Dublin-based software engineer Maahir Sharma has successfully demonstrated that autonomous AI agents can navigate real-world telecommunications to secure hotel accommodations. In a series of tests conducted in May, Sharma deployed a custom-built AI phone agent to contact hotels across the United States, handling availability inquiries and price negotiations without human intervention. The project, which ran between midnight and 3:30 a.m. Dublin time to align with US hotel operating hours, serves as a practical validation of generative AI capability in live voice conversations. The architecture combines Cursor for development, Bland AI for voice synthesis and telephony, Google Places API for hotel identification, and an OpenAI language model to refine conversational prompts. Sharma configured the system with specific travel parameters, including dates and budget constraints, instructing the agent to request direct booking rates and explore discount flexibility. Initial versions exhibited robotic phrasing and inefficient negotiation tactics, but over a week of iterative prompt engineering, the agent dialogue shifted to match natural human conversation patterns, significantly improving its success rate. During live testing, the agent engaged multiple hotel representatives, one of whom immediately questioned whether the caller was AI. The agent successfully deflected the inquiry while maintaining its negotiating posture. Across several trials, the system secured a five-dollar nightly discount at one property and identified a cash-payment incentive at another. Although Sharma did not finalize any bookings, the experiment confirmed that the agent could handle unexpected interruptions, adapt to human counteroffers, and extract measurable concessions. The broader objective extended beyond trivial cost savings. Sharma aimed to stress-test whether autonomous agents could manage time-intensive booking workflows, a capability with significant implications for travel technology and customer service automation. His development process underscores a critical industry insight: marginal adjustments in prompt design often yield greater performance gains than architectural overhauls. Sharma approach reflects a growing trend among technology professionals who dedicate substantial personal time to hands-on AI experimentation. With a background spanning Meta, Zomato, and current roles at major technology firms, he emphasizes that continuous practical engagement with emerging AI tools is increasingly essential for career resilience. He notes that barrier to entry for such projects is lowering rapidly, citing the rise of no-code AI development environments that enable non-engineers to deploy functional conversational agents. The hotel-negotiation experiment ultimately validates both the technical maturity of current voice AI systems and their expanding utility in automating routine commercial interactions.