In this episode of Appian Rocks, Stefan, Sandro, and Marcel tackle the controversial role of artificial intelligence in process implementation projects. While acknowledging AI’s impressive capabilities, they warn against the industry’s tendency to treat it as a universal solution. What demos well in sales meetings often falls short in practice, producing answers that only sound competent. The hosts argue that uncritical adoption leads to laziness, outsourcing of judgment, and a dangerous decline in deep problem-solving skills. Marcel frames the issue as the “hammer and nail” problem: with AI marketed as the hammer, everything starts looking like a nail. This obsession can stifle thoughtful analysis and push teams to skip the hard work of understanding processes. Stefan illustrates this with a client case where rethinking and simplifying steps—without AI—halved the workload. The real benefit came not from automation but from owning the thinking and redesign. If a team relies on a chatbot instead, it risks losing both control and learning. Still, the hosts emphasize that AI has valuable use cases, particularly where input is noisy or unstructured. Summarizing long documents, extracting fields from messy scans, or parsing communication are areas where probabilistic language models excel. But when data is already structured and clear, adding AI can actually reduce quality. As Stefan puts it, “the best part is no part”—if a step adds no value, eliminate it rather than overengineering with AI. The conversation then broadens to the societal and environmental costs of AI overuse. Marcel highlights the immense energy and water consumption of data centers, noting that a single AI query is vastly more resource-hungry than a standard Google search. Sandro compares the phenomenon to refrigerators: once they became widespread, people stopped considering older preservation methods and even began misusing fridges for foods that spoil faster inside them. Likewise, if developers only learn to solve problems through AI, they may never develop alternative methods, filling the industry with people who know no tools beyond the “fridge.” The panel also warns about economic risks. Current AI feels cheap because of heavy investment subsidies, but providers will eventually move to value-based pricing, charging for “man-hours saved.” This could trap organizations in costly dependencies once AI is deeply integrated into core processes. Consultants, they argue, must therefore frame adoption not only around use-case justification but also total cost of ownership, including volatile token-based pricing. In closing, the hosts underline that AI should be one tool among many. Its convenience is undeniable, but convenience alone is no justification. In low-code environments like Appian, the temptation to lean on AI for speed is strong, yet true transformation still requires creativity, critical analysis, and ownership of solutions. Overuse risks fragile systems and a loss of craft. For now, they agree: AI is powerful and promising, but it must be applied sparingly, thoughtfully, and only where it adds real value.