Use AI to Build Your Focus, Not Just Finish Your Tasks

Every useful invention starts the same way: someone wants life to be a little easier, a little faster, a little better. The calculator, the search engine and the smartphone all arrived with worries attached, and then they quietly became part of everyone's day. Even people who criticise a new technology tend to be using it by the end of the month. AI is following the same path, and I think the sensible question is not whether to use it but how to use it so that it makes us better.

That is a focus question as much as a technology question. Focus is a skill you can train, and the tools you work with shape the training. So here is a practical way to let AI help you without handing it the part of the work that makes you stronger.

Founder's Note

I cannot imagine my working life without AI. My work depends on research, and AI does it for me. While building GFF, I used to research on the internet and consult many sites. It took a long time and caused delays in the project. I was already using AI, but not majorly. Then AI tools kept coming up with updates, and the results of my prompts improved with each one. I later shifted all my research work to AI and concentrated on building a quality tool. As AI brings more advanced tools, life will become much easier for future founders. I see negativity around AI as a regressive way of thinking, and I would rather put our energy into learning to use it well.

Every new tool meets a worry first

It helps to remember that we have been here before. In one of Plato's dialogues, Socrates worries that writing will weaken people's memories because they will stop practising remembering. Centuries later, calculators in classrooms, search engines and spell-checkers each had their turn as the thing that was supposedly going to dull our minds. In every case the tool did change how we work, and in every case people learned where it belonged. Few of us would give up writing, and most of us are glad of a calculator when the sums get long.

What tends to separate a good outcome from a poor one is rarely the tool itself. It is whether people develop sensible habits around it. That is the encouraging part of this moment: habits are something you can design, starting today, with nothing more than a timer and a little intention.

In 2026 a team of researchers from Carnegie Mellon, UC Berkeley, Oxford, MIT and UCLA published a paper on AI assistance and persistence . They ran three experiments in which people worked on fraction problems and reading questions. While an AI assistant was available, people did better. When it was taken away, those same people solved fewer problems and skipped more of them than people who had never had the assistant.

It would be easy to read that as a warning against AI, but the authors do not frame it that way, and I do not think we should either. They tested short sessions, measured people straight after the tool disappeared, and used an assistant that gave complete answers on demand. They say plainly that they cannot speak to how long any effect lasts, and they urge caution about stretching the result to every kind of task. Their own conclusion is a design one: AI should help people finish today's task and also help them grow. They even suggest that a hint-style assistant is worth exploring.

That is the encouraging reading. The way we ask for help matters, and we get to choose it.

Think about the best mentor you have had. They did not take the pen out of your hand. They asked a better question, pointed at the part you had missed, and let you make the final move. Used that way, AI can be an endlessly patient coach available at any hour, which is a remarkable thing to have on a hard day.

The trick is to decide the order of play before you open the chat window. Your own attempt comes first, even if it is rough. The assistant comes second, to sharpen it. That small change keeps the useful struggle that builds focus while still giving you all the speed and clarity AI offers.

A timer is the easiest way to keep the order of play honest. Here is a simple routine you can try on your next hard task.

First block: think on your own. Set one Pomodoro and work on the problem without AI. Write down what you know, what you are unsure about and your best attempt, however messy. If you feel the urge to ask for the answer, that is just activation energy doing its job, and the timer is there to carry you across it.

Break: ask for a hint, not the answer. Now bring in AI with your attempt in hand. Try wording such as "Here is what I tried. Don't solve it. Tell me what I am missing" or "Give me one hint." You are using AI to point the way while you keep the steering wheel.

Second block: apply it yourself. Go back into a fresh session and use what you learned to finish the work in your own words. Finish by asking the assistant to check your result and explain anything you got wrong. You end up with the finished task and a stronger grip on how you got there.

A routine only matters if it survives a busy day, so here are three everyday pictures of how it might look. These are examples to spark ideas rather than rules.

The student. Imagine a student facing a tricky problem set. In the first block she works through the first two questions alone, noting where she gets stuck. During the break she pastes in her working and asks the assistant, "Where does my reasoning go wrong? Don't give me the answer." In the second block she redoes the questions in her own words, then asks for three similar practice problems to test herself. She finishes with the work done and with a clearer sense of what she can now do without help.

The writer. A writer with a half-formed idea spends the first block drafting a rough outline by hand, however ugly. Then he asks AI to point out gaps in the argument and suggest two counterpoints he has not considered. The second block is spent writing in his own voice, with the assistant brought back only for a final tidy-up. The piece still sounds like him, and it is stronger for the pushback.

The developer. Someone debugging a stubborn error gives herself one block to trace the problem and form a theory. If the timer ends and she is still stuck, she shares her theory and the error message with the assistant and asks it to explain what the message means, not to rewrite the code. She then makes the fix herself and writes one line in her notes about what caused it. Next time, she will spot it faster.

In all three pictures the person does something first, uses AI to move past a specific obstacle, and then does the final step personally. The assistant is generous with help, and the human stays in charge of the thinking.

Try it with a free focus timer

GetFocusFlow is a free Pomodoro timer built for exactly this kind of structured, distraction-free session. No sign-up needed to start.

Start a focus session: getfocusflow.info

Prompts that turn AI into a better coach

The wording you use shapes the help you get. If you ask a bare question, an assistant will usually answer it in full, which is wonderful when you want a fast result and less useful when you want to learn. A few small phrases change the role it plays, and they are worth keeping in a note you can paste from.

When you want a nudge, try "Give me one hint and wait for me to try again." When you want your thinking checked, try "Here is my reasoning. Tell me where it is weak and why." When you want to understand something, try "Explain it to me as if I will have to teach it tomorrow, then ask me two questions to see if I got it." When you want a challenge, try "Ask me questions about this topic one at a time and tell me how I did." Each of these keeps you actively involved, and that involvement is where the learning and the focus come from.

You can also set the ground rules once at the start of a session. Something like "In this conversation, act as a coach. Prefer hints and questions over full answers unless I ask for the answer" saves you from repeating yourself, and you can always lift the rule when you simply need the job done.

None of this means holding AI at arm's length. There are plenty of moments where handing it the work is the smart move. It is excellent at clearing the small chores that eat your attention: drafting a routine email, summarising a long document, reformatting notes, or turning a vague to-do into a clear first step. Each of those frees up mental room for the work that deserves your best hours, which is the idea behind protecting a focus block in the first place.

A useful rule of thumb is to let AI take the tasks you would be happy never to do again, and keep the ones that grow you. Learning a new skill, wrestling with a design, writing something that has to sound like you: these are worth your first attempt. If you are curious how this fits with deep work more broadly, our piece on what Cal Newport says about Pomodoro is a good companion.

Good intentions go further with a little preparation. Before you start a try-first block, decide what done looks like in one sentence, so that the assistant, when it arrives, has a clear target to help with. Close the tabs you do not need, and keep the chat window out of sight until the break. Putting the phone out of reach for twenty-five minutes does more than most willpower tricks.

It also helps to treat the break as a real break. Stand up, drink some water and look at something further away than a screen. When you return to the chat with your attempt, you will usually find your question has become sharper, and sometimes the answer has quietly arrived while you were away from the desk. Many people recognise that experience of solving a problem the moment they stop staring at it, and a Pomodoro break gives it room to happen.

Persistence is a bit like a muscle that responds to regular use. Short, repeatable sessions are a gentle way to train it. If you already follow a rhythm such as four to eight Pomodoros a day , you have a ready-made structure for practising: one or two of those blocks each day can be your no-AI blocks, and the rest can be fully assisted.

Notice how it feels when you stay with a problem for ten more minutes and it finally opens up. That small win is worth collecting. Over weeks it teaches you that being stuck is a temporary state, and with AI as your coach on the other side of the break, you rarely have to stay stuck for long.

Our earlier look at AI and focus covers the wider data on how people's working days are changing. The through-line is the same: the tools are getting better, and what we do with them is up to us.

You do not need a dashboard to know whether this approach is helping. A few honest questions at the end of the week will tell you plenty. Did you finish the things that mattered most? Could you explain, without looking, how you solved the hardest problem? When you got stuck, did you stay with it a little longer than you would have a month ago? Did the work still sound like you?

If the answer to most of these is yes, the mix is right. If you notice yourself reaching for the assistant out of habit before you have thought at all, nudge the balance back towards your own first attempt for a few days. There is no perfect ratio, and it is fine for the balance to shift with the kind of week you are having. The goal is simply to stay the person who is doing the thinking.

The same idea scales beyond one desk. A teacher can invite students to submit their first attempt alongside any AI-assisted version, so that the learning is visible and the assistant is welcome. A team lead can encourage people to share their best coaching prompts the way they might share a useful template. Parents can encourage a "try first" rule at homework time that feels like a game rather than a restriction.

What these have in common is a positive framing. Instead of asking people to avoid a powerful tool, they ask people to use it in a way that leaves them more capable, which is an invitation most people are happy to accept.

No. Used well, it is a powerful helper. The research suggests the way you use it matters, so a try-first, ask-second routine is a sensible way to get the benefits while keeping your own skills growing.

Put a timer between you and the chat window. A single 25-minute block of your own effort is usually enough to make real progress or to arrive at a much better question.

Share what you have tried and ask for one hint, a question that points you in the right direction, or a check of your reasoning after you have finished. Saying "do not solve it for me" in your prompt helps.

Not necessarily. The study behind the idea looked at short sessions with maths and reading questions, so treat the routine as a practical suggestion rather than a proven rule. Routine chores are a great fit for full delegation.

A timer gives you a clear, small commitment, which makes it easier to begin and to stay with a task. Many people find that once they start, the hardest part has already passed.

One Pomodoro is a good starting point. If the task is small, ten or fifteen minutes may be plenty, and if it is a deep problem you might give it two blocks. Adjust to what feels challenging without feeling stuck for no reason.

Then use AI freely. The routine is for the work where you want to learn or where your own judgement matters most. Plenty of tasks are best finished quickly, and that is a perfectly good use of the tool.

It can. Starting is often the hardest part, and a short, low-pressure first block is an easy way in. You are only asking yourself for a rough attempt, and the assistant is waiting in the break to help you improve it.

Build your focus, one block at a time

Pair a try-first block with an AI-assisted block and see how your focus changes over a week.

Open the free timer: getfocusflow.info

Sources

Liu, G., Christian, B., Dumbalska, T., Bakker, M. A., & Dubey, R. (2026). AI Assistance Reduces Persistence and Hurts Independent Performance .

UC Berkeley Research. Using AI for Just 10 Minutes Erodes Your Ability to Persist at Hard Things . October 2026.