Call

Using AI at work: what it is good for and what you should never paste in

First draftPrivateKeep it out
Original drawing: a counter under the ridge with a first draft marked with a check, beside a locked box marked Private and a sign that says Keep it out
01Quick answer5 questions / 5 min read

The answer first

AI assistants like ChatGPT, Claude and Gemini are good for first drafts, rewording, summaries of public documents and spreadsheet formulas, as long as a person checks the result. Never paste in customer details, health or payment information, passwords or login codes, or anything you've promised to keep private, because on personal accounts what you type may be kept, seen by human reviewers or used to train the AI.

  • 72 hoursGemini keeps chats even with Keep Activity off
  • 5 yearsClaude keeps chats if you allow training
  • 30 daysClaude keeps them if you don’t

What is AI actually good for at a small business?

We use AI every day and check every piece of its work by hand. Here's where it earns its keep for most small teams:

  • First drafts and rewording. "Write a polite reply saying we can't come until Thursday," or make a blunt text kinder. Then you fix the details.
  • Summaries of public things. A long city notice, a product manual, a vendor's terms.
  • Spreadsheet help. "How do I add up column C only when column B says Fruita?"
  • Job posts, social posts and FAQs, as a starting point you rewrite in your own voice.

It's not good at being right on its own: it makes things up with confidence, so a person checks anything with a price, a date, a law or a promise before it goes out.

What should staff never paste into an AI tool?

Our rule of thumb: if you wouldn't read it out loud at the counter with a line of customers behind you, don't paste it. Take the detail out first: a reminder about a Tuesday appointment works without the customer's name, phone and address.

  • Customer personal details. Names tied to addresses, birth dates, driver's license or Social Security numbers.
  • Payment information. Card numbers, bank account and routing numbers.
  • Health information. Anything about a patient or client's health, treatment or appointments.
  • Passwords, login codes and keys. Even the six digit code that just hit your phone.
  • Employee records. Pay, reviews, discipline, medical leave.
  • Anything under a confidentiality promise. A customer's contract, a private bid, numbers a partner shared in confidence.

For a covered health practice, patient information falls under HIPAA, and sharing it with an outside service can require a written agreement HHS calls a business associate contract. Don't paste patient details into an AI tool unless your practice has confirmed it is set up and contracted for that. This isn't legal or medical compliance advice: start with HHS's HIPAA pages for professionals and your compliance person or attorney.

CardDrawn in code by Mesa Valley Marketing

Good to use AI for

  • First drafts
  • Rewording
  • Summaries of public documents
  • Spreadsheet formulas
  • Job and social post starters

A person checks it before it goes out.

Never paste in

  • Customer personal details
  • Payment info
  • Health information
  • Passwords and login codes
  • Employee records
  • Anything confidential

Use the company account. Take out names and numbers first.

If you wouldn't say it out loud at the counter, don't paste it.

Why does it matter which plan you're on?

Personal and business accounts follow different rules, and the rules change. The three big vendors, as of the date above:

Google Gemini. Google's Gemini Apps Privacy Hub says human reviewers, including trained reviewers from its service providers, review some of the data it collects, and asks: "Please don't enter confidential information that you wouldn't want a reviewer to see or Google to use to improve our services". Chats are kept for 72 hours even with Keep Activity off, and chats a reviewer has looked at are kept for up to three years, even if you delete your activity.

Anthropic Claude. Anthropic's privacy center says that on its consumer plans (Free, Pro and Max) it uses your chats to improve its models if you choose to allow it, if a chat is flagged for safety review, or if you opt in another way. Its consumer terms update says chats are kept for five years if you allow training and 30 days if you don't, and that these consumer rules don't apply to its business plans (Claude for Work, which includes Team and Enterprise) or its API.

OpenAI ChatGPT. OpenAI's help page on model training says: "When you use our services for individuals, such as ChatGPT and Codex, we may use your content to train our models." You can switch that off in Settings, Data controls, "Improve the model for everyone." By default, the page says, OpenAI doesn't use inputs or outputs from ChatGPT Business, Enterprise, Edu or its API to improve its models.

All three also have a temporary or incognito chat. OpenAI and Anthropic say those chats aren't used to train their models, and Google says its temporary chats aren't used to improve its AI with help from human reviewers. That's a backstop, not permission to paste secrets. A personal account on someone's phone is not the place for customer or company secrets: if your team uses AI for work, pay for a business plan and read its data terms.

How do I set a simple rule for my team?

Put it on one page and tape it by the computer. For a five or ten person shop this covers most of the risk, in ten minutes at a staff meeting:

  1. Use the company account, not your personal one.
  2. Take out names, numbers and anything private before you paste.
  3. Never paste passwords or codes. Ever.
  4. A person checks anything that goes to a customer.
  5. If you're not sure, ask before you paste.

Can AI work with our real data safely?

Yes, by design rather than by pasting. We build tools that let your own AI assistant do specific jobs, like drafting a reply or pulling up tomorrow's schedule, with access you choose and changes you approve. The data stays in your systems; the AI gets a narrow door, not the whole building.

Want to talk through where the line should be for your team? Call us at (970) 314-6088.

FAQ

What should staff never paste into ChatGPT or other AI tools?

Customer personal details, payment information, health information, passwords and login codes, employee records, and anything you've promised to keep confidential.

Is it safe to use AI for work emails?

For drafting and rewording, yes, if you leave out private details and a person checks the message before it's sent.

Does Google review Gemini chats?

Google says human reviewers review some Gemini data, and asks people not to enter confidential information they wouldn't want a reviewer to see.

Does ChatGPT train on what I type?

On personal plans, OpenAI says it may use your content to train its models unless you turn off "Improve the model for everyone" in Data controls. It says it doesn't train on ChatGPT Business, Enterprise, Edu or API data by default.

Do business AI plans have different data rules?

Often, yes. OpenAI says it doesn't train on its business plans by default, and Anthropic says its consumer training and retention rules don't apply to its business plans or API. Check the terms for the plan your team actually uses.

Can a medical or dental office use AI with patient information?

Only if the practice has confirmed the tool is set up and contracted for HIPAA. See HHS's HIPAA pages and ask your compliance person or attorney.

Sources

7 sources
  1. Google, Gemini Apps Privacy Hub
  2. Anthropic Privacy Center, Is my data used for model training? (dated March 16, 2026)
  3. Anthropic, Updates to Consumer Terms and Privacy Policy
  4. OpenAI Help Center, How your data is used to improve model performance
  5. OpenAI Help Center, Data controls in ChatGPT
  6. US HHS, HIPAA for Professionals
  7. US HHS, Business Associate Contracts
Tell us what you want to change.Grand Junction / Fruita / Palisade / Clifton

Want a hand
with any of it?

(970) 314-6088hello@mesavalleymarketing.com

Call or write and tell us where you are stuck. We answer in plain words.

Born of this place.A little local
intelligence.

Where shall we go?

Find your next step, try a job with Mesa, or follow a few hidden trails.

Site guide & interactive demos
Let’s make something better.
1 / Your focus2 / Your details3 / Review & send

What are you thinking?

Choose a starting point
Good things reward a closer look.

A small adventure around the valley

Grand Valley
Passport.

Eight tiny details are hiding in the artwork. One on each page. Collect them all and you’ll earn a free AI widget, custom-designed for your business.

Still out thereHomeSomething is growing in a photograph.
Still out thereServicesThe orchard has a detail of its own.
Still out thereCustom AppsEven the control room has local roots.
Still out thereStudioLook up. Someone took the scenic route.
Still out thereWineriesBetween the rows, a small beginning.
Still out thereHome servicesA familiar line beside the home.
Still out thereHospitalityA room with a view. Look a little higher.
Still out thereProfessional servicesFollow the current past the paperwork.
0 of 8 discoveries
Bonus stampCustom AppsPair your AI on Custom Apps. It brings back a trail word.

Trail word

Got a trail word from your AI?

No bright markers. No treasure boxes. Just a peach, a silhouette, a little line that belongs to the landscape. Touch a detail when it catches your eye.

Saved on this device. No account or tracking.