AI Can Answer Almost Anything — But Can You Ask the Right Question?

AI can give you an incredible answer to almost any question you know enough to ask.

That sentence is doing a lot of work.

Because for all the excitement around AI—and I use AI constantly—there’s a quiet assumption hiding underneath a lot of the conversation. The assumption is that the hard part is getting the answer. If we can just ask AI the right thing, it will respond with something useful, detailed, fast, and often surprisingly good.

And that is true.

But what happens when you don’t know the question?

That, to me, is one of the most important issues in the AI era. Not whether AI is useful. It obviously is. Not whether AI can help a business owner, developer, marketer, writer, designer, or entrepreneur move faster. It absolutely can. The bigger question is whether the person using AI knows enough about the subject to recognize what needs to be asked in the first place.

I use AI for development, marketing, troubleshooting, research, strategy, writing, planning, and all kinds of day-to-day problem solving. I’m not coming at this as someone who is anti-AI or hoping the technology goes away. I think AI is one of the most powerful tools we’ve seen in a long time.

But the more I use it, the more convinced I become that the quality of what I get back usually depends on the quality of what I put in.

Most people describe that as prompting.

Write a better prompt, get a better answer.

And yes, that’s true. Prompting matters. Context matters. Specificity matters. If you ask a vague question, you often get a vague answer. If you provide background, constraints, goals, examples, and what you’re trying to avoid, the answer usually improves dramatically.

But I think there’s another problem that “prompt engineering” alone can’t necessarily solve.

What if you don’t know enough about the subject to know what belongs in the prompt?

That’s the part I think a lot of people underestimate.

AI is very good at responding to what you ask. It can also infer a lot. It can suggest related considerations. It can sometimes notice gaps and warn you about them. But if you’re working in an area where you don’t know what matters, it’s very easy to ask a question that sounds complete but isn’t.

And the danger is that AI may still give you a very polished, confident, useful-looking answer.

The answer might be organized. It might sound professional. It might give you steps, best practices, code, copy, recommendations, a checklist, or a plan. It might feel complete.

But complete compared to what?

If you don’t know the field, how do you know what’s missing?

That’s where expertise becomes interesting in the AI era.

I don’t think AI eliminates experts. I think AI changes what expertise is worth.

If all an expert does is provide basic information that someone could get by asking AI, then yes, that part of their value is probably declining. A lot of generic information is now easy to access. Basic explanations, first drafts, introductory tutorials, common recommendations, and surface-level strategy are widely available in seconds.

But a good expert does more than provide answers.

A good expert knows what questions need to be asked.

That distinction matters.

Let’s take a simple example: building an application.

Suppose you want to create a customer login system. You tell AI, “Build me a customer login system.” Depending on the tools you’re using and the level of detail in your request, AI can probably help you build one. It may generate the database tables, the login form, the authentication logic, the page customers see after they log in, and even a decent user interface.

You might tell it, “I want a login page with an email field and password field. After customers log in, send them to their dashboard.”

Great.

It can do that.

But did you ask about session security?

Did you ask about password hashing?

Did you ask about rate limiting?

Did you ask about secure password resets?

Did you ask about account lockouts?

Did you ask about audit logs?

Did you ask about user permissions?

Did you ask about roles?

Did you ask about whether customers should be able to see specific records and not others?

Did you ask about data privacy?

Did you ask about what happens if an employee account is compromised?

Did you ask about how long a session should last?

Did you ask about two-factor authentication?

Did you ask about how the system should behave if someone tries to brute-force passwords?

Did you ask about what happens when someone leaves the company?

Did you ask about compliance requirements?

Did you ask about backups?

Did you ask about logging sensitive information accidentally?

Did you ask about how authentication integrates with the rest of the system?

Maybe you did.

But maybe you didn’t.

And if you didn’t, it may not be because you were careless. It may be because you didn’t know those were questions you needed to ask.

That is the problem.

A beginner may ask for “a login system” and think they’ve asked for the whole thing. An experienced developer hears “login system” and immediately thinks about security, edge cases, abuse, account recovery, data exposure, user roles, sessions, tokens, permissions, storage, encryption, and failure modes.

The difference is not just that the experienced developer knows more answers.

The experienced developer knows more questions.

That’s one of the key forms of expertise.

And it applies far beyond development.

Think about going to a doctor.

Imagine walking in and saying, “My shoulder hurts.”

You don’t hand the doctor a complete checklist of every possible thing they should investigate. You don’t say, “Please evaluate whether this is referred pain, rotator cuff damage, nerve impingement, arthritis, bursitis, a fracture, overuse, poor posture, or something else. Also consider whether my age, activity level, injury history, medications, and lifestyle are relevant.”

Most people don’t do that.

You say, “My shoulder hurts.”

Then the doctor starts asking questions.

Where does it hurt?

When did it start?

Did you injure it?

Was it sudden or gradual?

Does this movement hurt?

Does that movement hurt?

Is the pain sharp, dull, burning, or radiating?

Does it hurt at night?

Do you have numbness or tingling?

Have you had this before?

What do you do for work?

What physical activities do you do?

Are there other symptoms?

The expertise is not only in answering your original question. The expertise is in knowing what information is missing.

A good doctor doesn’t just respond to the sentence “my shoulder hurts.” A good doctor knows that the sentence is only the starting point. They know there are patterns, risks, possibilities, and consequences that the patient may not know how to identify.

Business works the same way.

If you ask AI, “How should I advertise my business?” it will give you an answer. It may suggest Google Ads, search engine optimization, social media, local marketing, email marketing, content marketing, referral programs, partnerships, retargeting, video, and maybe some kind of funnel.

That may be a perfectly reasonable answer.

But did you tell it your margins?

Did you tell it your customer lifetime value?

Did you tell it your close rate?

Did you tell it your average sale size?

Did you tell it your geographic limitations?

Did you tell it your capacity?

Did you tell it how quickly you can fulfill new demand?

Did you tell it which customers are actually profitable?

Did you tell it which customers are a drain on the business?

Did you tell it what you’ve already tried?

Did you tell it what worked, what failed, and why?

Did you tell it whether your sales process happens online, over the phone, in person, or through referrals?

Did you tell it whether you need leads immediately or whether you’re building long-term visibility?

Did you tell it whether you can handle a flood of low-quality leads or whether you need fewer, more qualified prospects?

Did you tell it whether your website converts?

Did you tell it whether your reviews are strong?

Did you tell it whether your offer is clear?

Did you tell it whether your competitors are outspending you?

Did you tell it whether your business is seasonal?

Maybe not.

And again, maybe that’s not because you were careless.

If you don’t know those factors matter, why would you include them?

That is where an expert changes the conversation.

A beginner may ask, “How do I get more leads?”

An expert may ask, “What kind of leads do you want, what are they worth, what can you afford to pay for them, what happens after they contact you, and are you sure more leads are actually the thing you need?”

Those are very different conversations.

Sometimes the obvious question is not the right question.

A business owner may think the issue is traffic. “I need more people on my website.” Maybe that’s true. But maybe the bigger issue is that the website doesn’t explain the offer clearly. Maybe it gets traffic but doesn’t convert. Maybe the contact form is too complicated. Maybe the phone number is hard to find on mobile. Maybe the service pages are too generic. Maybe the reviews are weak. Maybe the business is attracting the wrong audience. Maybe the pricing is misaligned. Maybe the sales process is leaking opportunities after the lead comes in.

If you ask AI how to get more website traffic, it can give you a good answer about traffic.

But if traffic is not actually the bottleneck, then a good answer to that question may not solve your problem.

That’s the risk.

The biggest risk is not always that AI gives you a bad answer.

Sometimes the bigger risk is that AI gives you a good answer to the wrong question.

And because the answer is well-written and plausible, it can give you confidence.

That confidence can be dangerous.

This is true in marketing. It’s true in software. It’s true in finance. It’s true in law. It’s true in health. It’s true in operations. It’s true in hiring. It’s true in strategy. It’s true in almost every field where decisions have dependencies and consequences.

If you ask AI, “What’s the best way to save money on my taxes?” it may provide general suggestions. But did you give it your entity structure, income sources, state, expenses, goals, timing, payroll situation, asset purchases, retirement plans, and audit risk? Did you even know which of those details matter?

If you ask AI, “Can I use this contract template?” it may help you understand it. But are there industry-specific terms you need? State-specific rules? Liability issues? Indemnification problems? Payment risks? Intellectual property concerns? Termination clauses? Insurance requirements? Data privacy obligations?

If you ask AI, “How do I improve my website SEO?” it may suggest keywords, content, meta descriptions, page speed, backlinks, local listings, and internal links. All useful. But did you clarify whether your issue is technical indexing, content quality, local competition, weak authority, poor conversion, bad targeting, cannibalized pages, thin service pages, or search intent mismatch?

Again, AI can help with all of this.

But the person using it needs enough context to either ask the right questions or recognize when they don’t know what they don’t know.

That phrase gets used a lot, but it really matters here.

There are things you know.

There are things you know you don’t know.

And then there are things you don’t know you don’t know.

AI is excellent for the first two categories.

If you know what you know, AI can help you move faster. It can draft, summarize, compare, code, rewrite, research, organize, and generate ideas.

If you know what you don’t know, AI can help you learn. You can ask for explanations, examples, pros and cons, terminology, checklists, and step-by-step guidance.

But the hardest category is the third one: the things you don’t know you don’t know.

That’s where experience has real value.

An expert has seen patterns. They’ve seen things go wrong. They know where projects usually break. They know which shortcuts are harmless and which shortcuts become expensive later. They know when a decision that seems efficient today creates a maintenance problem six months from now. They know when something sounds reasonable but doesn’t fit the specific situation.

An expert has probably watched people make the obvious decision without considering the second and third consequences.

That is a big part of what experience is.

Experience is not just having more facts stored in your head. AI has access to more facts than any individual expert ever could. If expertise were only about recalling information, AI would make a lot of experts far less useful.

But real expertise is often about judgment.

It’s about knowing which facts matter.

It’s about knowing what to ignore.

It’s about knowing when the standard recommendation doesn’t apply.

It’s about knowing when a simple solution is actually better.

It’s about knowing when a complex solution is necessary.

It’s about recognizing risk before it becomes visible.

It’s about identifying the missing piece.

It’s about asking, “What happens next?”

That last question is one of the most important ones in almost any project.

What happens next?

You can build the feature. What happens when users start using it?

You can run the ad. What happens when leads come in?

You can lower the price. What happens to perceived value, margins, and customer quality?

You can automate the process. What happens when the automation fails?

You can add a new service. What happens to your team’s capacity?

You can redesign the website. What happens to existing SEO rankings?

You can migrate platforms. What happens to old URLs, integrations, data, and workflows?

You can publish AI-generated content. What happens if it is inaccurate, generic, off-brand, or legally risky?

The first answer is rarely the entire answer.

And that’s where I think experienced professionals using AI become incredibly powerful.

Because AI can dramatically amplify expertise.

A professional who knows what questions to ask can use AI to investigate those questions faster than ever. They can explore options, identify edge cases, generate drafts, compare approaches, write code, test assumptions, summarize documentation, brainstorm alternatives, debug issues, and move through a problem with incredible speed.

That combination is powerful: expertise determining what needs to be asked, AI helping investigate and execute.

That’s very different from blindly handing the whole problem to AI and assuming the first confident answer is the right one.

Used well, AI can make a skilled person dramatically more effective.

Used poorly, AI can make an inexperienced person dramatically more confident.

That’s where the danger is.

Confidence and competence are not the same thing.

One of the reasons AI feels so powerful is that it communicates well. It can produce clean, structured, persuasive responses. It can sound like it knows what it’s talking about. And a lot of the time, it does provide helpful information.

But if you are not familiar with the field, it can be difficult to tell the difference between:

A complete answer and an incomplete one.

A best practice and a context-dependent suggestion.

A harmless shortcut and a dangerous shortcut.

A current recommendation and an outdated one.

A technically correct statement and a practically useless one.

A solution that works in theory and a solution that survives real-world use.

This doesn’t mean you shouldn’t use AI for subjects you’re not an expert in. I think you should. It’s an amazing learning tool.

But there’s a difference between using AI to educate yourself and using AI to make high-stakes decisions without enough understanding.

If you’re planning a vacation itinerary, the downside of a mediocre AI answer may be wasting an afternoon or picking a bad restaurant.

If you’re building a payment system, drafting legal terms, handling customer data, making medical decisions, restructuring a business, or launching a large ad campaign, the downside may be much more serious.

The stakes matter.

The reversibility of the decision matters.

The cost of being wrong matters.

The complexity of the field matters.

Your own ability to evaluate the answer matters.

That’s why I think the question is not simply, “Can AI do this?”

AI can do a lot.

The better question is, “Can I responsibly evaluate what AI gives me?”

That changes the frame.

For example, suppose I ask AI to help me write code in an area I understand well. I can look at the output and evaluate it. I can spot issues. I can test it. I can recognize when it’s solving the wrong problem. I can push back. I can ask follow-up questions. I can modify the approach. AI becomes a powerful assistant.

But if I ask AI to do something in an area where I have no real background, I may not know whether the output is good. I may only know whether it sounds good.

That is a major difference.

I can ask AI to explain legal language to me, and that may help me understand a document. But that doesn’t make me a lawyer.

I can ask AI about symptoms, and that may help me ask better questions. But that doesn’t make me a doctor.

I can ask AI for investment concepts, and that may help me learn. But that doesn’t make me a financial advisor.

I can ask AI for code, and it may generate something that works. But that doesn’t mean it is secure, maintainable, scalable, or appropriate for the project.

AI can narrow the gap between beginner and expert in some ways, especially for learning and basic execution. But it does not automatically give a beginner the judgment that comes from experience.

And sometimes judgment is the whole game.

Consider website projects. Someone may ask AI, “Build me a website for my business.” AI can help create copy, layouts, code, imagery prompts, page structure, calls to action, and even technical implementation. That’s useful.

But what pages does the business actually need?

What should be on the homepage?

What does the customer need to understand before they call?

What objections need to be addressed?

Which services deserve their own pages?

How should the site be structured for search engines?

What local signals matter?

What should the navigation include?

What can be removed?

How should the mobile experience work?

Where are users most likely to get confused?

What kind of trust signals are necessary?

What is the primary conversion goal?

What should be tracked?

What happens after someone fills out the form?

A website is not just a collection of pages. It is a business tool. If the questions going into it are shallow, the output may be shallow even if it looks polished.

That’s one of the biggest changes AI brings: it can make mediocre thinking look professional.

That may sound harsh, but I think it’s true.

A weak strategy can now be wrapped in beautiful language, clean formatting, and convincing explanations. A bad idea can get a great-looking presentation. A poorly considered plan can be turned into a professional document.

The polish can hide the weakness underneath.

So we need to get better at separating presentation from substance.

Just because something is clear does not mean it is correct.

Just because something is detailed does not mean it is complete.

Just because something is persuasive does not mean it is wise.

Just because something is fast does not mean it is good.

That doesn’t diminish AI. It just puts it in the right place.

AI is a tool. A very powerful tool. But tools amplify the person using them.

A skilled carpenter can do more with better tools. A beginner with expensive tools can still build something crooked. The tool matters, but so does the operator.

The same is true here.

AI can make a knowledgeable person faster, broader, and more productive.

AI can make a beginner learn faster, experiment more, and get unstuck.

But AI can also make a beginner overestimate their understanding.

That’s not a reason to avoid it. It’s a reason to use it with humility.

One of the best ways to use AI when you’re outside your expertise is to ask it to help you discover the questions, not just answer the first one.

Instead of only asking, “How do I advertise my business?” you can ask, “What information would a marketing strategist need before recommending an advertising plan for my business?”

Instead of only asking, “Build me a customer login system,” you can ask, “What security, privacy, and architecture considerations should be addressed before building a customer login system?”

Instead of only asking, “What should my contract say?” you can ask, “What are the major categories of risk I should discuss with an attorney before using a contract like this?”

Instead of only asking, “How do I improve my SEO?” you can ask, “What diagnostic questions should I answer before deciding on an SEO strategy?”

This is a better use of AI because it helps expose the unknowns.

You can ask AI to interview you.

You can ask it to identify missing inputs.

You can ask it to list assumptions.

You can ask it to point out risks.

You can ask it to explain what an expert would want to know.

You can ask it to challenge your plan.

You can ask it to give you the strongest argument against the thing you’re about to do.

Those are excellent uses of AI.

But even then, there are limits. AI can suggest what might matter. It can broaden your thinking. It can help you prepare for a conversation with an expert. It can make you a better client, a better manager, a better decision-maker.

But it may still miss something. Or it may not know which risk is most important in your specific context. Or it may treat all considerations as equal when experience would say, “This one matters most.”

That’s why I keep coming back to the role of expertise.

The future is not simply experts versus AI.

The future is experts using AI, and non-experts using AI more intelligently.

The people who will get the most value from AI are not necessarily the people who type the fanciest prompts. They are the people who understand the problem deeply enough to direct the tool.

Prompting is useful. But subject knowledge is better.

A good prompt from someone who understands the domain can produce an incredible result. A good prompt from someone who doesn’t understand the domain may still be missing critical context.

And sometimes the prompt itself reveals the problem.

If I ask, “How do I get more traffic?” maybe I’ve already assumed traffic is the solution.

If I ask, “How do I make this cheaper?” maybe I’ve already assumed cost is the problem.

If I ask, “How do I automate this?” maybe I’ve already assumed the process is worth preserving.

If I ask, “How do I rank number one?” maybe I’ve already assumed that ranking for that keyword will produce profitable customers.

If I ask, “How do I add this feature?” maybe I haven’t asked whether users need it.

Experts are valuable because they often challenge the premise.

They don’t just answer, “How do we do this?”

They ask, “Should we do this?”

They ask, “What are we trying to accomplish?”

They ask, “What happens if this works?”

They ask, “What happens if this fails?”

They ask, “What are we not considering?”

They ask, “Is this really the bottleneck?”

They ask, “What tradeoff are we making?”

AI can ask those things too, if prompted. But an expert knows when to ask them without being prompted.

That is the difference.

The real value of expertise is not only knowledge. It is discernment.

Discernment is hard to automate because it depends on context, priorities, tradeoffs, and experience. It depends on understanding not just what could be done, but what should be done given the situation.

A lot of business decisions are not purely technical. They involve constraints.

Budget.

Time.

People.

Risk tolerance.

Customer expectations.

Brand position.

Operational capacity.

Regulations.

Existing systems.

Maintenance burden.

Long-term goals.

Short-term survival.

AI can help analyze these things, but only if they are brought into the conversation. And if they are not brought in, the answer may be missing the most important part.

That’s why, before deciding AI can handle something by itself, I think there’s one question worth asking:

Do I know enough about this subject to recognize what I haven’t asked?

That question is simple, but it cuts through a lot.

If the answer is yes, AI may be an incredible DIY tool.

If you understand the field well enough to evaluate the output, spot gaps, test the result, and ask the right follow-up questions, then AI can help you move much faster. It can be a research assistant, junior developer, brainstorming partner, editor, analyst, and problem-solving companion.

But if the answer is no, be careful.

Not because AI is useless.

Not because AI is always wrong.

But because the biggest risk may not be that AI gives you the wrong answer.

The bigger risk may be that AI gives you a fantastic answer while neither of you realizes you asked the wrong question.

That’s the part worth remembering.

A wrong answer is sometimes easy to catch. It breaks. It contradicts known facts. It produces an obvious error. It looks suspicious.

A good answer to the wrong question is harder to catch.

It may look right.

It may feel helpful.

It may move you forward.

It may even solve the immediate thing you asked about.

But if the framing was wrong, you may be optimizing the wrong part of the system.

You may be solving a symptom instead of the cause.

You may be creating a future problem.

You may be making a decision without the context that would have changed the decision entirely.

This is why I think we need a more mature view of AI.

The conversation shouldn’t be, “AI can replace experts” or “AI can’t replace experts.”

That’s too simplistic.

AI will replace some tasks. It will reduce the value of some basic information work. It will make certain services faster, cheaper, and more accessible. It will also raise expectations. People will expect faster drafts, faster analysis, faster answers, and more informed conversations.

But expertise still matters.

In some cases, it may matter more.

Because when answers are abundant, questions become more valuable.

When everyone has access to information, judgment becomes more valuable.

When polished output is cheap, deep thinking becomes more valuable.

When speed increases, knowing where to aim becomes more valuable.

AI gives us access to more answers than we’ve ever had before. That is an enormous opportunity. I don’t want to minimize that at all. Used thoughtfully, it can help individuals and businesses learn, build, improve, and compete in ways that would have been much harder just a few years ago.

But answers are not the same as wisdom.

And output is not the same as strategy.

And confidence is not the same as correctness.

So use AI. Use it often. Use it creatively. Use it to learn. Use it to get unstuck. Use it to explore ideas that would have taken hours or days to investigate before. Use it to challenge your assumptions and expand your thinking.

But when the stakes are high, when the field is complex, or when you don’t know enough to recognize what might be missing, don’t confuse a good answer with a complete understanding.

That’s where an expert can still be invaluable.

Because sometimes the most valuable thing an expert knows is not the answer to the question you asked.

Sometimes the most valuable thing an expert knows is the question you didn’t know you needed to ask.