How I built a brand skill in Claude (so my decks stop looking generic)

Wispr Flow turns your voice into clean, professional text you can send the moment you stop talking. Not rough transcription you have to clean up. Actual polished text — ready for email, Slack, or any app.

Speak the way you think. Go on tangents. Change your mind mid-sentence. Flow strips the filler, fixes the grammar, and gives you text that reads like you spent five minutes writing it.

89% of messages sent with zero edits. Millions of professionals use Flow daily, including teams at OpenAI, Vercel, and Clay. Works on Mac, Windows, and iPhone.

Start flowing free

Client work spiraled, some questions about where this newsletter is going got loud in my head, and the result was that I didn’t publish.

In the meantime, I’ve been building a lot of cool stuff with AI and feeling a bit guilty for not sharing it, so I’m pushing through the inertia and getting something out.

I’ll start with one I should have shared three weeks ago.

A few weeks ago, I was prepping a free workshop on becoming an AI-enabled consultant, and I needed a deck.

In the “old” days that mean building something myself in Power Point or Google Sheets.

Since my design skills are, let’s just say, minimal, most of my decks were what you could at best call minimalist. But really, you’d say ugly.

I got slightly better when I tried Gamma.

Then I met Claude.

If you ask Claude to make you a PowerPoint, unless you are some kind of PowerPoint design maven, it will build you a deck multiple times better than anything you could have made.

And if you combine that with a brand skill, then it can build on-brand decks using your logo files, fonts and colors.

Now, I had logo files, fonts and colors. But the logos weren’t that strong, and the different brands I’ve created (personal, two newsletters, and my consulting business) weren’t aligned.

I always love a good excuse to do some productive procrastination, so I figured before I go all in on building a brand skill, which I needed to make my PowerPoint, I’d rebrand.

Don't overthink it. Which is like telling a duck not to swim.

But I figured anything Claude could come up with would be better than what I had.

So I started throwing stuff in. The colors I've been using for the past six years. My existing logos, etc.

Since Claude already knew everything about me and what I do, I was able to jump right in.

I told Claude I wanted to unify my branding so everything felt coherent, and asked it for some ideas.

We had the usual back and forth, and I saw something I liked.

I didn’t even recognize at first that it represented a keycap on a keyboard, but Claude had created a complete concept behind it:

That pretty much sold me. And I liked how the design worked for multiple brands:

Once we landed on a direction, we built out color variations, banners, etc.

Then full brand guide that explains the colors, typography, voice rules, and when to use which.

Lastly, I turned this into a skill so I could use it to create presentations and decks using my logos and colors.

Without this, every time you want to use your branding, you’d have to upload your logos and brand book and rexplain how you like things.

With it, every future "make me a slide" or "design a banner" call automatically uses your brand.

Claude has a skill-creating skill from Anthropic (you can check if it’s turned on by going to Customize > Skills and looking for skill-creator in your list).

All you have to do is tell Claude something like “now turn this in to a brand skill” and it will do the work.

This was the hardest part.

Not because of the content, but because of the iteration it required to get what I wanted.

Even with the skill, the first dozen versions of the PowerPoint Claude created weren’t want I wanted.

Eventually, at version 15, we had things to look good, and at that point, I told Claude to update the skill with what we had and called it good enough:

I felt proud of it, and should have stopped.

I didn't.

I went back in the next day to "polish the colors." Six rounds of palette iteration. An hour and a half. The deck didn't get materially better. I just got more frustrated. At one point I literally typed "what is hard about getting this right? what should have taken 5 minutes is now at an hour plus" into the chat.

The frustration only ended when I stopped describing what I wanted in words and gave Claude a clear visual example of what I wanted instead. Two minutes after that screenshot landed in the chat, we were done.

In the weeks since, I’ve made several more decks with the skill, and each time I give feedback about what’s working and what’s not so it gets better.

The deck I made this past week required only a few minutes of editing to be ready to go.

AI will let you tweak forever. Your brand skill is a tool, not a perfectionism machine. Good enough is good enough. Ship it.

When you must keep iterating, stop describing and start showing. Words don’t always land. Screenshots can be easier for the AI to understand.

(Btw, I wrote a whole issue about that polishing trap a couple weeks back: When AI Polishes Turds.)

If you've been making decks or one-pagers in AI and they keep coming out generic, the fix isn't a better prompt.

The fix is teaching the AI what your brand is, once, and packaging it as a skill so you don't have to teach it again next time.

Pick one upcoming deliverable. A deck, a one-pager, an email banner.

Open Claude in whatever flavor you use most. Drop in whatever brand assets you have, even three colors and a logo you don't love. Ask it to help you build a brand skill you can call later.

You won't get it perfect in one sitting.

You'll get something usable. The next deliverable will look better than the last one. The one after that will look better still.

Know someone whose decks look decent but never quite feel like their brand? Forward this. The skill is the missing piece.

Thanks for reading!

Nathan Rodgers

👋 Say hello on Substack and LinkedIn

Were you forwarded this email? Click here to subscribe.