Earned automation means learning a skill well before you hand it to AI. It works like a calculator at school: your teen learns the math first, then uses the tool. It's also how people get the best results from AI at work. Kairo is an AI tutor for teens built on this idea.
Maya is a made-up teen practicing equations. Here she is in Kairo twice, two weeks apart.
How do I solve 3x + 5 = 20?
What have you tried so far?
I don't know where to start.
Let's get x by itself. What's being added to 3x?
5. So I take away 5 and get 3x = 15. Then x = 5.
That's it! Why did you take away 5 first?
To undo the + 5, so only 3x is left.
Kairo gave hints. Maya did the work and explained why.
Next practice problem: 4x − 7 = 21.
I've got these now. Can you do this one so I can try word problems?
Sure, you've learned this one. Add 7 to both sides to get 4x = 28. Then divide both sides by 4. x = 7.
Ready for word problems? Let's read the first one. What is it asking you to find?
Maya proved the skill over many problems, she asked first, and her mom had turned on full help. Word problems are new, so it's back to hints.
On her homework, Maya still gets hints, even for a skill she's learned.
AI is part of more and more jobs. It works best for people who know the work and use AI to do it faster. If you don't know the work, AI's mistakes sound just as sure as its right answers, and you can't tell them apart.
In a study of 758 consultants, AI helped them do 12% more work, 25% faster, at higher quality. On a problem the AI got wrong, the ones using it were 19 points less likely to find the right answer than the ones working alone.
Read the study (2023)In a survey of 319 office workers, the ones sure of their own skills thought harder about AI's answers. The ones who trusted AI more thought about them less.
Read the study (2025)Kairo builds that habit now. Your teen learns a skill, proves it, and only then lets AI help with it. That's how they'll use AI well at work, too.
Every Kairo session is the first two: hints, then explain why. Handing it off and keeping it come later, one skill at a time.
If the skill slips, it goes back to hints
These checks run in Kairo's own code, so they hold on every message.
AI works the same way. If your teen doesn't know a skill, they can't tell when AI gets it wrong. Once they do, AI helps them do more, faster.
In a study of nearly a thousand high school math students, the ones who practiced with a plain chatbot did worse once it was taken away. A tutor built to protect learning mostly avoided that drop.
Read the study (2025)A 2026 study found the best results when the AI tutor was part of a routine where students had to show they'd learned a skill before moving on.
Read the study (2026)Kairo only does a step on practice problems it picked itself. If your teen types, pastes or snaps a photo of a problem, that's their own work. It doesn't matter what they call it. On their own work, Kairo gives hints, not the answer.
Yes. Help starts on Hints only. The first time your teen learns a skill in a subject, your dashboard asks if Kairo may do learned skills. Nothing changes until you say yes.
Hints only: Hints, one step at a time. Your teen explains why. Full: Everything in Hints only, plus two things on Kairo practice problems. The worked answer shows after 3 real tries, and Kairo does a learned step when your teen asks. Homework still gets hints. Off: No help in this subject. A real try is an answer, like 4 or B. "I don't know" doesn't count. New accounts start on Hints only.
About 5 right answers in a row on the first try, explaining why along the way. A wrong answer sets it back, so most teens need more. Answers that needed a lot of hints count for less.
Lock it. Kairo never does a locked skill, even after your teen learns it. Only you can unlock it.
Kairo stops doing it and goes back to hints until it's solid again. Your weekly report shows any skill that slipped.
Yes, on practice questions from Kairo's library. Your teen's own writing stays theirs: Kairo helps with questions and feedback.
On homework and graded work, Kairo gives hints, not answers. You can read every chat word for word. Every school sets its own AI rules, so it's worth checking yours.
Your teen's chats are not used to train AI models. Homework photos go to our AI provider to be read. We don't keep them. We can't remove details inside a photo, so have your teen photograph only the problem. Read our privacy policy.
Rather talk it through? Book a 15-minute walkthrough.
For researchers, writers and professionals
Automation and skill. Lisanne Bainbridge's ironies of automation (1983) explains why the skill matters most when the automation fails.
Related ideas. Earned automation puts the earning on the person. Earned autonomy puts it on an AI agent (Taneja, 2026), and "earn the right to automate" puts it on a process (Edmondson, 2026; Spencer, 2026). It builds on gradual release of responsibility (Pearson and Gallagher, 1983), desirable difficulties (Bjork, 1994) and cognitive offloading (Risko and Gilbert, 2016).
Research. Bastani et al., PNAS, 2025 · Oreopoulos, Liut, Sungu and Low, NBER, 2026 · Lee et al., CHI 2025 · Dell'Acqua et al., HBS working paper, 2023
To cite this page: Kairo. (2026). What is earned automation? learnkairo.com/earned-automation