• The AI Journey — Part 2: AI Cannot Read Your Mind

    The AI Journey — Part 2: AI Cannot Read Your Mind

    I thought AI understood me.

    That was my mistake.

    A few days after my first conversation with AI, I found myself sitting in a small café beside a lake. I had not gone there with any particular plan. I simply wanted a quiet morning.

    The café was built slightly above the lakeshore, with wide windows facing the water. I chose a table in the corner, close enough to the glass that I could look outside without turning my chair.

    It was still early.

    The lake was almost completely covered in mist. Every now and then, a faint movement appeared across the water, then disappeared again. The mountains beyond it were barely more than dark shapes in the distance.

    Inside, the café was slowly waking up.

    The first coffee of the morning was being made behind the counter. I could hear the soft grinding of beans, followed by the familiar hiss of steam. A cup touched a saucer somewhere behind me. Someone pulled a chair across the wooden floor.

    There was quiet music playing in the background, so soft that I couldn’t recognize the song.

    I wrapped both hands around my coffee cup. It was warm against my fingers.

    Outside, a drop of condensation slowly moved down the window beside me.

    I watched it for a few seconds before opening my laptop.

    The screen looked unusually bright against the grey morning outside.

    I opened an AI chat.

    For a moment, I didn’t type anything.

    I thought about my first conversation with AI.

    Back then, I had been trying to understand what AI actually was.

    Now I was becoming curious about something else.

    How well could AI understand me?

    I looked at the empty chat window.

    There was something almost funny about the question.

    So I decided to find out.

    I typed:

    “Write something for me.”

    The answer appeared almost immediately.

    “Sure. What would you like me to write?”

    I smiled.

    Fair enough.

    I took another sip of coffee.

    “Write an article about AI.”

    A few seconds later, I had an article.

    It was clear. It was organized. It explained artificial intelligence, chatbots, image generation and the growing role of AI in everyday life.

    I read the opening paragraph. Then the next one. Then I stopped.

    Nothing was wrong with it.

    That was the problem.

    It was the kind of article that could have been written for almost anyone.

    I looked through the window.

    The mist was still sitting heavily over the lake.

    I knew there was something I wanted to say.

    I just hadn’t told the AI what it was.

    So I tried again.

    “I want it to feel personal. Like someone is discovering AI for the first time. It should be curious, not technical.”

    The answer changed.

    The writing became more personal. It talked about curiosity, uncertainty and the feeling of discovering something new.

    I read it slowly.

    Better.

    But still not quite right.

    I could feel the difference, even if I couldn’t completely explain it.

    I took my notebook from my bag and placed it beside the laptop.

    Then I typed:

    “Make it sound like I’m telling the story to a friend.”

    AI replied:

    “Sure. I’ll make it more natural and conversational.”

    The next version felt warmer.

    Less like an article.

    More like someone talking.

    I leaned back.

    A waiter walked past my table carrying two coffees. The smell of freshly ground beans followed him for a moment before disappearing into the room.

    Outside, the mist had begun to loosen.

    A small patch of water appeared between the trees.

    I looked back at the screen.

    Something was bothering me.

    The AI hadn’t suddenly understood what I was thinking.

    I had simply given it more information.

    I wanted to test that.

    So I typed:

    “Write something beautiful.”

    This time, the AI didn’t immediately start writing.

    It asked:

    “Sure. What kind of thing do you have in mind—a story, a poem, a reflection, or something else?”

    I looked at the question.

    Then I laughed quietly.

    The woman at the next table glanced over for a second, then returned to her coffee.

    I typed:

    “I don’t know. Just something beautiful.”

    The answer came back:

    “Okay. I’ll choose the direction.”

    It wrote about morning light, hope and new beginnings.

    I read it.

    It was nice.

    Actually, it was quite good.

    But it still wasn’t what I had imagined.

    I stared at the screen.

    For some reason, I felt slightly frustrated.

    Then I realized something.

    The AI wasn’t failing.

    I was leaving too much unsaid.

    So I typed:

    “Why doesn’t this feel like what I had in mind?”

    The reply appeared:

    “Probably because I had to make several choices for you. You gave me a general feeling, but not enough information about the specific idea you had in your head.”

    I stopped.

    Outside, the mist moved slowly across the lake.

    I watched the mountains for a moment before looking back at the screen.

    Then I typed:

    “So you don’t know what I’m thinking?”

    The answer was simple.

    “No. I only know what you share with me in the conversation. I can make guesses based on your words and the context, but I can’t access your private thoughts.”

    I leaned back in my chair.

    The café suddenly seemed quieter.

    Or maybe I was simply paying more attention.

    I could hear the soft hum of the refrigerator behind the counter. Somewhere near the entrance, the door opened and closed, letting in a brief breath of cool morning air.

    I looked at the lake again.

    The mist had lifted a little more.

    And suddenly, something became clear.

    I had been expecting AI to understand something I had never actually told it.

    That was my mistake.

    I opened my notebook.

    At the top of a blank page, I wrote:

    AI cannot read my mind.

    I looked at the sentence for a few seconds.

    Then I added:

    If I want a better answer, I have to give it a better picture of what I want.

    That felt more accurate.

    I wasn’t discovering a secret command.

    I was learning how to communicate.

    So I tried one more time.

    This time, I told the AI where I was.

    I told it about the quiet café, the lake outside the window, the mist over the mountains and the feeling of sitting there with a coffee while trying to understand something new.

    I told it that I didn’t want a technical article.

    I wanted the reader to feel like they were sitting there with me.

    Then I typed:

    “Can you write it from that feeling?”

    The answer was different again.

    Not because the AI had suddenly become more intelligent.

    I had simply given it more of me.

    I read the new version slowly.

    This time, I recognized what I had been trying to say.

    I looked outside.

    The mist was disappearing.

    More of the lake was visible now, reflecting the pale morning sky. The mountains had begun to emerge properly, their edges becoming clearer against the light.

    They hadn’t suddenly appeared.

    They had been there the whole time.

    I smiled.

    Maybe my ideas were like those mountains.

    Sometimes they were already there, somewhere inside me.

    But when I tried to explain them, I gave AI only the mist.

    A few words.

    A vague feeling.

    A general instruction.

    Then I wondered why the result didn’t look like the picture inside my head.

    I closed the notebook.

    My coffee had gone cold.

    I hadn’t even noticed.

    Before leaving, I asked one final question:

    “So what should I tell you when I want you to give me a better answer?”

    The AI replied:

    “Tell me what you’re trying to accomplish, give me the details that matter, and describe what you want the result to feel or look like. If my first answer isn’t right, tell me what’s missing. We can adjust it from there.”

    I read it twice.

    There was nothing revolutionary about the answer.

    But something had changed for me.

    I had arrived that morning thinking that AI needed to understand me better.

    I was leaving with a different thought.

    Maybe I needed to explain myself better.

    By then, the café had changed.

    More people had arrived. Conversations had replaced the quiet of the early morning. Cups moved across tables. The coffee machine hissed again.

    Sunlight finally reached through the windows and touched the wooden table in front of me.

    The lake had almost completely lost its mist.

    I closed the laptop.

    For the first time, I wasn’t thinking about finding the perfect prompt.

    I was thinking about something much simpler.

    A conversation works both ways.

    I tell AI what I mean.

    AI gives me something back.

    I look at it.

    I notice what’s missing.

    Then I explain again.

    And somewhere between those attempts, the idea inside my head slowly becomes clearer.

    I finished my coffee and packed my laptop.

    As I stood up, I looked through the window one last time.

    The mountains were completely visible now.

    They hadn’t changed.

    Only my view of them had.

    I walked out of the café and into the morning.

    And somewhere between the café door and the lakeside path, another question began forming in my mind.

    This time, I knew what I would do when I asked it.

    I would try again.

    To be continued…

    The AI Journey — Part 3: The Second Try

  • AI Is Coming for White-Collar Jobs: What the Data Actually Says

    AI Is Coming for White-Collar Jobs: What the Data Actually Says

    The debate has shifted. No longer “will AI take jobs?” but “which jobs, how many, and how fast?” A wave of new economic research is finally giving us sharper answers — and they are more nuanced, and more urgent, than either the optimists or the doomsayers predicted.

    The Data So Far

    A McKinsey Global Institute report released this quarter found that 30% of tasks in knowledge-worker roles — lawyers, accountants, analysts, marketers — are now automatable with current AI tools. Goldman Sachs estimates that 300 million full-time jobs globally are exposed to AI-driven automation. Crucially, “exposed” does not mean “eliminated” — but it does mean transformed, often significantly.

    Who’s Already Feeling It

    Paralegal work, first-draft copywriting, basic data analysis, customer support tier-1, and entry-level software testing have already seen headcount reductions at major firms. Junior roles that once served as training grounds for senior careers are shrinking — raising hard questions about how the next generation of professionals will develop expertise.

    The Counterargument

    Economists who study past technological transitions point out that new categories of work always emerge. Prompt engineering, AI auditing, model fine-tuning, and human-AI workflow design are already growing fields. The challenge is not the long-term equilibrium — it’s surviving the transition period, which historically disadvantages those without resources to retrain or relocate.

    The Number That Matters: Exposure Is Not Elimination

    One of the biggest mistakes in the AI-and-jobs debate is treating “exposure” and “replacement” as the same thing.

    Goldman Sachs’ estimate of roughly 300 million jobs exposed to AI automation refers to jobs containing tasks that could potentially be automated. It does not mean 300 million people are expected to lose their jobs. Goldman Sachs has repeatedly emphasized that many occupations are only partially exposed and that AI can complement workers rather than replace them.

    That distinction is becoming increasingly important as real-world evidence arrives.

    Goldman Sachs reported in September 2026 that AI adoption among major developed economies is around 15%–20%. Its analysis found that employment growth has slowed in some highly AI-exposed industries, including information and communication services, software publishing, advertising and call centers. At the same time, the measured effect on economy-wide hiring remains relatively limited.

    In other words, the transformation is visible, but the data does not yet support the idea that AI has simply erased huge numbers of jobs across the entire economy.

    AI Is Changing Tasks Before It Replaces Jobs

    A job is rarely one single activity.

    A lawyer researches cases, summarizes documents, drafts correspondence, communicates with clients and exercises judgment. A marketer researches audiences, writes copy, studies campaign performance and develops strategy. An accountant processes information, checks records, explains financial results and advises clients.

    AI can attack individual tasks inside these jobs.

    This is why task-level automation may be a better way to understand the coming disruption than simply asking whether an occupation will disappear.

    McKinsey’s recent research found that existing technologies could theoretically automate activities representing more than half of current U.S. work hours. But McKinsey explicitly cautions that this is not a forecast that half of U.S. jobs will disappear. The research instead points toward major changes in how work is divided between people, AI agents and other technologies.

    That distinction changes the question.

    Instead of:

    “Will AI replace accountants?”

    The more useful question is:

    “Which accounting tasks will AI perform, and what will accountants do with the time that remains?”

    That same logic applies to almost every white-collar profession.

    The Biggest Pressure May Be on Junior Workers

    One of the most important developments is the potential impact on entry-level employment.

    Traditionally, junior workers learn by doing relatively routine tasks. A young lawyer reviews documents. A junior analyst cleans spreadsheets. A new marketer prepares reports. A junior developer writes basic code and fixes straightforward bugs.

    These tasks may also be among the easiest for AI systems to assist with or automate.

    Goldman Sachs’ latest analysis specifically notes that junior workers may face stronger headwinds to hiring as AI adoption increases.

    This creates a potentially difficult cycle.

    If companies need fewer people to perform routine entry-level work, fewer graduates may get the traditional first step into a profession. But those junior employees were also the people who would eventually become experienced managers, specialists and leaders.

    The result could be a new kind of career problem:

    If AI removes some of the first rungs of the career ladder, how do workers climb it?

    This question may become just as important as the number of jobs technically capable of being automated.

    The White-Collar Jobs Most Exposed

    Exposure is not evenly distributed.

    Jobs that involve large amounts of structured digital information, text processing, repetitive analysis or predictable computer-based workflows are generally more exposed to generative AI and automation.

    Examples include:

    • Basic content drafting
    • Document summarization
    • Routine research
    • Data classification
    • Customer-support responses
    • Standard financial reporting
    • Basic spreadsheet analysis
    • Simple software coding
    • Document review
    • Administrative processing
    • Repetitive marketing tasks

    But exposure does not automatically mean disappearance.

    A customer-support representative might use AI to answer routine questions while handling difficult cases personally.

    A financial analyst might use AI to prepare the first version of a report and spend more time interpreting the results.

    A software developer might use an AI coding agent to produce routine code while focusing on architecture, testing and product decisions.

    The job changes because the machine takes over part of the workflow.

    The Human Skills That Become More Valuable

    AI can generate text, summarize information, produce code and analyze enormous amounts of data.

    That does not eliminate the need for human judgment.

    In fact, some research suggests that the skills required alongside AI may become more sophisticated.

    The International Labour Organization’s 2026 research on skills in the AI era reports increasing demand for cognitive, socioemotional, digital and AI-related skills. It also highlights AI literacy, adaptability, resilience and human agency as important capabilities as workplaces change.

    This means that “knowing how to use AI” may become only the starting point.

    Workers may increasingly need to know:

    What should AI do?

    What should humans do?

    How do we check the answer?

    What happens when the AI is wrong?

    How do we turn an AI-generated output into a useful business decision?

    Those questions require judgment rather than simple prompting.

    The Rise of the AI-Augmented Worker

    The most realistic near-term scenario may not be “human versus AI.”

    It may be human plus AI versus human without AI.

    Imagine two employees with similar experience.

    One spends several hours researching, drafting, formatting and summarizing a report manually.

    The other uses AI to produce a first draft, identify relevant information, generate alternative approaches and organize the material — then personally verifies the important facts and makes the final decisions.

    The second worker may be able to complete more work in the same amount of time.

    That changes competition within companies.

    The question may gradually become less about whether a worker’s occupation is technically automatable and more about whether that worker knows how to incorporate AI effectively into the job.

    New Jobs Are Appearing Too

    Technological disruption does not only destroy tasks.

    It also creates new demand.

    The World Economic Forum’s Future of Jobs Report 2025 projected that, across the broader set of technological, economic, demographic and environmental trends, 170 million jobs could be created globally by 2030 while 92 million could be displaced, producing a projected net increase of 78 million jobs. These figures are employer expectations and projections, not a guarantee of future employment outcomes.

    AI-related roles are among the areas expected to grow.

    The WEF identifies AI and machine-learning specialists, big-data specialists and related technology roles among the fastest-growing occupations in its employer survey.

    At the same time, the ILO’s 2026 research points to emerging technical work involved in developing and maintaining AI systems. It describes these roles as still relatively small and specialized, but growing as AI adoption spreads.

    The new employment landscape therefore isn’t simply:

    Old jobs → AI → no jobs

    It is more complicated:

    Old tasks → automation → redesigned jobs → new skills → new jobs

    What Happens to Writers?

    Writing is one of the clearest examples of this transition.

    AI can already produce:

    • Blog drafts
    • Product descriptions
    • Email drafts
    • Social-media copy
    • Summaries
    • Headlines
    • Ad variations
    • Basic reports
    • Outlines
    • Rewrites

    That creates obvious pressure on writers whose work consists primarily of producing predictable first drafts.

    But writing also contains activities that are harder to automate reliably:

    • Developing an original point of view
    • Interviewing people
    • Understanding an audience
    • Building trust
    • Investigating claims
    • Making editorial judgments
    • Developing a distinctive voice
    • Understanding cultural context
    • Taking responsibility for accuracy

    The likely result is not necessarily the disappearance of writers.

    It may be a shift toward fewer hours spent physically producing words and more time spent deciding which words should exist in the first place.

    What About Lawyers and Accountants?

    Legal and financial professions provide another useful example.

    AI can help review large quantities of documents, identify patterns, summarize contracts and prepare preliminary analyses.

    But legal advice involves responsibility, interpretation and judgment. Financial work similarly involves decisions that can have significant consequences for individuals and businesses.

    That means automation may remove parts of the workflow without removing the profession itself.

    The highest-value human contribution may increasingly move toward:

    judgment + verification + communication + responsibility.

    The more consequential the decision, the more important those elements become.

    The Productivity Question

    There is another side of the debate that receives less attention: what happens if AI allows the same number of workers to produce significantly more?

    The ILO’s 2026 review of empirical evidence finds that productivity gains from generative AI are real in some settings, although they remain uneven and are not yet consistently translating into higher measured output, earnings or employment. The review also identifies risks around inequality, younger workers’ employment opportunities and changes in job quality.

    This matters because productivity growth can change the economics of a business.

    If a company can serve more customers without proportionally increasing its workforce, it may expand.

    That expansion could create new jobs.

    But if productivity improvements mainly reduce the number of employees needed for existing output, employment pressure could increase.

    The outcome depends on how businesses use the additional capacity.

    The Transition Will Not Be Equal

    Perhaps the most important point is that AI’s impact will not be distributed evenly.

    A highly experienced professional who knows an industry deeply may use AI to become dramatically more productive.

    A new worker who has not yet developed foundational skills may find it harder to distinguish themselves.

    A worker with access to training may adapt quickly.

    Another worker may not have the time, money or educational resources to retrain.

    This is why the transition could create winners and losers even if the overall economy eventually produces new categories of employment.

    The ILO’s recent research specifically highlights the risk of growing inequalities and reduced employment opportunities for younger workers.

    What Workers Should Do Now

    The practical response is not to panic about every new AI model.

    It is to understand where AI intersects with your own work.

    Start by dividing your job into three categories:

    1. Tasks AI can assist with

    These are repetitive or information-heavy activities where AI can save time.

    2. Tasks AI can partially automate

    These require human review but can be accelerated significantly.

    3. Tasks that depend heavily on human judgment

    These involve relationships, accountability, strategy, leadership, creativity, negotiation or decisions where context matters.

    Then build your skills accordingly.

    Learn the AI tools relevant to your profession.

    But do not stop there.

    Develop domain expertise, communication skills, critical thinking and the ability to verify AI-generated information.

    The worker who understands both the profession and the technology may be in a stronger position than someone who understands only one.

    The Bigger Question Isn’t Whether AI Takes Jobs

    History suggests that technological change can eliminate particular tasks while creating new forms of work.

    But the transition can still be disruptive.

    For today’s workers, the uncomfortable reality is that the labor market does not have to experience mass unemployment for millions of careers to change.

    A job can remain technically intact while its responsibilities, salary structure, entry requirements and career path change dramatically.

    That may be the defining feature of the AI transition.

    The office worker of 2030 may still be a lawyer, accountant, marketer, designer, analyst or developer.

    But the daily work may look very different.

    The most important skill may not be competing against AI.

    It may be learning how to work effectively with it.

    The Bottom Line

    The available evidence does not support a simple story in which AI automatically eliminates hundreds of millions of jobs.

    It does support a more complicated story: AI is changing the tasks inside jobs, reshaping hiring, increasing demand for certain skills and putting particular pressure on routine and entry-level work.

    The transition is already underway.

    For workers, the safest assumption may be that the job description they have today will not be exactly the job description they have several years from now.

    The winners of the AI transition will not necessarily be the people who know the most about artificial intelligence.

    They may be the people who understand their own profession deeply enough to recognize what AI should do, what humans should do, and where the two work best together.

Category: Free Resources

I thought AI understood me.

That was my mistake.

A few days after my first conversation with AI, I found myself sitting in a small café beside a lake. I had not gone there with any particular plan. I simply wanted a quiet morning.

The café was built slightly above the lakeshore, with wide windows facing the water. I chose a table in the corner, close enough to the glass that I could look outside without turning my chair.

It was still early.

The lake was almost completely covered in mist. Every now and then, a faint movement appeared across the water, then disappeared again. The mountains beyond it were barely more than dark shapes in the distance.

Inside, the café was slowly waking up.

The first coffee of the morning was being made behind the counter. I could hear the soft grinding of beans, followed by the familiar hiss of steam. A cup touched a saucer somewhere behind me. Someone pulled a chair across the wooden floor.

There was quiet music playing in the background, so soft that I couldn’t recognize the song.

I wrapped both hands around my coffee cup. It was warm against my fingers.

Outside, a drop of condensation slowly moved down the window beside me.

I watched it for a few seconds before opening my laptop.

The screen looked unusually bright against the grey morning outside.

I opened an AI chat.

For a moment, I didn’t type anything.

I thought about my first conversation with AI.

Back then, I had been trying to understand what AI actually was.

Now I was becoming curious about something else.

How well could AI understand me?

I looked at the empty chat window.

There was something almost funny about the question.

So I decided to find out.

I typed:

“Write something for me.”

The answer appeared almost immediately.

“Sure. What would you like me to write?”

I smiled.

Fair enough.

I took another sip of coffee.

“Write an article about AI.”

A few seconds later, I had an article.

It was clear. It was organized. It explained artificial intelligence, chatbots, image generation and the growing role of AI in everyday life.

I read the opening paragraph. Then the next one. Then I stopped.

Nothing was wrong with it.

That was the problem.

It was the kind of article that could have been written for almost anyone.

I looked through the window.

The mist was still sitting heavily over the lake.

I knew there was something I wanted to say.

I just hadn’t told the AI what it was.

So I tried again.

“I want it to feel personal. Like someone is discovering AI for the first time. It should be curious, not technical.”

The answer changed.

The writing became more personal. It talked about curiosity, uncertainty and the feeling of discovering something new.

I read it slowly.

Better.

But still not quite right.

I could feel the difference, even if I couldn’t completely explain it.

I took my notebook from my bag and placed it beside the laptop.

Then I typed:

“Make it sound like I’m telling the story to a friend.”

AI replied:

“Sure. I’ll make it more natural and conversational.”

The next version felt warmer.

Less like an article.

More like someone talking.

I leaned back.

A waiter walked past my table carrying two coffees. The smell of freshly ground beans followed him for a moment before disappearing into the room.

Outside, the mist had begun to loosen.

A small patch of water appeared between the trees.

I looked back at the screen.

Something was bothering me.

The AI hadn’t suddenly understood what I was thinking.

I had simply given it more information.

I wanted to test that.

So I typed:

“Write something beautiful.”

This time, the AI didn’t immediately start writing.

It asked:

“Sure. What kind of thing do you have in mind—a story, a poem, a reflection, or something else?”

I looked at the question.

Then I laughed quietly.

The woman at the next table glanced over for a second, then returned to her coffee.

I typed:

“I don’t know. Just something beautiful.”

The answer came back:

“Okay. I’ll choose the direction.”

It wrote about morning light, hope and new beginnings.

I read it.

It was nice.

Actually, it was quite good.

But it still wasn’t what I had imagined.

I stared at the screen.

For some reason, I felt slightly frustrated.

Then I realized something.

The AI wasn’t failing.

I was leaving too much unsaid.

So I typed:

“Why doesn’t this feel like what I had in mind?”

The reply appeared:

“Probably because I had to make several choices for you. You gave me a general feeling, but not enough information about the specific idea you had in your head.”

I stopped.

Outside, the mist moved slowly across the lake.

I watched the mountains for a moment before looking back at the screen.

Then I typed:

“So you don’t know what I’m thinking?”

The answer was simple.

“No. I only know what you share with me in the conversation. I can make guesses based on your words and the context, but I can’t access your private thoughts.”

I leaned back in my chair.

The café suddenly seemed quieter.

Or maybe I was simply paying more attention.

I could hear the soft hum of the refrigerator behind the counter. Somewhere near the entrance, the door opened and closed, letting in a brief breath of cool morning air.

I looked at the lake again.

The mist had lifted a little more.

And suddenly, something became clear.

I had been expecting AI to understand something I had never actually told it.

That was my mistake.

I opened my notebook.

At the top of a blank page, I wrote:

AI cannot read my mind.

I looked at the sentence for a few seconds.

Then I added:

If I want a better answer, I have to give it a better picture of what I want.

That felt more accurate.

I wasn’t discovering a secret command.

I was learning how to communicate.

So I tried one more time.

This time, I told the AI where I was.

I told it about the quiet café, the lake outside the window, the mist over the mountains and the feeling of sitting there with a coffee while trying to understand something new.

I told it that I didn’t want a technical article.

I wanted the reader to feel like they were sitting there with me.

Then I typed:

“Can you write it from that feeling?”

The answer was different again.

Not because the AI had suddenly become more intelligent.

I had simply given it more of me.

I read the new version slowly.

This time, I recognized what I had been trying to say.

I looked outside.

The mist was disappearing.

More of the lake was visible now, reflecting the pale morning sky. The mountains had begun to emerge properly, their edges becoming clearer against the light.

They hadn’t suddenly appeared.

They had been there the whole time.

I smiled.

Maybe my ideas were like those mountains.

Sometimes they were already there, somewhere inside me.

But when I tried to explain them, I gave AI only the mist.

A few words.

A vague feeling.

A general instruction.

Then I wondered why the result didn’t look like the picture inside my head.

I closed the notebook.

My coffee had gone cold.

I hadn’t even noticed.

Before leaving, I asked one final question:

“So what should I tell you when I want you to give me a better answer?”

The AI replied:

“Tell me what you’re trying to accomplish, give me the details that matter, and describe what you want the result to feel or look like. If my first answer isn’t right, tell me what’s missing. We can adjust it from there.”

I read it twice.

There was nothing revolutionary about the answer.

But something had changed for me.

I had arrived that morning thinking that AI needed to understand me better.

I was leaving with a different thought.

Maybe I needed to explain myself better.

By then, the café had changed.

More people had arrived. Conversations had replaced the quiet of the early morning. Cups moved across tables. The coffee machine hissed again.

Sunlight finally reached through the windows and touched the wooden table in front of me.

The lake had almost completely lost its mist.

I closed the laptop.

For the first time, I wasn’t thinking about finding the perfect prompt.

I was thinking about something much simpler.

A conversation works both ways.

I tell AI what I mean.

AI gives me something back.

I look at it.

I notice what’s missing.

Then I explain again.

And somewhere between those attempts, the idea inside my head slowly becomes clearer.

I finished my coffee and packed my laptop.

As I stood up, I looked through the window one last time.

The mountains were completely visible now.

They hadn’t changed.

Only my view of them had.

I walked out of the café and into the morning.

And somewhere between the café door and the lakeside path, another question began forming in my mind.

This time, I knew what I would do when I asked it.

I would try again.

To be continued…

The AI Journey — Part 3: The Second Try