I thought an AI assistant was just a smarter chatbot.

After using one for a while, I realized it still cannot handle everything for me. But it has started taking over a whole category of chores: the ones I know I should do, but that make me tired just thinking about them.

These are the three things I hand off most often.

1. A scheduled news scan that filters my information anxiety

A few times a day, at fixed times, I have it round up important news, tech industry developments, and technical articles and community discussions actually worth reading. I skim once in the morning and catch an update in the afternoon, instead of opening a dozen tabs and hunting through headlines myself.

It feels like having an assistant who reads everything first and tells me: these are the few things today that are genuinely worth your attention.

2. Slow travel planning, with prices updated every day

With kids, planning a family trip around the school calendar is rarely hard because information is missing. It is hard because there are too many options: dates, departure airports, airlines, hotels, cash prices, and award redemptions. By the time you finish comparing, you are exhausted before the trip even starts.

So I let the AI compare itineraries and destinations, and track flights, hotels, and award availability daily, consolidating everything into one easy-to-compare report.

The best part: I do not have to define every requirement on day one. Today I might ask it to try a different airport; tomorrow, to compare two more hotels. I can hand it new constraints whenever they occur to me. Work that used to take one long, painful sitting now accumulates a little every day into a complete report.

3. Finding the savings opportunities that are easy to ignore

Comparing insurance, asking for refunds, renegotiating rates, or finding a better deal is never actually difficult. What makes it painful is opening a dozen websites, lining up the conditions, comparing back and forth, and going rounds with customer service.

Letting AI handle the research and the tedious comparisons — while I keep the final judgment — saves a real amount of time. And the most convincing moment so far was when I let it make the call itself.

A real case: two rounds of negotiation, from $480 to $1,320

I often receive AT&T new-customer offers, usually a $150–$200 Visa gift card. Two years ago, I tried switching our home internet from my name to my wife’s name to qualify, but eligibility kept getting rejected — the two accounts had no gap between them, so we did not count as a new customer. I went through several agents and a long back-and-forth, and it failed anyway.

This time, I handed the whole thing to the AI.

In round one, the first two attempts failed too. The difference was that I never had to sit on hold, listen to the waiting music, or explain the same story again. After each failure, all I said was: try again tomorrow with a different agent. On the third attempt, it worked — a $20 monthly discount for 24 months, saving $480.

A few days later, I saw someone else land an even better deal, so I had the AI call again the moment customer service opened. Round two came back with:

Against the 1 Gig list price of $90 a month, that round saved $1,320 in total. (The $1,320 is the second deal’s saving against list price — it replaces, rather than adds to, the first round’s $480.)

The most useful thing about an AI agent right now: patience

I have come to think the most practical thing about AI agents today is not answering hard questions. It is handling the chores I know I should do but really do not want to spend my own time on.

It does not always succeed on the first try. But it has the patience to keep trying — and simply not having to wrestle with customer service yourself already makes it feel like a real personal assistant.

It is not an all-powerful assistant yet. Some websites block automated lookups, and some people hang up the moment they hear an AI on the line. But handing the time-consuming searching, comparing, and retrying to AI — while keeping accounts, points, payments, and final decisions in your own hands — is a division of labor that already works today.

What chores have you handed off to AI lately? And has there been a moment when it genuinely felt like a personal assistant to you?

Merged and translated from three posts originally published in Chinese on Facebook: Notes on using the Muse AI assistant (2026-09-16) and AI called customer service and won back $480 (2026-09-19), and Saving money with Muse, part 2 (2026-09-24).