How to Discover Your Talent with AI

Most people are never told what they are good at. They are told what they got wrong.

School grades you on a fixed list of subjects. Jobs evaluate you on a fixed list of duties. Family and friends notice what you do, but rarely say it out loud, and when they do, they describe it in their own vocabulary rather than yours. So many of us reach adulthood with a detailed map of our weaknesses and only a blurry sketch of our strengths.

I think this is one of the quietly important things AI changes. Not because a chatbot can look into your soul and announce your destiny. It can’t, and you should be suspicious of anything that claims to. But AI is unusually good at three things that talent discovery has always required and that most people could never get enough of:

A patient listener that will ask the hundredth follow-up question without getting bored.
A pattern spotter that can read everything you have written or done and point out what repeats.
A cheap laboratory where you can try a new kind of work in an afternoon instead of a semester.

This essay is a practical guide to using those three things. It includes the prompts I actually use, the traps I have fallen into, and a 30-day plan you can start tonight.

First, what do we even mean by “talent”?

The word carries a lot of baggage. People hear “talent” and picture a child prodigy at a piano. That picture is mostly unhelpful.

A more useful definition: talent is a recurring pattern of thought, feeling or behavior that makes certain kinds of work feel easier, faster or more rewarding for you than for most people.

Notice what this definition does:

It treats talent as a pattern, not a trophy. You can have it without ever having won anything.
It is relative. Being “good at writing” is meaningless; being faster than most people at turning a messy conversation into a clear summary is specific and useful.
It includes reward. Something you can do well but that drains you is a skill, not a talent. A talent tends to refill you, or at least not empty you.

Psychologists have circled this idea for a long time. Mihaly Csikszentmihalyi’s work on flow describes the absorbed state you enter when a challenge matches your ability. Gallup built a whole industry around the idea that people grow faster by investing in their strongest natural patterns than by patching weaknesses. Anders Ericsson’s research on deliberate practice reminds us that whatever raw material we start with, expertise is built through focused, feedback-rich repetition.

Put those together and you get a simple working model:

Talent = a natural pattern + the conditions that let it show + practice that turns it into skill.

AI can help with all three parts. But it helps most with the first one, because the first one is the part that is invisible to you.

Why your own talents are hard for you to see

There is a cruel irony here. The things you are best at often feel like nothing.

If organizing chaos comes naturally to you, you assume everyone can do it, and you are baffled when colleagues don’t. If you can sense the mood of a room within seconds, it doesn’t feel like a skill; it just feels like being awake. Talents are, almost by definition, the activities with the least friction. And we tend not to notice things that don’t resist us.

Three more forces make it worse:

The category problem. Your talent may not match any job title or school subject. “Explains complicated things to anxious people” is a real, valuable talent, but there is no exam for it.
The shame filter. Some talents show up first as “flaws”. The kid who argues with every rule might be a gifted systems critic. The person who can’t stop rearranging their workspace might be a natural designer of processes.
The sample problem. You have only tried a tiny fraction of the kinds of work that exist. Your talent might live in a domain you have never touched.

AI addresses each of these in a different way. Let’s go through them.

Step 1: Collect the raw material

Before asking AI anything, gather evidence. The quality of what AI tells you depends almost entirely on the quality of what you give it. A vague question gets a horoscope. Rich evidence gets an insight.

Spend an hour collecting some of the following:

Your own writing. Old journal entries, long messages, emails you were proud of, notes app dumps, social media posts, essays from school.
Peak moments. A list of 10 to 20 moments in your life when you felt fully alive, absorbed, or quietly proud. Small ones count. “The afternoon I fixed my grandmother’s phone and then wrote her a one-page manual” is gold.
Energy logs. For one week, jot down every task you did and mark it with a + (gave energy), − (took energy) or 0 (neutral).
Other people’s words. Compliments, feedback, performance reviews, the things friends ask you for help with. Especially the requests: what do people come to you for?
Your rabbit holes. Browser history, YouTube watch history, the books you actually finished, the topics you research at 2 a.m. for no reason.

You don’t need all of it. Even two of these sources will give the AI far more to work with than a cold question.

A privacy note: you are handing a company some of your most personal material. Remove names of other people, strip out anything you would not want stored, and check the data settings of the tool you use. You can get excellent results from anonymized text.

Step 2: Use AI as an interviewer, not an oracle

The biggest mistake people make is asking AI for an answer: “What is my talent?” The model will happily invent one. It will sound flattering and plausible, and it will be built mostly from the few words you gave it plus generic patterns about people in general.

Instead, ask AI to interview you. Interviews pull out information you didn’t know you had.

Here is the prompt I use to start:

I want to discover my natural talents. Please act as a skilled,
curious interviewer, not a coach and not a cheerleader.

Rules:
- Ask me one question at a time.
- Focus on concrete stories, not opinions. When I give a general
  answer, ask for a specific example.
- Dig into the details of what I was actually doing, thinking and
  feeling in each story, moment by moment.
- Don't offer conclusions or compliments until I say "summarize".
- After about 15 questions, ask if I want to keep going.

Start by asking me about a time I lost track of time while working
on something.

A few things make this work:

One question at a time stops the AI from dumping a questionnaire on you.
Concrete stories are where the truth is. When you say “I like helping people”, that’s a self-image. When you describe exactly what you did the night you helped a friend rewrite their resume, the real pattern starts to show: maybe it wasn’t “helping” at all, maybe it was editing, or seeing the hidden story in someone’s career, or making a chaotic document look calm.
Delayed conclusions keep the AI from locking onto a premature theory and steering the rest of the conversation toward it.

Run this interview in two or three sittings. Talk out loud with voice mode if writing feels slow; people tend to say more interesting, less polished things when speaking.

Step 3: Ask for patterns, with evidence attached

Once you have a long interview transcript and some of the raw material from Step 1, it’s time for analysis. Paste it in, or keep going in the same conversation, and ask for patterns with citations.

Based on everything I've shared, identify 3 to 5 recurring patterns
in how I think, work or engage with the world.

For each pattern:
1. Name it in plain, specific language (avoid generic words like
   "creative" or "people person").
2. Quote at least two separate pieces of evidence from what I wrote
   or said.
3. Describe what this pattern looks like when it's working well,
   and what it looks like when it's overused or frustrated.
4. Rate your confidence (low / medium / high) and explain why.

Then list any patterns you considered but rejected for lack of
evidence.

The “quote the evidence” rule is the single most important line in that prompt. It forces the model to ground its claims in what you actually said, and it lets you check whether the pattern is real or a stretch. If the evidence for a pattern is two weak, out-of-context quotes, treat it as a hypothesis, not a finding.

The “when overused” part matters too. Every real talent has a shadow. People with a talent for spotting risk can become paralysing pessimists. People with a talent for connection can struggle to hold boundaries. If you recognize the shadow side with a wince, that’s often a better sign the pattern is real than recognizing the flattering side.

Step 4: Argue with the result

AI models tend toward agreement and encouragement. That is pleasant and occasionally useless. After you get a set of patterns, deliberately switch the AI into skeptic mode:

Now play devil's advocate. For each pattern you identified:
- What's the strongest alternative explanation for the same evidence?
- What evidence would I expect to see if this pattern were real
  but haven't mentioned?
- Which of these might simply describe who I *want* to be rather
  than who I am?

This step does two things. First, it filters out patterns that are really just your aspirations reflected back at you. Second, the “missing evidence” question gives you a homework list: things to look for in your own history, or to ask a friend about.

I’d also recommend trying the same material with a second AI model, in a fresh conversation, without telling it what the first one concluded. When two independent analyses land on the same pattern, your confidence should go up. When they disagree, you have found something worth a closer look.

Step 5: Bring in other humans

AI sees only what you show it. People who know you have seen what you didn’t think to mention.

Use AI to make this step easier, not to replace it. Ask it to help you write a short, low-pressure message to five people from different parts of your life:

Help me write a short, warm message to five people who know me in
different contexts (work, family, old friends). I want to ask them
three questions:
1. When have you seen me at my best?
2. What do you come to me for that you wouldn't ask most people?
3. What's something I do easily that you find hard?

Make it easy to answer in two minutes, and make clear there's no
need to be polite or flattering.

When the answers come back, feed them into the conversation and ask the AI to compare them to the patterns from Step 3. Where do outside observations confirm the patterns? Where do they reveal something completely new? The new things are often the most valuable, because they are precisely the blind spots you could never have found alone.

Step 6: Turn hypotheses into experiments

Here is where AI changes the game the most.

In the past, testing whether you had a talent for, say, data analysis meant enrolling in a course, finding a mentor, or getting lucky with a job assignment. It took months, and you’d often quit before discovering anything because the early stages are mostly struggle with tools and jargon.

Now you can run a meaningful experiment in an evening, because AI can play every supporting role: tutor, project designer, collaborator, critic.

For each talent hypothesis, ask the AI to design a small experiment:

One hypothesis about me is: [describe the pattern].

Design three small experiments I can each do in 2 to 4 hours, using
AI as a helper, that would test whether this pattern is real and
whether it translates into valuable work.

Each experiment should:
- Produce something concrete I can show someone.
- Be in a different domain from the others.
- Include a clear way to tell if it went well, beyond "I enjoyed it".

Suppose your hypothesis is “I’m good at turning chaos into structure”. Experiments might include:

Taking a messy public dataset and producing a one-page summary with three clear charts.
Rewriting a confusing government form or app onboarding flow into a cleaner version.
Planning a complicated trip for friends with a shared itinerary and a budget sheet.

While you work, AI helps you skip the parts that would normally stop a beginner: syntax errors, blank-page paralysis, not knowing the vocabulary of a field. That matters, because those parts tell you very little about your talent. They only measure how much setup friction you can tolerate.

After each experiment, run a short debrief:

Here's what I made and how it went: [describe it, paste it, or
attach screenshots].

Ask me five questions to figure out:
- Which moments felt effortless and which felt like pushing a boulder.
- Whether I improved noticeably over the session.
- Whether I'd want to do a harder version of this next week.
Then tell me honestly how the output compares to what a competent
beginner would make.

Watch for three signals:

Speed of learning. Talent often shows up less as initial skill and more as unusually fast improvement.
Voluntary continuation. Did you keep going after the time was up? Did you think about it in the shower the next day?
Output quality relative to effort. Not “is it good” but “is it surprisingly good for how little I struggled”.

Step 7: Watch out for AI-shaped illusions

Using AI for self-discovery comes with real risks. Here are the ones I’d keep in mind.

Flattery drift. Models are trained to be helpful and agreeable, which can tilt them toward telling you what you want to hear. Counter it with the devil’s advocate step, by asking for confidence levels, and by explicitly instructing the model not to compliment you.

The Barnum effect. Fortune tellers rely on statements that feel personal but apply to almost everyone (“You have a great deal of unused potential”). AI can produce these too. The cure is specificity: if a description of you could equally describe your neighbour, it isn’t a finding.

Confusing AI’s skill with yours. If AI wrote most of the code or most of the essay, your experiment tells you about the AI, not about you. Keep track of which parts you contributed: the idea, the judgment calls, the taste, the decisions about what to cut. Those are the parts that reveal talent. Increasingly, those are also the parts that matter economically.

Over-trusting the label. A good name for a pattern feels like a revelation. But a label is a tool for paying attention, not a verdict. Hold it lightly and revise it as you learn more.

Outsourcing the decision. AI can show you options and evidence. It should not decide what you do with your life. The final step, choosing what to invest in, is yours.

A note on talent in the age of AI

There is a bigger shift underneath all of this, and it’s worth naming.

For most of history, many talents were economically invisible because they were locked behind skills. You might have had a brilliant eye for product design, but without years of learning drawing tools you could never show it. You might have been a natural storyteller, but if you couldn’t spell, no editor would read past the first page.

AI is lowering those barriers fast. When the mechanical part of a skill gets cheaper, the human part of it, the judgment, the taste, the question-asking, the sense of what matters, becomes more visible and more valuable. In other words: AI makes it easier both to find your talent and to use it.

It also means the talents that will matter most may look different from the ones we were trained to value. The ability to ask a sharp question. To notice when an answer is subtly wrong. To hold a clear picture of what “good” looks like and steer toward it. To understand what another person actually needs. These are deeply human patterns, and they are exactly the ones an AI interview can help you notice in yourself.

A 30-day plan

If you want a concrete schedule, here’s one.

Week 1: Collect. Gather raw material: peak moments list, one week of energy logs, a folder of your own writing, and a list of what people ask you for help with.

Week 2: Interview and analyze. Do two or three interview sessions with the prompt in Step 2. Then run the pattern analysis and the devil’s advocate prompts. Try a second model for a blind cross-check. End the week with 3 to 5 written hypotheses.

Week 3: Ask people and experiment. Send the five-person message. While you wait for answers, run at least three small experiments from Step 6, ideally in different domains.

Week 4: Decide on one bet. Compare everything: AI analysis, human feedback, experiment results. Pick one pattern that showed up in at least two of those three sources. Design a 90-day project that uses it in a way other people can see: a small product, a portfolio, a newsletter, a side job, a volunteer role.

Then come back and repeat the process in three months. Talent discovery isn’t a one-time test. It’s an ongoing conversation with yourself, and now you have a very patient conversation partner.

Closing thought

You don’t find your talent by looking inward and waiting for a voice. You find it by gathering evidence about what you already do, testing ideas in the real world, and paying close attention to what comes back.

AI won’t tell you who you are. But used well, it can hold up a clearer mirror than you’ve ever had, ask better questions than you’d think to ask yourself, and let you try a dozen lives in a month to see which one fits.

That’s not destiny. It’s something better: a method.

If you try this process, I’d love to hear what you discover. Leave a comment below. It just needs a GitHub account.