March 9, 2026 · The Hollis team
What a personal assistant actually does (vs. what AI promises)
Most of what makes a great assistant work is invisible: context, judgment, follow-through. Here's the real job description.
The marketing for AI assistants describes them as a person you can text. They're not. The gap is the things a good human assistant does that don't show up in a feature list.
Remembering context you didn't say twice
A good assistant remembers that your kid is gluten-free, your father-in-law prefers a window seat, your dentist runs late on Tuesdays. Not because you told them last week — because they figured it out across many tasks and kept the note. That memory compounds. Software has to be told every time.
Reading judgment from one sentence
'Send flowers to Linda' is enough for a person who knows Linda is your mother-in-law, the occasion is bereavement, and your usual budget is $80–120. They'll pick the arrangement, draft the card in your voice, and confirm before sending. A bot needs the full spec or it'll send roses for a funeral.
Follow-through over days
Most life-admin tasks aren't single calls — they're sequences. Dispute filed, awaiting response, follow up Tuesday, escalate to a supervisor if not resolved by Friday, document the outcome. A real assistant tracks the thread and brings it home. Software hands it back to you after step one.
Knowing when to push and when to back off
Retention departments give in to firm, friendly pressure. Insurers respond to specific, well-cited appeals. Doctors' offices respond to politeness and persistence. None of these instincts come out of a prompt — they come from doing this work a lot.
Saying no, kindly, on your behalf
Sometimes the work is making a graceful exit — declining a meeting, ending a negotiation, telling a vendor to stop calling. A person can do this in your voice. Software can't.