What Is the 30% Rule for AI? A Plain-English Explanation

You may have heard someone mention “the 30% rule” for AI: in a news story, at work, or from a grandchild talking about school. It sounds official, like something you’re supposed to follow.

It isn’t. There’s no law, standard or single definition behind it. Different people use the same phrase for quite different ideas, and two of the most common versions point in opposite directions.

This guide lays out what the 30% rule usually means, where the number comes from, and a sensible version you can use in everyday life, with a worked example and a short checklist.

What is the 30% rule for AI?

The 30% rule for AI is an informal rule of thumb about keeping people in charge of work that AI helps with; it is not a law, and there’s no single official definition. Most often, it means AI handles the routine bulk of a task while a person keeps roughly 30% for judgment, checking and the personal touch.

We couldn’t find a reliable source that credits the 30% rule to a particular person or organization. The tech news site BGR, in a June 2026 explainer, suggested it may have started in education and called it “less a hard, defined rule, and more a guideline” for workers. One business guide on the topic says plainly: “It’s not a regulation.”

So when you see the phrase, the useful question isn’t “Am I following the rule?” It’s “Which version does this person mean?”

The different meanings of the 30% rule

Here are the main versions you’ll come across, from most to least common.

1. The 70/30 split: AI does the routine work, people keep the rest

This is the version most business articles describe, and it’s sometimes called the 70/30 rule. AI takes on roughly 70% of the repetitive, data-heavy work. People keep the other 30%: quality control, leadership, ethical judgments, and creative or critical thinking, as BGR describes it.

For example, in an office, AI might sort incoming emails and draft routine replies. A person decides what actually gets sent, handles the tricky cases and deals with upset customers.

2. The 30% ceiling: AI does no more than 30% of the work

This version flips the numbers. It says a person should do at least 70% of the thinking and writing, and AI should contribute no more than about 30% (help with spelling, an outline or ideas, for example). You’ll mostly hear it about school essays and writing.

BGR notes that in education, some people take it to mean a piece of work should score below 30% on AI detection tools such as Turnitin. But that’s a popular interpretation, not a Turnitin standard. Turnitin’s own guidance says its AI detection “may not always be accurate” and should not be the sole basis for action against a student. It doesn’t even show a number for scores between 1% and 19%, because false alarms are more common in that range.

Each school, teacher and publisher sets its own AI rules. If a grandchild is worried about “the 30% rule”, the answer is to check their school’s actual policy.

3. “About 30% of work could be automated”

This one isn’t a rule at all. It’s a prediction that often gets mixed up with the others.

In 2023, the McKinsey Global Institute estimated that by 2030, activities accounting for up to 30% of hours currently worked across the US economy could be automated, with generative AI speeding that up. McKinsey called its 29.5% figure the midpoint of a very wide range, from 3.7% to 55.3%.

Two things are worth knowing. It’s about tasks, not whole jobs. And it’s an estimate about the future, not a limit or a target. A similar idea shows up online as a slogan, “automate a third, amplify the rest”: let AI take the most routine third of the work, and spend the time saved on the parts that need people.

4. A budget rule for businesses

The least common version is aimed at companies: spend about 30% of AI investment on data quality, oversight and managing risk. BGR mentions it alongside the others. Unless you run a business, you can safely ignore this one.

Why the meanings don’t agree

Here’s the same phrase, four different ideas:

Notice that the first two are opposites. One says AI does about 70% of the work. The other says AI does at most 30%.

But they agree on the part that matters: a person stays responsible for the result. That’s the useful idea hiding inside all the versions.

Why people are searching for this

“What is the 30% rule for AI?” has become a common Google question, and plenty of business blogs and tech sites have written about it. A few reasons people look it up:

• It sounds official. People worry they’re breaking a rule they’ve never been told about.
• More people are using AI to write. Once you’ve had a chatbot help with a letter, it’s natural to wonder how much help is “too much”.
• School rules come up at the dinner table. Grandchildren and students hear a lot about AI rules and AI checkers.
• Jobs are in the news. Headlines about AI and work often quote the McKinsey-style “30%” figure.

The short answer for everyday life: unless your school, employer, club or publisher has a written AI policy, there’s no 30% rule you have to follow. If they do, their policy is what counts.

A sensible 30% rule for everyday life

Here’s a version that works for older adults and anyone else using AI at home:

Let AI do the first draft or the legwork. You do the final part yourself. The exact percentage doesn’t matter. The order does.

What AI can do well:

• Write a first draft of a letter, email, speech or card
• Summarize a long article or instruction booklet
• Suggest options: recipes, gift ideas, day trips
• Explain a confusing word or form in plain English

What you do:

• Check the facts. Names, dates, prices, phone numbers and opening hours. AI can produce answers that sound believable but are wrong or made up. IBM calls these “hallucinations”.
• Add your own voice. Change words you’d never use, and add the detail only you know.
• Make the decisions. What to buy, where to go, what to send and who to trust.
• Keep health, money and legal decisions human. Use AI to understand a topic and prepare questions, then check with your doctor, pharmacist, bank, financial adviser or lawyer. Our guide to using AI for medical questions safely shows how.

Worked example: a thank-you note

For example, say your neighbor watered your garden for two weeks while you were away. Here’s how the split works in practice.

What AI does: gives you a tidy draft in seconds. Ask for “three different versions” and you’ll have choices.

What you do:

1. Pick the version closest to how you talk.
2. Cut anything too flowery. “I am eternally grateful for your kindness” might become “Thank you so much.”
3. Add the detail only you know: “The roses have never looked better, and I owe you a jar of my plum jam.”
4. Check your neighbor’s name is spelled right.
5. Write it out by hand, or at least sign it yourself.

The AI saved you from the blank page. The part that makes the note mean something, the detail, the voice and the signature, is yours.

A quick checklist before you use what AI wrote

Run through these questions before you send, print or act on anything AI helped with:

• Did I check every name, date, number and price?
• Does it sound like me?
• Did I make the final decision myself?
• Did I keep private details out of the chat?
• If it touches health, money or legal matters, have I checked with a professional?
• Does my school, club, employer or publisher have its own AI rules?

If you can answer yes to the ones that apply, you’re using AI the sensible way, whatever percentage it works out to.

Where to go from here

If you’d like ideas for the “first draft and legwork” part, our guide to what seniors can use AI for has practical uses with ready-made prompts. And if the terms in AI news still feel like alphabet soup, the 7 types of AI, explained in plain English sorts them out.

Our book AI for Seniors, Plainly Explained (coming soon) goes further, with step-by-step help setting up an assistant and asking good questions. In the meantime, our Scam-Stopper Kit (free when you join our email letter) includes 25 easy AI prompts to try.