How a Claude Skill Converts Leads With a Quiz

I just finished the design of a new Claude Skill and went to put it to work. So I gave it this brief.

It was about a page of plain text — no code, no diagrams, no special template — describing the problem domain I wanted it to handle: how companies adopt AI, and where they get stuck.

Some companies, the brief said, tried AI a couple of years ago, weren't impressed, and quietly decided it can't do their work. Others bought every employee a subscription and watched nothing change. Some people are sure the trick is a magic phrase you type into the box. Some pile a year of conversation into one endless chat. Some point AI at the flashy problem that shows up twice a year instead of the boring task that eats six hours a day. And under all of it, people are afraid — of being replaced, of looking inauthentic, of doing it wrong and freezing, or of running ahead and torching the budget in a month.

The brief ended with a simple ask: an assessment that tells a company where it actually is with AI, and what to consider doing next.

A few minutes later I had two things back. A clear design document that laid out the whole framework in language I could read and correct over a Diet Coke. And an actual working assessment — a real interactive quiz, ready to host and make available on a website, that asks thirteen questions and hands back an honest, personal read.

That round trip — plain-language brief in, finished diagnostic out — is the thing I want to show you today.

Quick recap, in case you missed Thursday

On Thursday I made the case that everyone needs a custom assessment. Not everyone everyone — but every domain expert whose customers are non-experts.

The argument was the mall map. Walk into an unfamiliar mall and the first thing you hunt for is the red dot: you are here. Until you find it, no map helps and no directions make sense. Your customers walk in exactly that lost — they haven't done the work you've done for a decade, and they don't have the words to place themselves. So an expert who opens with advice is reading turn-by-turn directions to someone who doesn't know which floor they're on.

Dot > Map > Directions. That's the whole thesis in four words.

I've been building these assessments for three decades. They used to take me anywhere from a few months to over a year. Over the last three months I've built four — the shortest in three days, the longest maybe two weeks. It still blows me away.

What changed is the part I want to walk you through: I stopped building each one by hand and built a skill that builds them. So let me explain what that even means.

What a "skill" actually is

Most people have used an AI chatbot for a clever one-off: draft this email, summarize this report. Useful — but it's a single shot. Ask the same thing next week and you might get something shaped completely differently.

A skill is the opposite of a one-off. It's packaged expertise — a repeatable process Claude follows the same careful way every time. Think of it as handing Claude a seasoned specialist's playbook: not just what good work looks like, but the exact steps, the order to run them in, the traps to dodge, and the standards the finished product has to clear before it's allowed out the door.

I called mine Red Dot, for obvious reasons. Its one job is hard: take an expert's hard-won judgment and turn it into a diagnostic their customers can actually use.

The clever part: the gap, and the honesty

Finding the dot is only half of it. The thing that makes someone want to move is the distance between where they are and where they could be. The skill calls that distance the gap and treats it as the real product. A dot with no gap is trivia. A visible gap is motivation.

Two choices keep that gap honest, and they're what separate this from every rigged quiz you've ever bounced off of.

First, the questions ask what actually happens in your company — never how good you think you are. Not "rate your AI skills" but "when something doesn't work on the first try, what usually happens next?" People can't reliably rate the very thing they'd hire an expert to see. And a behavior question hides its own "right" answer, so nobody can game their way to a flattering result.

Second — and this is the line the whole thing lives or dies on — the dot has to be honest even when that's inconvenient. The internet is full of "assessments" that are really email-collectors with a foregone conclusion: surprise, you need our thing. People smell it instantly, and the second they do, your credibility runs in reverse. So Red Dot is built to sometimes conclude you're doing great, keep going — even though that sells nothing. The honesty is the asset.

What came back

For the AI brief, the skill drew a map with four places a company can land — named so a leader will say them out loud:

The Operator — genuinely skilled and spreading it across the org. The rare healthy state.

The Lone Champion — one or two people are great with AI, but the company can't absorb what they know. Skill with no system.

The Stocked Toolbox — real money spent, licenses for everyone, maybe a mandate — and thin actual skill. A system with no practice.

The Onlooker — still watching from the sidelines, usually because an early experiment underwhelmed.

Then it added a sharper second gap: the distance between how much a company has invested in AI and how it actually works day to day. That's the quiet pain behind "we've paid for this for a year — why are we still stuck?" The assessment computes it and names it, instead of leaving the leader to feel it with no words for it.

And it handed me all of it twice: once as a plain-English design document I could mark up and correct, and once as a finished, shareable assessment that runs the moment you open it.

Why building it as a skill matters

The obvious win is speed — months of work compressed again, into an afternoon.

But the deeper wins are quieter.

It's repeatable. This isn't one lucky output. The same playbook works for a marriage counselor, a leadership coach, a financial advisor — anyone whose customers need to know where they stand before they can take advice.

The expert stays the expert. Claude doesn't invent your knowledge. It manufactures the instrument and hands you back a draft to correct — and correcting a draft is far easier than starting from a blank page. The judgment that can't be automated — which gap to lead with, what to name the four types, what's actually true — stays firmly with you.

And it won't ship something broken. Before Red Dot releases an assessment, a built-in checker refuses to publish one that leaves a question unanswered, makes a claim it never measured, or quietly slides into "buy the thing." Every sentence a customer reads was written and approved in advance. Nothing is improvised in front of them.

What this won't do

Let me be straight about the edges.

What I've built here is an automated quiz builder. It's great for a sharp little diagnostic — the kind that places a dot, names a gap, and works as a lead-gen front door to your expertise.

It is not a rich, deep, complex assessment. I build those too, and I use AI for those as well — but that work can't be compressed into a simple skill (yet). The depth still needs me in the loop in ways a packaged playbook can't replace.

Come along

I told you Thursday I'd build one of these in the open. This skill is how I'm going to do it — and I'll show you every prompt, every guideline, every agent, and every bit of code that comes out of it.

The expertise was always yours. I just built the machine that delivers it.