Tell someone you used a spelling checker and nobody cares. Tell them you used a calculator and nobody demands to see your long division. Tell them you looked something up on Google and nobody accuses you of outsourcing your intelligence.
Tell them AI helped you with something, though, and the temperature in the room suddenly drops five degrees. There is often a pause. Maybe a raised eyebrow.
“Oh. You used AI.”
The words are delivered with the same tone someone might use after learning that the homemade lasagna came from the frozen-food aisle.
I understand some of the reaction. We have all seen terrible AI-generated articles, soulless illustrations, confidently incorrect answers, mangled code, fake photographs, disposable marketing copy, and social media posts that somehow use seventeen paragraphs to say absolutely nothing.
There is a lot of AI-generated garbage in the world. There was also a lot of garbage written before AI arrived. AI did not invent incompetence. It merely gave incompetence a faster keyboard.
The Pencil
I think of AI as a pencil. The first thing you are taught about a pencil is how to hold it. Before you can draw, shade, sketch, write, or calculate, you need to learn how the tool behaves in your hand.
Then you begin discovering what else it can do.
Hold it upright and it produces a narrow line. Angle it and you can shade. Apply more pressure and the line becomes darker. Sharpen it and you get finer detail. Sharpen it too aggressively and you eventually discover that it is possible to spend more time preparing a pencil than actually using it.
The pencil is simple. Using it well is not.
Give a pencil to a toddler and you may get an enthusiastic circle accompanied by several unexplained marks on the table. Or wall. Give the same pencil to a skilled illustrator and you may get a portrait that looks ready to breathe.
The pencil did not create the difference. The person holding it did.
AI works much the same way. It can help a capable person explore an idea, challenge an assumption, review code, organize research, test an argument, or identify something they overlooked. It can also help someone generate twelve pages of convincing nonsense before lunch. The existence of the second person does not invalidate the work of the first.
“She Used a Word Processor”
I sometimes wonder whether the arrival of the word processor irritated professional typists as much as AI irritates people today. I imagine an office conversation sometime in the early days of personal computing, “Oh, look, Carol. That memo she’s holding? She did that using a word processor. She’s not a real typist.”
The word processor offered spell checking, easy revisions, movable paragraphs, formatting, templates, and the miraculous ability to correct a mistake without applying correction fluid and waiting for it to dry. It unquestionably changed the work.
It probably eliminated the need for some skills while making other skills far more valuable. Typing a flawless page on the first attempt mattered less. Organizing information, editing clearly, communicating effectively, and understanding the new technology mattered more. The tool did not make communication unnecessary. It changed which parts of the process required the most human effort.
That is what tools do.
Desktop publishing did not eliminate design. Digital cameras did not eliminate photography. Calculators did not eliminate mathematics. Integrated development environments did not eliminate programming. Search engines did not eliminate research. They changed the work.
Some old skills became less important. Some remained essential. New skills appeared.
AI is another change in that progression, although admittedly a larger and stranger one. Unlike a pencil or word processor, it can produce something that resembles a finished answer. That makes it unusually easy to confuse operating the tool with understanding the result. But that is an argument for learning to use it properly, not for pretending it does not exist.
I Heard This Story Before
As ColdFusion developers, we’ve been living this argument for decades. One of ColdFusion’s earliest taglines was “Making hard things easy.” Sound familiar? I’ve lost count of the number of times I’ve heard someone dismiss ColdFusion because it “hid” too much complexity.
Apparently writing <cfquery> was cheating.
Real programmers wrote fifty lines of boilerplate to connect to a database. They hand-parsed form submissions. They manually escaped everything. Real programmers suffered. "I used to walk to school and back home five miles each day! Uphill! Both ways!"
Except… they didn’t.
The good programmers used whatever tools let them spend less time solving yesterday’s problems and more time solving today’s. ColdFusion didn’t eliminate understanding. It eliminated repetitive plumbing. That’s a very important distinction.
Nobody ever looked at <cfquery> and concluded, “Well, I guess SQL is obsolete now.” You still needed to understand indexes. Transactions. Joins. Performance. Security.
The tag didn’t replace knowledge. It amplified it. AI feels remarkably similar. Knowing how to ask an LLM for code doesn’t eliminate the need to understand the code it produces. If anything, it makes understanding more important.
Just because something is easier doesn’t make it cheating. Sometimes it just means we’ve invented a better pencil.
Why Does AI Make People Cringe?
Some of the discomfort is justified. People cringe because they have been forced to read AI-generated writing that nobody bothered to edit. They cringe because companies use AI as an excuse to replace experienced workers with systems that are cheaper, faster, and significantly worse. They cringe because AI can confidently fabricate facts, sources, methods, and explanations. They cringe because people present generated work as evidence of expertise they do not possess. They cringe because questions about copyright, training data, consent, privacy, and attribution have not been resolved to everyone’s satisfaction.
Those are real concerns.
But there is another layer to the reaction: the belief that using AI is inherently dishonest or intellectually lazy. That conclusion does not follow.
Using AI badly can be lazy. So can copying code from a decade-old forum answer without understanding it. So can installing a library because its README promises to solve everything. So can reusing the same architecture on every project because thinking about alternatives would take too long.
Laziness is not a feature of the tool. It is a decision made by the operator.
AI Isn’t Magic. It’s Marketing.
Let’s be honest for a minute. Every tech company on Earth has apparently concluded that slapping “AI” onto their product increases its valuation by 37%. Your email has AI. Your browser has AI. Your IDE has AI. Your search engine has AI. Your note-taking app has AI.
I’m half expecting my garage door opener to announce that it now features “AI-powered opening technology.” Congratulations. You invented… opening.
The problem isn’t AI. The problem is companies desperately trying to solve problems nobody actually has. Nobody has ever looked at a calculator and thought, “You know what this needs? A chatbot.”
Yet here we are. Every week another startup proudly announces that they’ve “revolutionized” something by inserting a text box that talks back. I don’t need AI to summarize my grocery list or to rewrite my calendar reminder. I certainly don’t need AI deciding whether I really meant to order crunchy peanut butter.
Sometimes software should just... do the thing.
Good AI disappears into the workflow. Bad AI interrupts the workflow to remind you it exists. That’s the difference.
AI Does Not Give You Judgment
One of the most dangerous misconceptions about AI is that it gives inexperienced people expertise. Whoa, Nellie, does it not. It gives inexperienced people output. Those are very different things.
A language model can produce code that looks reasonable. It may even work. But it does not know whether the code fits your application, follows your conventions, creates a security vulnerability, introduces a race condition, ignores an edge case, or quietly makes future maintenance miserable.
Someone still needs to evaluate it. That evaluation requires knowledge and experience. In my own work, AI can help me investigate unfamiliar code, compare implementation strategies, generate test cases, explain an error from another angle, or challenge my interpretation of a problem. It can help me move through possibilities faster. But the useful part is not simply getting an answer. The useful part is knowing what to ask next.
- Why did it choose this approach?
- What assumptions is it making?
- What breaks when the data is incomplete?
- How does this behave under load?
- What security boundaries have been ignored?
- What is the simpler alternative?
- What evidence would prove this explanation wrong?
Those questions come from experience. AI can participate in the investigation, but it cannot assume responsibility for the conclusion. The person using it still owns the work.
The Tool Amplifies the Operator
AI is an amplifier. It can help a good writer explore more variations, but it can also help a bad writer produce more bad writing.
AI can help an experienced developer investigate a problem quickly, but it can also help an inexperienced developer install a remarkably sophisticated production outage.
AI can help someone organize genuine knowledge, or it can help them disguise the absence of it.
This is why arguments about whether AI output is “good” often go nowhere. The quality depends heavily on who is operating it, how it is being used, and what happens after the initial response appears. A professional does not accept the first answer merely because it arrived in a neatly formatted box. They question it. Test it. Rewrite it. Verify it. Remove the unnecessary pieces. Correct the mistakes. Add the missing context. Reject the entire thing when it heads in the wrong direction.
Sometimes the most productive result of an AI session is discovering that the proposed solution is completely wrong. That is still useful. A pencil sketch does not need to become the finished painting to have value. It may simply help the artist understand the composition.
AI Doesn’t Replace Good Employees. It Makes Them More Valuable.
One trend worries me far more than AI itself. Executives who look at it and immediately see payroll reductions. There’s a fantasy circulating through boardrooms that AI means you can fire half your developers, half your support staff, half your writers, and somehow produce twice as much work.
That’s not augmentation. That’s wishful thinking.
The companies getting the most value from AI generally aren’t replacing experienced people; they’re making experienced people dramatically more productive. That’s an important distinction.
A senior developer with AI is usually faster than the same senior developer without AI. That doesn’t magically make a junior developer interchangeable with twenty years of experience. In fact, it often makes expertise more valuable, because AI produces more ideas that require experienced people to validate. Someone still has to recognize bad architecture, catch the security hole, and notice that the generated SQL happily drops the production database.
Research into how people actually use generative AI already suggests that augmentation is at least as important as automation. An Anthropic analysis of more than four million Claude conversations classified 57% of observed usage as augmenting human capabilities and 43% as automating tasks. That does not settle what AI will eventually do to employment, but it does undermine the simplistic idea that its only purpose is to remove a person from the process.
Hiring patterns are similarly complicated. Employers are increasingly asking for AI skills across technical and nontechnical roles, suggesting that many jobs are being redesigned around the technology rather than simply erased. At the same time, some companies are cutting positions while hiring for different, often more specialized ones. AI is changing the composition of work, but anyone claiming that the final headcount arithmetic has already been settled is selling something.
Responsible companies don’t ask, “Who can we replace?” They ask, “What can we remove from people’s day so they can spend more time doing the work only humans can do?” That’s a much better question.
Prompting Is Not the Whole Skill
There is a temptation to reduce AI proficiency to “prompt engineering,” as though finding the correct magic phrase will cause the machine to dispense flawless answers. Prompting matters, but it is only the beginning. Knowing how to ask for something is useful. Knowing whether the response is any good is essential.
The real skill involves framing the problem, supplying relevant context, recognizing missing information, breaking large tasks into manageable pieces, verifying claims, identifying hallucinations, refining the output, and deciding when not to use AI at all.
In other words, the valuable skill is judgment. Learning to hold the pencil is important. It is not the same thing as learning to draw.
There Is a Difference Between Assistance and Substitution
The uncomfortable truth is that some people are using AI to avoid doing any meaningful work. Students submit generated assignments they have not read. Developers ship code they cannot explain. Companies publish support documentation containing invented features. Applicants send cover letters that could be addressed to any company in any industry on any planet.
That deserves criticism.
But the problem is not that AI participated. The problem is that the human being abandoned responsibility. There is a meaningful difference between asking AI to help improve an argument you have developed and asking it to manufacture an opinion you do not have.
There’s a difference between using it to review code you understand and using it to produce code you cannot maintain. There’s also a difference between using it to identify possible research directions and citing whatever sources it invents. There's a difference between assistance and substitution. The dividing line is not always how much AI was used. It is whether the person submitting the final work understands it, stands behind it, and accepts responsibility for its accuracy.
Nobody Awards Points for Avoiding Tools
There is a peculiar kind of professional pride built around performing work the difficult way. Sometimes that is appropriate. Understanding fundamentals matters. A developer should understand the code they deploy. A writer should be able to express an idea without asking a machine to invent one. A designer should understand composition rather than endlessly generating variations until something looks acceptable.
But difficulty is not automatically virtue. Nobody should receive extra points for spending four hours on a task that could have been completed carefully in one.
The objective is not to avoid assistance. The objective is to produce good work. That requires choosing tools intelligently. Sometimes AI is the right tool. Sometimes documentation is better. Sometimes a search engine is better. Sometimes the best option is to close every browser tab, stare at the problem, and think.
Professionalism means knowing the difference.
We Are Still Learning to Hold It
AI is in its infancy. The tools will improve. The rules around them will evolve. Some current uses will look embarrassingly primitive in a few years. Some of today’s loudest promises will collapse. Some of today’s loudest objections will become irrelevant.
During that process, we need skepticism. We should question accuracy and challenge unethical implementations. We should resist replacing expertise with generated plausibility, and we should care about attribution, privacy, intellectual property, and the consequences of automating decisions.
But skepticism should not become performative helplessness. Refusing to learn about AI does not make someone more authentic. It simply leaves them less informed about a technology that is already changing how work gets done. You don't have to admire every use of it. You don't have to accept every claim made about it. You don't have to use it for every task.
But dismissing all AI-assisted work because some people produce garbage with it is like dismissing pencils because someone drew on the wall. The pencil is not the masterpiece. The pencil is not the artist. The pencil is a tool. Learn how to hold it.
AI helped me refine this article. There. I said it.
AI also generated all of the monkey photos used throughout this website. Do you think I can afford a real monkey to pose in a stupid orange hoodie? In this economy? I think not.