A special serving by BLOCK71 Indonesia and Innovation Factory, in collaboration with BelajarGPT and AI Connect.
On Thursday, 16 July 2026, BLOCK71 Indonesia and Innovation Factory hosted the July 2026 edition of KopiChat, bringing founders and builders together for an honest conversation about how startups are really using AI agents today. Joined by speakers from BelajarGPT and AI Connect, the session tackled a question every founder is asking right now: when AI can build almost anything, what actually gives one startup a competitive advantage over another?
The panel featured William Jakhfar, CEO & Founder of BelajarGPT, and Mizan Lazuardi, Programme Lead at AI Connect, with BLOCK71’s Programme Manager, Andit Intasworo, as moderator. Here are six key takeaways from their discussion.
Building isn't the hard part anymore
One of the biggest shifts discussed by the speakers was that building a product is no longer a startup's biggest challenge. With AI, almost anyone can put together a working prototype in a short amount of time. In this modern age of AI, what actually sets a startup apart isn't the technology, it's whether the team is solving a real, specific problem instead of a generic one.
The panel also agreed that success isn't only about having a good product. As William put it, “if everyone can build the same product, why would customers choose yours?” Trust, branding, and strong execution matter just as much as the product itself. A great feature won't count for much if customers don't trust the company, or if the team can't deliver consistently. Better prompts and better AI output also depend on something AI can't generate on its own: strong business context and a real understanding of the customer.
Splitting the work between AI and humans
This is why the speakers kept coming back to how AI and humans should divide the work. Mizan stated how, “AI accelerates the process of thinking, but execution remains a human capability.” Let AI handle the repetitive tasks, automation, and first drafts, and let humans handle the parts that require judgment: decisions, creativity, and relationships. AI is excellent at producing ideas, but it can't validate them, sell them, or push them through to completion. That responsibility still sits with the founder.
The real challenge isn't building AI, it's integrating it
Mizan challenged the idea that building AI models is still the hard part. In his view, the bigger challenge today is integration. How to connect systems, get different tools to communicate, design proper governance, and build architectures that can actually scale. Scaling AI, he argued, requires thinking beyond the model itself and focusing on system integration.
He also emphasized starting with the problem, not the technology: understand why the product needs to exist and who the user is before adding AI into the mix. AI should solve a real user problem, not exist for its own sake.
Distribution is the new moat
This shift is also changing where startups should focus their energy. Since AI has made the technology easier to copy, the panel argued that marketing, audience building, and customer acquisition matter more than ever. If everyone can build the same thing, the winners are whoever can reach customers first.
William broke this down further on the sales side: AI agents are efficient for high volume, repetitive customer service, but they still struggle with context and meaningful follow up. Humans, he noted, remain better at generating sales and understanding customer nuance, so closing a deal and earning someone's trust still takes a human touch.
Where AI still falls short
Even with all these advantages of AI, the speakers was honest about where it still falls short. When asked how much they trust AI's output, both speakers landed around 50 percent. Useful, but not something you can rely on blindly. Human oversight still matters, especially for decisions with real consequences. Authenticity, storytelling, and genuine emotional connection also came up as areas where AI isn't there yet, and probably shouldn't try to be.
William was direct about the risk here: “vibe coding” with AI speeds up development, but relying too heavily on AI generated code without understanding it can introduce security risks. Good software architecture still requires real engineering thought, moving fast is great, until it isn't, a codebase that can't hold up under pressure will cost you later.
The takeaway for technical founders
The speakers wrapped up with a simple reminder for technical founders: knowing how to build isn't enough anymore. Founders also need to understand that human touch still matters greatly. In a world where AI makes building easy, the founders who win are the ones who know how to reach people.
KopiChat is BLOCK71 Indonesia's ongoing series of honest, community driven conversations for founders and builders. Want in on the next session? Follow Innovation Factory & BLOCK71 Indonesia for more event updates & insights.