Every six months, someone publishes a piece saying AI is about to replace virtual assistants. Every six months, the piece is wrong in the same way. The interesting question is not whether AI will eat the role. It is which 30% of the role moves to AI, and what the other 70% becomes when the boring 30% is gone.
What moves to AI in 2026
First-pass email triage. Calendar conflict detection. Receipt categorization. CRM data hygiene. The first draft of a status update. Transcription. Translation. Meeting prep doc assembly. Boilerplate research where the source is well-indexed.
This is real and it is happening on every account we run. We expect a vetted assistant to be fluent in three to five AI tools by the time they start. Not because we trained them. Because they were already using these tools in their previous role.
What does not move to AI
The judgment calls. The relationship work. The “this email cannot go out, the tone is wrong” call. The decision to escalate an investor concern to the founder before the next standup. The instinct to flag that a vendor’s invoice is up 14% from last quarter and is that normal? The capacity to chase a missing piece of information through three Slack threads without bothering the client.
The other 70% is, in fact, getting more valuable, not less. The boring work going to AI does not shrink the role. It refocuses the role on the work that actually compounds.
What this means for hiring
If you are evaluating someone to handle work for you, screen for judgment, not for typing speed. Ask them to walk you through a moment they decided not to do the thing the brief said to do, and why. Tools change every six months. Judgment is the through-line.
