Why technique should mean something to AI users
Technique in the digital realm is more important than we realize. The importance of thinking about technique when we use AI.
We cook a lot, and my wife is truly a student of cooking (she’s very good at it). Besides that getting us an immensely delicious and healthy diet, it makes her our de facto ‘head chef’ and me the resident sous chef in our two-person household.
I mostly enjoy the rhythm and mindless rhyme of chopping, but there are those recipes that call for an interesting level of precision. One of those meals, an outrageously tasty quinoa egg bowl, has us thinly slice radishes. And I mean like sixteenth-of-an-inch-is-too-thick thin.1
Key Takeaways
Strong technique leads to better results, led by faster prompting
The best AI users make a commitment to maximizing the quality, adaptability, and insight of every interaction with AI
Like any craft, AI has a set of techniques we can use to master it at the micro level
At 10-15 radishes in a bunch, you can imagine the time to ponder. As I prepped for our latest go-round of this salad, I thought deeper about the road to this type of consistency. At which point I realized two things:
My eyes can only help so much in achieving consistent results
Most important is leaning in at the muscular level (in this case, a slant of the knife)
That second component was key, because it speaks to the idea of technique. A pianist can play faster and better because of it. A sprinter shaves milliseconds off their time because of it. And a sous chef can work faster and more reliably because of it (fans of FX’s The Bear, anyone?).
Let’s explore what technique looks like in how we approach AI.
What does technique mean in AI use?
At its core, technique in AI is about intentionality and structure. It’s easy to oversimplify and say the use of AI is:
Just about typing a request and waiting for a response.
Not working at all if it doesn’t work easy.
Supposed to be messy.
Think of a writer using AI to generate multiple variations of an idea instead of settling for the first draft. A researcher structuring their questions to go one level deeper with each iteration. A marketer fine-tuning AI-generated so the their first message is a 90% match for brand voice. A software engineer who knows how to rapidly test, refine, and even discard AI-generated code.
The difference between an amateur and an expert is technique.
A good technique serves multiple goals
Good technique isn’t just about efficiency—it’s about maximizing impact. The best techniques don’t serve a single purpose; they accomplish multiple goals at once.
Take sprinters. Their technique isn’t just about speed; it’s also longevity. The right stride length and arm movement shave off milliseconds and they reduce wear and tear on the body. A sprinter who masters their technique isn’t just faster in one race; they’re able to maintain peak performance over a career.
As a meta goal (akin to “Be the fastest sprinter”), you want to be a faster user of AI (where faster means faster and bias-aware and accurate). When we think about techniques, there are seven goals we can think about:
Faster prompting: Reducing the time spent working with AI (formulating prompts + iterating).
Higher accuracy: Getting to your desired outcome in fewer prompts.
Reusability: Creating prompts and workflows that can be easily adapted for different cases.
Scaling ourselves: Taking a single concept and apply it across multiple use cases or channels simultaneously.
Bias awareness and mitigation: Structuring interactions to reduce AI-generated biases.
Depth of insight: Uncovering deeper layers of information or alternative perspectives (using iterative prompting or deep research).
Strategic enhancement: Leveraging AI to push past mental and output blocks, maximize our strategic view, and explore unexpected directions.
So what does good AI technique actually look like? For starters, it looks like a commitment to maximizing the quality, adaptability, and insight of every interaction with AI.
The hallmarks of good AI technique
Precision in prompting
Goals: Higher accuracy, Faster prompting
A chef’s knife technique dictates the consistency of their slices. A clean, confident cut makes all the difference. AI prompting works the same way—structure and clarity are everything.
Strong AI users develop consistency in their phrasing. They don’t call something a “blog post” in one part of a prompt and a “message” in another. They form habits around using special characters, placeholders, and structured inputs to maintain clarity. Precision reduces ambiguity, and reducing ambiguity leads to better results.
A/B testing
Goals: Higher accuracy, Depth of insight
Pianists don’t play a piece once and assume it’s perfect. They test subtle variations to improve and refine. AI users can do the same.
A well-crafted AI prompt isn’t one-and-done. Running the same prompt multiple times, tweaking the phrasing slightly, or testing different structures can reveal better results. A/B testing is a logical way to commit to iterating.
Playing opposites
Goals: Strategic enhancement, Depth of insight
Sometimes, the best way to make it clear it’s your writing and not AI’s is to flip what it gives you on the head. Take for example the nice, plain phrase:
“AI users should shift strategies when the first attempt doesn’t land.”
There’s a lot of ways to flip this. We can do it in adding one word:
“AI users should not shift strategies when the first attempt doesn’t land.”
Changing zero words:
“AI users should abandon strategies when the first attempt doesn’t land.”
Most of the words:
“When we fail the first time, the next step is to dig in.”
If you’re asking why I’d do any of that, it’s a helpful writing trick. Mostly because it wildly changes the next thing I want to say.
Reacting to models (when structuring input/output)
Goals: Scaling ourselves, Higher accuracy
A boxer reads their opponent and adjust and reacts. There’s an analog type of awareness in AI—especially the opponent concept. A boxer faces a single opponent pretty rarely, and every opponent is different in a lot of ways. With the breadth of tools and the models behind them, we can afford to think of them not as opponents, but as something we need to read and adjust to in the moment.
The tool you choose matters—all of which have different strengths. Then there’s each of the models (Sonnet, Claude, GPT-4.5, o1, Flash, take your pick).
By and large, this a is a huge factor on structure of our input. Using tables vs. bullet points and considering input length are two things you’ll bob and weave with in real time. The output, and how easy it is to copy and paste the outputs also affects us in ways we don’t always think about. And just like a boxer adapts mid-fight, AI users should be nimble, shifting with each round.
Committing to meta-prompting
Goal: Reusability, Faster prompting
Sprinters obsess over the smallest details—stride length, arm movement, starting block position—because micro-adjustments lead to massive gains. AI users can do the same by refining how they craft prompts.
Instead of treating AI like a one-way tool, use it to help write better prompts. Ask it to suggest alternative phrasing, request an optimized structure, or generate a self-improvement checklist. The goal isn’t just to get a response—it’s to improve the way you ask for that response.
Documenting prompts
Goals: Reusability, Faster prompting, Scaling ourselves
Great playwrights don’t just write dialogue; they meticulously document stage directions, character motivations, and scene transitions to ensure the final performance lands. AI prompting is no different.
Keeping a record of effective prompts (and the results they generate) isn’t just useful—it’s essential. The best AI users don’t rely on memory; they build a library of prompts, tracking what works and what doesn’t. Even small tweaks in phrasing can make a massive difference, and having those insights documented speeds up future work.
How to improve your AI technique
Learn from bad outputs
A poor AI response is nothing more than feedback. Just like a chef adjusts seasoning to balance a dish, you refine prompts. Generic responses get fixed with more specific constraints. Misunderstood requests need clearer instructions, objectives or examples.
Every bad response is an opportunity to improve a prompt, and sometimes improving a prompt means throwing it out.
Be a shortcut hunter
Create structured templates, reusable prompts. Once you master those, find ways to create automations and shortcuts to streamline processes. As an example, using iOS Shortcuts to preload structured AI prompts for repeated tasks—like summarizing articles.
The less time you spend reinventing the wheel, the more time you spend refining.
Study prompts like great recipes
Analyze and read well-crafted prompts from other AI users. Keep track, at least mentally, what makes your own prompts effective. Borrow structures and refine them for your own use cases.
It’s not always about starting from scratch.
Like any skill, the people who get the best results with AI aren’t the ones who use it the most. They’re the ones using it with the most intention. So next time your neck-deep in a prompt, ask yourself:
“How do you want to cut these radishes?”
Radishes sliced less than 1/16” thick (they’re nearly translucent).