The piece breaks down the economics of replacing call center workers with AI, focusing on the voice AI industry. It categorizes companies into three groups: foundation model companies like OpenAI and ElevenLabs that develop the underlying technology; pipeline companies which connect these models and help businesses implement them; and vertical startups that create voice agents tailored for specific industries such as healthcare and logistics. This structure illustrates the complexity of the market and highlights that many companies may straddle these categories. Using Vapi as a case study, the article details the AI process involved in handling calls. It outlines the steps from speech-to-text conversion by Deepgram, to text processing via GPT-4o, and finally converting text back to speech with Vapi’s TTS. Each step has specific latency and cost, totaling a minimum of $0.15 per minute for usage. Comparatively, it provides an overview of call center salaries across various countries, revealing that AI solutions can be significantly cheaper than human labor. For example, the average hourly rate for a call center worker in Egypt is about $1.31, while Vapi’s offerings can exceed $5.40 per hour, although they are still competitive with higher-cost countries like Canada and Poland. The article notes potential limitations in enterprise contracts, which might differ in pricing based on usage and commitment, but lacks concrete data on these arrangements. Testing of other platforms was limited, particularly due to bugs in Bland, indicating that the reliability of these technologies can vary. Overall, the analysis highlights the financial incentives for businesses to adopt AI solutions over traditional outsourcing while pointing out the nuances in implementation and service quality.