If I'm starting an AI automation agency today from a third-world country (Tunisia), with extremely limited capital, zero technical skills, no clients, and no portfolio, what niche and offer would you advise, and what should I learn or build first? Are niches different in developing nations versus developed ones?

Yes, optimal niches for a market like Tunisia will be very different from the optimal set of services for, say, North America. But you don't have to sell to people in your local target country — that's really the whole promise of the internet: you can sell into a market that's already well-established with clear, well-understood solutions to needs, the kind talked about and demonstrated on channels like mine. There's nothing that necessitates selling to your local market. If you're in Bangladesh, you don't have to sell only to your local Dhaka market — you can go global. The tradeoff as you go further away is cultural norms and time zones become less familiar. As long as you control for those two variables — either by offsetting your own schedule, or by consuming a lot of media from your target country to understand how people there think, which, keep in mind, is quite similar to how people in your own country think (human beings are human beings, 99.99% the same DNA, just some superficial design tweaks like CSS on an HTML website) — you'll be fine. You can follow the exact standard playbook and achieve good results wherever you are.