Will AI replace administrative assistants? The field has shrunk sharply over two decades, with medical roles the lone growth pocket.
Administrative assistants are still needed, but the numbers already show the routine core of the job shrinking, the field shrank sharply between 2004 and 2024, with the sole growth pocket in medical-office roles tied to healthcare's own expansion. Drafting and aggregation will keep getting faster as Copilot-style agents spread through more companies, but priority-based planning, person-by-person communication, mistake prevention, and proactive support for the overall flow will remain.
Administrative assistants do much more than create documents or answer phones. They keep work from stalling by handling meeting preparation, calendar coordination, stakeholder communication, paperwork, information gathering, and the prevention of missed steps for managers and departments. Much of the work happens behind the scenes, but it plays a core coordinating role that supports how an organization functions.
The value of this role lies less in simple data entry and more in anticipating what needs to happen next and putting the right steps in place ahead of time. AI can speed up document drafts and routine messages, but deciding who needs to see what, when it needs to move, and how to handle it without mistakes or awkwardness still tends to remain a human responsibility.
Calculated by ai-job-risk.net from four public datasets, on a 0–100 scale relative to the 188 occupations we cover. Higher means more of the job's tasks are ones generative AI can do or is already used for. It is not a probability of losing the job.
Task analysis by researchers (Eloundou et al.; ILO). Shown as a percentile among all occupations they rate.
Observed use of Microsoft Copilot and Anthropic Claude on this occupation's tasks. Shown as a percentile among all occupations they rate.
Productivity change - share decreases as technology, including artificial intelligence (AI) systems, assists workers with routine back office functions, such as preparing documents, scheduling, and notetaking.
US figures for the BLS occupation "Secretaries and administrative assistants, except legal, medical, and executive". BLS projections assume technology keeps developing at its historical pace; they are not a forecast of AI's effect alone.
Sources
Exposure index calculated by ai-job-risk.net from the datasets below (data as of October 15, 2026). US labour figures are from the BLS, unchanged.
The publishers did not prepare, review or endorse this site. Full citations, licences and method
AI will not eliminate administrative assistant work, but the numbers already show real contraction. The number of secretaries and administrative assistants in the U.S. fell sharply between 2004 and 2024, and BLS's lead economist for employment projections has said plainly that this occupation family has faced productivity-enhancing technologies limiting demand for employment for several projection cycles in a row, with generative AI now the newest technology in that pattern. The one clear exception is medical secretaries and administrative assistants, whose numbers are projected to keep growing because of overall healthcare-industry growth.
If you look only at the surface-level tasks, drafting documents, proposing time slots, summarizing meeting notes, or formatting request forms, the work can look highly compatible with automation, and enterprise adoption backs that up: Microsoft's 2026 Work Trend Index reported roughly 15-fold year-over-year growth in active agents across Microsoft 365, and a 2025 PwC survey of senior executives found 79 percent saying AI agents are already being adopted in their companies.
But the real difficulty in the job was never producing paperwork or writing messages. It lies in judging who needs to be prioritized, which requests are urgent, how wording should change depending on the recipient, and which internal route a request should follow. Administrative assistants are not simply people who clear away miscellaneous tasks. They are coordinators who keep a department moving and connect the people around them so nothing falls through the cracks. The useful line to draw is between the work AI is likely to automate and the parts people will continue to own.
AI is most likely to take over office tasks that follow fixed formats and involve few exceptional judgments. BLS's own data on this occupation shows that pattern already playing out at scale.
Routine communications such as meeting invitations, internal requests, submission reminders, and visitor notices can be drafted almost instantly with AI, and adoption data suggests most large organizations already have some version of this running. People still need to check whether the tone fits the relationship and the situation, which a template doesn't know.
AI is well suited to producing first drafts that organize meeting notes by topic or pull out action items, lightening the initial information-sorting work considerably. People still need to decide which decisions truly matter and who is actually responsible for them.
It's relatively easy to use AI to align the format and wording of approval requests, application forms, and meeting materials, work that is mainly about making documents look consistent is especially easy to automate. Even so, someone still needs to verify whether the required attachments and approval paths are actually correct.
AI-powered scheduling tools are good at listing candidate time slots, turning responses into a single overview, and organizing attendance information, clearly reducing the burden of aggregation. Deciding how to secure time from the highest-priority people is still a human task.
The parts of administrative assistant work that remain are the ones that require reading the situation, rebuilding the sequence of work, and adjusting the approach for each person involved, the layer BLS's own healthcare exception hints at, since even the growing corner of this field depends on judgment calls a scheduling tool can't make.
When requests pile up at once, someone still has to decide what should move first, what can be pushed back, and who needs to be consulted before anyone else. Two tasks can look identical on the surface while carrying very different real business importance, and people who can arrange the order so work doesn't stall remain highly valuable.
Executives, frontline staff, and outside business partners each need different levels of detail and different wording. Communicating in a way that is respectful while still covering what must be said remains human work; what matters is moving things forward without damaging relationships, not polished wording alone.
Work such as spotting missing attachments, approval-order mistakes, schedule conflicts, or missing meeting-preparation steps remains important. Even when a process appears to follow the rules, exceptions happen constantly in real operations, and people who notice that something feels off are trusted more.
Finding missing materials before a meeting, nudging a delayed approver, or reordering tasks based on travel schedules and time off are all forms of anticipatory support that remain human work. The strongest assistants do more than complete one-off tasks, they watch the whole flow and keep it moving.
For administrative assistants, the future depends less on typing speed and more on the accuracy of coordination and information handling. AI should clear routine work while people keep building the planning and sequencing ability only humans provide.
People who understand who checks what, where approvals tend to get stuck, and what each document requires are much stronger in this role. Having a mental map of internal procedure becomes a real advantage precisely because it's the thing current AI agents still can't infer on their own.
It's not enough to line up a few candidate dates. Strong administrative assistants also weigh the priorities of key participants, deadlines, and the backward planning needed to prepare for meetings. The job is about arranging work so it actually moves forward, not just filling a calendar.
The role requires the ability to prevent small but damaging mistakes such as misdirected emails, missing attachments, skipped approvals, or incorrect names. AI can draft a message, but the quality of the final human check still varies greatly from person to person, and that carefulness matters even more in communications that leave the company.
It's increasingly important to use AI for drafts and summaries while designing clear checkpoints for human review. Automation isn't helpful if it increases misdirected messages or misunderstandings. The people who stay valuable are the ones who reduce manual work without sacrificing quality control.
Administrative assistant experience goes beyond support work. It builds strengths in scheduling, information handling, error prevention, and internal coordination, strengths that transfer well even as the routine core of the role keeps shrinking.
People who have built strength in calendar coordination and anticipatory planning can move naturally into roles centered on schedule management, where balancing the priorities of different stakeholders becomes a direct advantage.
Exposure index 81/100 · US jobs 2025–35 -1.3%
Experience organizing multiple requests and keeping work moving without omissions also supports project-wide coordination, suiting people who want to expand their support perspective into upstream planning and sequencing.
Exposure index 70/100 · US jobs 2025–35 +6.7%
Reading people's situations and coordinating the right communication at the right time transfers well into onboarding and ongoing customer support, extending internal coordination strengths into customer-facing partnership work.
Exposure index 83/100 · US jobs 2025–35 +3.9%
Experience checking documents and organizing inquiries can be extended into broader internal office administration, fitting people who want to move into roles that stabilize the processing flow as a whole.
Exposure index 89/100 · US jobs 2025–35 -6.0%
The coordination skills built through applicant communication, internal notices, and schedule setting also translate into HR operations, applying careful interpersonal communication to recruiting and labor administration.
Exposure index 85/100 · US jobs 2025–35 +6.4%
Experience preventing things from falling through the cracks while watching how a department operates can also lead into process analysis and improvement work, moving from daily operational support into roles that improve the workflow itself.
Exposure index 88/100 · US jobs 2025–35 +10.1%
Administrative assistants are still needed, but the numbers already show the routine core of the job shrinking, the field shrank sharply between 2004 and 2024, with the sole growth pocket in medical-office roles tied to healthcare's own expansion. Drafting and aggregation will keep getting faster as Copilot-style agents spread through more companies, but priority-based planning, person-by-person communication, mistake prevention, and proactive support for the overall flow will remain. As the work changes, career prospects will depend less on how many tasks someone can process and more on how smoothly they can keep people and work moving.
These roles appear in the same industry as Administrative Assistant. They are not the exact same job, but they make it easier to compare AI exposure and career proximity.
Administrative Assistant scores 91 out of 100 on our generative-AI exposure index (rank 26 of 188). That measures how much of the work AI can do or is used for, not whether the job will disappear. The BLS projects US employment of Administrative Assistant to change by -6.0% from 2025 to 2035.
AI is most likely to take over office tasks that follow fixed formats and involve few exceptional judgments. BLS's own data on this occupation shows that pattern already playing out at scale. Examples: Drafting routine emails and notices; Summarizing minutes and notes; Formatting applications and documents; Proposing candidate dates and consolidating schedule information.
The parts of administrative assistant work that remain are the ones that require reading the situation, rebuilding the sequence of work, and adjusting the approach for each person involved, the layer BLS's own healthcare exception hints at, since even the growing corner of this field depends on judgment calls a scheduling tool can't make. Examples: Reworking plans based on priorities; Tailored communication and relationship management; Catching omissions and exceptional conditions; Supporting department-wide progress and acting ahead of problems.
For administrative assistants, the future depends less on typing speed and more on the accuracy of coordination and information handling. AI should clear routine work while people keep building the planning and sequencing ability only humans provide. Examples: Understanding workflows and approval paths; Precision in scheduling and progress management; Document checking and risk sensitivity; Designing AI-assisted office efficiency.
