Art deco sunburst emblem split into a large gear-filled section and a smaller section with a human hand holding a fountain pen and magnifier.

Quick answer

The 30% rule in AI is a rule of thumb, not a law or an official standard. In business it usually means letting AI handle the bulk of the routine work, such as first drafts, summaries and sorting, while a person keeps the part that needs judgement, checking and accountability. Do not treat 30% as a target to hit. Use it as a reminder: the draft can come from a tool, but the decision still comes from you.

Key takeaways

  • It is a reminder, not a regulation. There is no single official “30% rule”, and people online use the phrase in different ways.
  • The useful idea is the split: AI does the routine drafting; a person decides, checks and stays responsible.
  • Apply it one task at a time by labelling each step as “AI drafts” or “a person owns”.
  • Measure the checking time. If checking takes as long as doing the job yourself, the split is not working for that task.
  • Some steps always stay human: anything sent to customers, prices and promises, and legal, financial, tax or HR decisions.

In our latest video on how to use AI in your small business, I answer a question I see a lot: what is the 30% rule in AI? The short version is that it is a rule of thumb about who does what. This guide turns that into something practical you can apply to the one task you are testing.

📘 Main article: This guide goes deeper on one step from AI for Business: A Practical Starter Guide for Small Firms. Start there if you want the full one-task-first plan, the risks and the checklist.

What is the 30% rule in AI?

It is a loose way of describing how to share work between AI and people. The version most relevant to small businesses goes like this: let the AI tool do roughly the first 70%, the repetitive drafting and sorting, and keep a person on the remaining 30%, the part where judgement, checking and responsibility matter.

The numbers are not the point. Some tasks will need a person for far more than 30% of the effort, especially at first. Others may need only a quick check. The point is that a person stays in charge of the result.

Is the 30% rule a law or an official standard?

No. It is not a law, and there is no single official 30% rule that businesses have to follow. You will see the phrase used in different ways online, and some people quote it as if it were fixed. It is not.

What is real are the rules around the data you put into AI tools. In the UK, personal data is covered by UK GDPR, so check how a tool stores and uses what you type. Our guide on checking whether new AI rules affect your business gives a simple routine to keep up.

🔎 Unconfirmed / rumour: If you see someone claim the 30% rule comes from a specific law, regulator or study, ask for the source. We have not found one official definition, so treat precise claims about it with care.

How do you apply the 30% rule to one task?

Break the task into steps, then label each step by who owns it. This takes about 15 minutes with a task you already do every week.

  1. List the steps as you do them now. You may already have this from step one of the starter plan.
  2. Label each step. Use “AI drafts” for repetitive, text-heavy work where a mistake is easy to spot. Use “person owns” for anything involving judgement, facts that must be right, or sending something out.
  3. Write one check rule for every AI step. For example: “check every price against the price list” or “read it aloud as the customer”.
  4. Name the owner. One person is responsible for the final result, even if the tool wrote most of the words.
  5. Time it for two to four weeks. Record how long each run takes, including the checking and fixing.

💡 Tip: Put your check rules into the prompt itself, for example “mark anything you are unsure of with [CHECK]”. Our guide on writing a reusable AI prompt shows how.

What does the split look like for a real task?

Here is one way to split a weekly job, writing replies to common customer questions. It is an illustration, not a measured result.

StepWho owns itCheck rule
Sort new emails into common questions and everything elsePersonAnything unusual, angry or about money goes straight to a person
Draft replies to the common questionsAI draftsUses the saved prompt with your real policies pasted in
Check facts, prices, dates and namesPersonCompare against your own notes, never the draft
Adjust tone for the customerPersonRead it as if you were them
SendPersonNothing goes out unread
Note what had to be fixedPersonOne line per reply, used to improve the prompt

In this example the AI does one step out of six, yet that step may be where most of the typing happens. That is the 30% rule working as intended: the tool saves effort on the routine part, and the person keeps every decision.

Which parts should always stay with a person?

Some steps should stay with a person whatever percentage you aim for.

  • Anything that reaches a customer. The tool can draft. You remain responsible.
  • Prices, promises, dates and refunds. These are easy for a tool to get confidently wrong.
  • Legal, financial, tax and HR questions. Treat AI output as a starting point only and get qualified advice before you act.
  • Decisions about personal data. Including whether it should go into the tool at all.
  • Complaints and sensitive conversations. These need a human tone and human judgement.

If your team uses AI, write these down in a short policy. Our guide on writing an AI acceptable use policy for a small team has a template.

How do you know the split is working?

Look at three numbers over two to four weeks: total time per run, time spent checking, and how often you had to make a real correction.

What you seeWhat it suggestsWhat to do
Total time falls and corrections dropThe split suits this taskKeep it, and keep checking
Total time falls but corrections stay highThe prompt needs workTighten the prompt and add missing facts
Checking takes as long as doing it yourselfAI is not saving time hereDrop it and try a different task
Errors slip past the checkThe check rule is too looseMake the person-owned steps stricter

Be realistic. AI will not run your business for you, and I cannot promise how much time you will save. Some jobs get quicker after a couple of weeks. Some barely move because the checking still takes as long. Our guide on measuring the ROI of an AI tool has a simple worksheet, and setting confidence thresholds for human review helps when volumes grow.

What are the common mistakes with the 30% rule?

  • Treating 30% as a target. If a task needs a person for 60% of the effort, that is fine.
  • Skipping the check because the drafts look good. Confident wording is not the same as correct facts. See how to test an AI tool for made-up answers.
  • Leaving out checking time when you judge whether it is paying off.
  • Splitting too many tasks at once. Start with one, learn from it, then add the next.

Watch: the source video

What this video covers: In just over four minutes, I walk through the one-task-first approach: pick one repetitive task, write it down and time it, get your examples ready, choose one general assistant, write a reusable prompt, check every output, and measure for two to four weeks before you decide.

Video: Prodify Digital on YouTube (our own channel), presented by Garry Knight. No sponsor. Tools named are examples only, not paid recommendations.

Frequently asked questions

What is the 30% rule in AI?

A rule of thumb for sharing work between AI and people: AI handles most of the routine drafting, while a person keeps the judgement, checking and accountability. It is not an official standard.

Is the 30% rule a legal requirement?

No. It is not a law. Laws that do apply include data protection rules such as UK GDPR, which cover what personal data you put into AI tools.

Should exactly 30% of the work be done by a person?

No. Do not treat 30% as a target. The share depends on the task, and it is normal for a person to do more at the start.

Which tasks suit the 30% rule best?

Repetitive, text-heavy tasks where mistakes are easy to spot, such as email replies, first drafts of posts, document summaries and answers to common questions.

How do I know if AI is really saving time?

Time each run for two to four weeks, including checking and fixing. If checking takes as long as doing the job yourself, try a different task.

Does the 30% rule mean AI will take 70% of jobs?

Not in the sense used here. For most small businesses, AI takes on parts of tasks rather than whole jobs, and people stay responsible for the result.

Your next step

Take the one task you are testing, list its steps and label each one “AI drafts” or “person owns”. Add a check rule to every AI step, then time the next few runs.

If you have not chosen a task yet, start with AI for Business: A Practical Starter Guide for Small Firms and How to Choose the Next Business Workflow to Automate With AI.

author avatar
Garry Knight
I'm Garry Knight, the person behind Prodify Digital. I write about email list building, email marketing, SEO, AI search and the tools that connect them. My aim is to make online marketing easier to understand, so creators and small business owners can make informed decisions about building an audience and keeping people engaged. Here you'll find straightforward guides and product reviews that explain what something does, where it fits and which limitations matter. The focus is on clear explanations and useful next steps—not hype, shortcuts or promises of easy earnings.
One thought on “What Is the 30% Rule in AI? How to Apply It to One Business Task”

Leave a Reply