Since Sharma Brands® has sold, I’ve been DEEP in the AI world. At first, it wasn’t to build anything particular, I just wanted to see what all is even possible; where are the guardrails, and what can I start to see as patterns and then how can I get better at prompting and using the tool, in the first place.
A fun/free site I put up a couple weeks ago was DTC101.com — I created a network of agents to read through my newsletters, extract facts, organize them by category and for 3 main roles at a company, then build a brand book, design the UX/UI of the website, develop it, and maintain it. As my newsletters continue to send out, the site will automatically update.
Another one was Zikes.com — since ambulance-chasing, scum-of-the-earth attorneys go after eCommerce brands with frivolous ADA claims for settlements, I thought, why not get ahead of these issues and be aware of what could be coming? Now anyone can run the same tests those attorneys run.
4 years ago when starting Character, I paid an agency $10k to build me fully custom brand identities, then develop them into performance-friendly Wordpress themes so I could drive new customer acquisition with advertorials. The process took 3-4 weeks PER publisher. I did the same thing with a new publisher, MaeveMag.com, and while the site isn’t complete yet, its taken me a total of 8-10 hours to build so far — visual identity, custom theme, customizing the site, building a couple of custom WP apps, 200+ properly-formatted articles, etc.
This is only a fraction of the stuff I’ve been able to build. Custom reporting dashboards. Better tools to filter through CRM lists. A creative brief generation bot inside Slack. A shared knowledge base for a brand.
Even deeper than that, building training documents for different AI tools to use. Whether I’m using Claude (the app), OpenClaw (or ClawdBot), OpenAI, Gemini, N8N, Higgsfield, Kiva, Manus, Figma, Replit, or anything in between, I have a bible of training documents I can upload and contribute/improve anytime, so everything feels/sounds/looks consistent and uniform.
While half of the internet screams loudly at the downsides of AI, even my mom is learning how to incorporate Cowork in her life. If you’re not becoming an adopter and spending even 30 minutes a day playing with AI tools, you will get left behind.
Here’s an analogy you may find familiar. In 2014, when you ran Meta ads, everything was manual. You had placement choices from right-hand-rail (back when RHR was programmatic-network inventory) to being able to select audiences of people who were going to take a specific airline and go to a specific city. When something didn’t work, you’d go in and find where in the setup you broke something. Today... all of this happens behind-the-scenes, no one chooses anything really. That’s why 2014 media buyers know the platform better, understand the “algorithm” more and can diagnose when things aren’t working or scaling, better than anyone else. They’ve done it the hard way, they’ve seen and learned the patterns, they know how the broken thing is not working and where to be able to fix it.
That same behavior/pattern what is happening with AI too — every two weeks when another model comes out, it becomes easier to use and more powerful, which is great, but you want to know how its working BTS. That’s why you need to be using it consistently, all the time.
Will AI takeover everything and nothing feels real again? No. But like with anything else, we’ll see drop-shippers and vice-industry brands use it first, and then as it becomes more normalized, bigger brands will adopt it faster. I imagine by EOY, 80% of ad accounts will be using AI creative. Tools like Marpipe, Higgsfield, Butter, even TripleWhale’s Moby all incorporate AI models and workflows in their apps. They are the interface, and leverage different model’s APIs to create the product.
In just a few months, you’ll likely just use the model’s interfaces directly — whether it’s Gemini, Kling, Wan or something else.
Everyone has already most-likely tried to use tools like ChatGPT for writing ad copy or analyzing raw data, but here are a few things I encourage you to try doing (all take under an hour):
Build a reporting tool for yourself. It doesn’t matter what you’re building it for (creative analytics, media reporting, internal task analytics, logistics/shipping dashboards, etc.), but this will walk you through the process of design, development, connecting sources to one-another, and gives you something to use on a daily basis, fully custom to your needs. Previously you’d have to use tools like Whatagraph, now you can build it in <60 minutes.
Build a landing page with AI. Whether you use Figma’s AI, Replit, Cursor, Lovable, or Claude Code, going through the process of design, development, writing on-brand copy (or listacles or advertorials), incorporating research, curation and creation of content, and integrating the domain gets you setup to build repeatable landing pages. As you get deeper, you can customize more — the design, the animations, the flow. Dynamic pricing based on a user’s location or their real-time weather, all within seconds.
Start building your second brain. Connect/integrate tools directly or create automations and flows to export files and upload them to your AI model of choice. I find Claude to be better for this, by creating projects, and using their connectors (plus, Claude is like your co-worker/employee, whereas ChatGPT is your buddy you go get drinks with or ask questions to). Integrate your call recorder app to keep it updated with your conversations/meetings. Give read-access to your email, calendar, drive, Dropbox, and more. As much context you can feed it, the more it can work for you. When you leave a meeting where you discussed a new promotion, how it would flow and its next steps, but need to gather past promo data to benchmark against it, you can run that with one Claude prompt, because it has all of it already.
Create markdown (.md) files with your rules. Markdown (.md) is a file format similar to plain text, or when you see a .txt file, except the AI models love it because it incorporates some basic formatting. The more you can create .md files with your own rules, the easier and more cohesive things stay. Remember a few weeks ago, my newsletter was about “The Cohesion Tax”; it only gets more important with AI. Here are examples of .md files you can create and continue to update (using AI) so no matter what tools you use, they feel in sync:
- brand-voice-guide.md
- copywriting-rules.md
- brand-story.md
- founder-story.md
- product-stories.md
- positioning-statement.md
- messaging-hierarchy.md
- taglines-and-phrases.md
- ideal-customer-profiles.md
- objections-and-rebuttals.md
-
email-playbook.md
- social-playbook.md
- ads-playbook.md
- product-catalog-brief.md
- promotions-and-offers.md
- cx-voice-and-macros.md
- policies-summary.md
- creator-bio-pack.md
- personal-rules.md
- content-frameworks.md
- brand-strategy-guardrails.md
-
ethics-and-redlines.md (this can even be your brand's universal rules of what to look for in vendor agreements to draw redlines, so everyone in your organization can skip the step of sending things to legal, until final review)
4 things to start getting your juices flowing with AI. Why is a whole newsletter (that should be tactical) dedicated to Nik ranting about WHY you need to jump on the AI train? Because I don’t want you to get left behind, and when I talk to brands today who are doing $5M, $50M or $500M, rarely are they leveraging AI in the ways its available today. And with the 4 things I mentioned here, you’ll go down pathways and rabbit holes that start to show you what is possible.
If you’re building something cool with AI, email me what it is, I want to see it (either reply here or email [email protected]). And if you want to know how the early adopter brands are using AI to build creative teams, ad-ops teams, or CX teams and more, make sure you come to the AI Summit in June.