For a while, AI in my business looked like everyone else's: I'd open a chat, ask a big vague question, get a big vague answer, nod, and go back to doing the work by hand. It was a novelty I visited, not a tool I used.
The shift came when I stopped treating AI as an oracle and started treating it as a teammate with a job description. The turning point was a small idea: instead of re-explaining my context every single time, I built skills — reusable instructions that carry my standards, my stack, and my market into every task automatically. That one change is what turned the hype into a habit. Here's how it actually plays out, on both sides of what I do.
The business side
Most of my week isn't code. It's content, positioning, and getting the right people to notice the right thing. AI does the heaviest lifting here — but only because I've taught it how I think, not just what I want.
Writing content and blogs. This blog is the example. I don't ask for "a blog about X" and paste whatever comes back. I have a writing skill that knows my voice — problem first, no throat-clearing, real examples over abstractions. So the output starts as a draft in my register, and I edit from there instead of fighting a generic one.
Marketing. I built a campaign skill that already knows my defaults — WhatsApp Business and LinkedIn as primary channels, the Zimbabwean context baked in (dual currency, WhatsApp-first buyers, trust-sensitive markets). "Help me market this" produces a real, structured plan I can execute, not a listicle of tips I've read a hundred times.
Lead generation and outreach. Drafting cold messages, shaping a funnel, turning one blog into a week of LinkedIn posts, writing the WhatsApp follow-up that doesn't sound like a robot. The grind of consistent output is exactly where a tuned assistant earns its place.
The pattern on this side: AI removes the friction between having an idea and shipping it. The bottleneck was never ideas. It was the hours between idea and published thing.
The code side
On the engineering side the discipline is different, because bad code is worse than no code — especially anything touching money, identity, or health data, which most of my products do.
Writing and reviewing. I move fast on React/Next.js and Tailwind, so AI scaffolds components, drafts API routes, and writes the boilerplate I'd otherwise type in my sleep. But the real value is the second pass. I built a code-quality skill that reviews for the things I'd catch on a good day and miss on a tired one — edge cases, error handling, security holes, missing tests — with extra scrutiny anywhere money or personal data flows.
Architecture and thinking. Before building, I use it to pressure-test decisions. Talking through a schema or a data flow with something that pushes back is faster than doing it alone and finding the flaw three sprints later.
The pattern here is the inverse of the business side: AI is a force multiplier on judgment, not a replacement for it. It writes more, so I have to review better. The skill is what makes the review consistent instead of dependent on my mood.
What actually made it a habit
The honest answer isn't a clever prompt. It's three small commitments:
The Habit
- Give it your standards once, reuse them forever. A skill beats a perfect prompt you have to remember to type. The context travels with you.
- Give it a role, not a wish. "Review this for security and edge cases" gets a useful answer. "Make this better" gets mush.
- Stay the editor. On content, I ship my judgment, not the model's first draft. On code, I own every line that goes to production. AI drafts; I decide.
The hype told me AI would do my work. The habit is quieter and more useful than that: AI does the first 70% of a lot of things, fast, in my voice and to my standards — so I spend my time on the 30% that actually needs me. That's the whole trick. Not magic. Just a teammate I finally bothered to train.
