tier “general office assistant” roles. Executive assistants supporting senior leaders show more resilience because the work is less standardised.
AI exposure is high but not total. Anthropic’s Economic Index shows only about 4% of occupations use AI for 75%+ of their tasks. Office and admin work is high-exposure but augmentation (52% of Claude conversations) outweighs automation (45%) across the board. EAs who use AI as a force multiplier rather than a replacement are seeing wider scope, not narrower.
The skill premium is real and measured. The WEF Future of Jobs Report 2025 ranks “AI and big data” and “technological literacy” as two of the top five fastest-growing skills employers expect to need. For EAs, this shows up in job postings: the share of executive-assistant listings that mention AI tools by name has roughly tripled since 2023 according to Indeed’s hiring lab data.
The counterintuitive finding
The data point worth pinning to the wall: the hardest parts of an EA’s job are getting harder, not easier, in the AI era. Inboxes are noisier because every counterparty also has AI to draft more email. Calendar requests pile up faster because everyone is more efficient at proposing meetings. The screening, triage, and trust-based filtering an EA does is more valuable when the volume is up and the signal-to-noise is down.
The EAs who treat AI as the assistant’s assistant, not the assistant’s replacement, are the ones building larger remits, not shrinking ones. Many are now running cross-functional projects, owning parts of communications, or taking on chief-of-staff-adjacent work because the routine load that used to fill the day is gone.
The actual job in 2026
Asked what changed, most experienced EAs describe the same shift: the work is more interesting and less repetitive. Less typing meeting notes, more producing the brief that comes out of those notes. Less booking flights, more building decision frameworks for which trips matter. The mechanical parts of the role are leaving. The judgement parts are growing.
The career question is no longer whether AI will replace executive assistants. It’s whether you’ll be the EA whose principal can’t imagine running their week without you, or the one whose remit gets steadily compressed back into the parts of the work AI can’t do well yet. The training matters because that gap widens monthly.