← Back to all free guides
A free guide by Mariah Brunner ↗ INSTAGRAM ↗ TIKTOK

How To Build
Taste With AI

Taste is three things you can practice. Pick the right problem, check the output, finish the last 20 percent. Here’s the drill for each one, plus the two habits that turn them into instinct.

Everyone is getting faster with AI. Almost nobody is getting better. Taste is the gap.

I’ve built two eight-figure businesses on AI systems, and what decides whether someone builds something great or forgettable is almost never the model. It’s taste. People think you either have it or you don’t. You can drill it, and every drill below runs on work you already have in front of you this week.

1. The Pick

You can build almost anything now, and that’s the trap. When building was hard, the difficulty picked for you. Now people build constantly and end up with fifteen half-finished tools that solve problems nobody had. The people winning with AI aren’t building more. They’re picking better.

01

The one-sentence rule

Before you build anything, write one specific sentence naming what it will make or save. "This saves me ___ hours a week." "This makes ___ dollars." "This stops ___ from happening again." If you can’t finish it, don’t build it.

It takes twenty seconds and kills about half your ideas. Every idea it kills is a weekend you get back.

Passes: "This saves me the four hours a month I spend rewriting the same client update." Or: "This catches the pricing mistake that cost us a customer in March."

Fails: "This will make my workflow better." "This would be really cool to build." "Everyone is building agents right now."

One more filter: you should have felt this problem in the last seven days, by losing time to it or apologizing for it. Problems you haven’t felt lately are usually ones you invented for the idea you already wanted to build.

Not sure? Make AI argue with you. This prompt tests the problem instead of approving the idea.

The Pick Test
Role: You are my strategy partner, and your job today is to talk me out of this. Do not be encouraging. Do not help me plan it. I want the problem tested, not the idea approved. Here is what I am thinking about building: [describe it in a few sentences] Here is who it is for: [me, my team, my customers, be specific] Run these six questions on it, one at a time, and answer each in two or three sentences: 1. In one sentence, what does this make or save, in hours or dollars? If I have not given you enough information to answer that, say so and stop there. 2. What is the real cost of this problem today, per week? Show your math using only numbers I gave you. Do not invent numbers. 3. How am I solving this right now, and what is actually wrong with that? If the current way is mostly fine, tell me. 4. Is this a problem I have felt in the last week, or a problem I am imagining? Ask me for evidence before you decide. 5. What is the smallest version that would tell me if this is worth it, and could I test it in under an hour? 6. What would have to be true for this to be a waste of my time? Give me the three most likely ways this ends up unused in a month. Then give me one of three verdicts, and commit to it: BUILD IT, with the one sentence I should have written. TEST IT FIRST, with the smallest test and what result would greenlight it. DROP IT, with the specific reason. Rules: Do not soften the verdict. Do not give me a fourth option. If I push back, only change your answer if I give you new facts, not if I just sound disappointed.

Practice This Week

Write the sentence for everything you build for seven days, even small things, in one note. At the end of the week, read them back. The ones you couldn’t write clearly are the ones you shouldn’t have started.

2. The Check

AI output is confident, clean and well formatted, which is exactly what your brain uses to decide something is correct. That’s why bad output slides past you. AI is much better at looking right than being right.

02

Change one thing

Every time AI hands you something, find one thing to change before you accept it. One word, one number, one step. Every time, even when it looks perfect.

It sounds pointless. It’s the most useful habit in this guide, because you can’t find a change without actually reading. Skimming is where bad output gets through. If you truly can’t find anything, write down why it’s good. Same rep, other direction.

Then run three checks on anything that matters.

  1. Trace every number. If it didn’t come from something you gave the AI or something you can open and look at, it isn’t a fact yet.
  2. Find what it assumed that you never said. Budget, audience, timeline, tone. Ask it to list every assumption. Expect some of them to be wrong.
  3. Ask what would make it wrong. That flips you from grading into thinking, and it surfaces the edge case you were about to ship.

Run them in a fresh chat, or better, a different model. A model defending its own work is about as reliable as you proofreading your own email at midnight.

The Cold Check
You are a skeptical editor checking work that someone else wrote with AI. You did not write it, and you have no reason to defend it. Your goal is to find what is wrong before it ships. WHAT THIS IS FOR: [who will read or use it, and what they need it to do] WHAT I GAVE THE AI TO WORK FROM: [paste or describe the inputs, facts and numbers you provided] THE OUTPUT TO CHECK: [paste it here] STEPS: 1. Numbers. List every number, date, name and quote in the output. For each one, say whether it came from the inputs I listed, or whether it is unsupported. Mark unsupported ones "VERIFY." 2. Assumptions. List every assumption the writer made that I never stated (budget, audience, timeline, tone, tools, policies). Be specific. 3. What would make it wrong. Name the 3 most likely situations where this output would fail or mislead someone, including edge cases (an unusual customer, a missing file, a deadline already passed). 4. The finish. Point to the 3 spots that most sound generic or AI-written, and say what kind of detail only I would know that belongs there. Don't write it for me. OUTPUT FORMAT: - Four short sections with the headers above, as bullet points. - End with one line: SHIP, FIX FIRST, or REDO, and the single biggest reason. GUARDRAILS: - Don't rewrite the whole thing. Point, don't polish. - Don't invent facts to replace the ones you flag. - If something is fine, say so in one line and move on.

The Limit

The check only works on topics you know something about. Without some footing in the topic, a prompt can’t save you. Your job then is to find one real source, or one real person, who can check it.

3. The Finish

AI gets you about 80 percent of the way, fast. That last 20 percent is the difference between work that’s obviously AI and work that’s obviously good. The 80 feels almost done, so people ship it and wonder why it looks like everyone else’s. Everyone has the 80. Only you have the 20.

03

Your Last 20% list

Open a note called Last 20%. Every time you finish something AI helped with, write down what you had to fix yourself.

Within two weeks you’ll see the same six or seven fixes every time. AI fails in the same places in your work, because your work has a shape it doesn’t know. That list becomes your checklist, and it’s worth more than any prompt you’ll save.

A starter list to react to. Keep what’s true for you and toss the rest.

  1. Edge cases. AI writes for the normal path. It forgets the customer with two accounts or the refund after the deadline.
  2. Tone. Right meaning, wrong sound. Read it out loud and cut anything you’d never say.
  3. Formatting. Headers and bullets where a person would write a paragraph.
  4. The detail only you know. That one client hates phone calls. That number is seasonal. Usually the most valuable sentence in the piece.
  5. The stakes. AI writes every line at the same temperature. You know which line has to land.
  6. The cut. The 80 is almost always too long. Deleting a third is often the whole finish.

Do This Once

Take something you already shipped and rebuild it with AI from scratch. Put the two side by side. Everything in your version that’s missing from the AI version is your last 20 percent. It takes about fifteen minutes.

The Two Habits

The drills build the skill. These two habits make it automatic. Give each one two weeks.

  1. Change one thing, every time. It forces you to read, and reading is where the check comes from.
  2. Log your Last 20%. It turns your finish into a checklist you can run on everything.

Pair them with the one-sentence rule before you start anything, and you’re practicing all three parts of taste every single day.