Using AI for scheduling, bookkeeping help, and research

Writing replies gets most of the attention, but AI is also handy for the dull office work. It's good at turning a mess into a draft. It's not good at being the final word on anything with real consequences.

Here's how that plays out for three jobs.

Scheduling

Building the weekly shift schedule is a good example of a task where AI helps with the first pass and needs a careful second one.

What works: paste in who's available when, the shifts you need covered, and your rules. "Maria can't close. No one works more than five days. Two people minimum from 11 to 2." Ask for a draft grid. You'll often get something usable in a minute, and it beats starting from a blank page.

What doesn't: it doesn't know that two of your cooks don't get along, that Jamal always needs Thursdays for his class, or that one of your rules is unwritten. And it will break your rules now and then without saying so. Check the finished grid against your rules line by line, the way you'd check a new hire's first work.

Treat it as a draft, and you stay the scheduler.

Bookkeeping help

This is the one where I'd be most careful. AI can be useful around your books. It shouldn't be your books.

Good uses:

Explaining something. "What is an accrual, in plain words?" or "What does this line on my statement mean?"
Drafting a message to your bookkeeper or accountant with your questions in order
Suggesting a category for a list of expenses, which you or your bookkeeper then review
Making a checklist for month-end

Bad uses:

Doing the math you'll file with. It makes arithmetic mistakes and won't tell you.
Deciding what's deductible, how to classify a worker, or what you owe. Those are your accountant's calls.
Anything where it sees raw account numbers, tax IDs, or bank logins

If your accountant uses a specific tool, ask what they'd like you to send. They'd rather answer one question now than fix a mess in March.

Research

AI is a good way to get oriented on a topic you know nothing about. "I'm choosing a point-of-sale system for a food truck. What should I ask vendors?" will give you a decent list of questions in seconds. Comparing options, drafting an interview guide, summarizing a long document you'll then read yourself: all fine.

Where it goes wrong is facts. It may state a rule, a price, or a local regulation with total confidence and be wrong or out of date. For anything you'll act on, find the real source: the vendor's own page, the city's website, the actual contract. Use the AI to know what to look for, then look.

A useful trick for a long contract: ask it for "the five things I should ask a lawyer about in this." That's safer than asking whether the contract is fine to sign. It's not a lawyer and can't tell you.

What not to paste in

Here's the practical rule: if you wouldn't read it aloud in a coffee shop, think twice before pasting it into a chat box.

Bank logins, card numbers, account numbers, passwords
Social Security numbers and tax IDs
Patient or client health information
Employee reviews, complaints, or anything about discipline
Customer lists with names, phone numbers, and emails

Many tools have settings that control whether what you type can be used to train future versions. Look at the settings before you start, and read what business or paid plans promise about your data. Plans differ and change, so check the tool's own current terms rather than trusting a summary, including this one.

When you need to work with sensitive material, strip it down first. Swap names for "Customer A." Delete account numbers. Keep what the tool needs and leave the rest out.

Start small

Pick one job: the shift schedule, a monthly checklist, a vendor comparison. Try it for three weeks, with checks. If it saves you an hour and doesn't cause trouble, keep it. If it makes you nervous, stop. You don't have to use it for everything.

And keep a person in the loop for anything that touches money, people, or the law. What AI gets wrong, and how to catch it covers how.