With the rapid development of conversational AI technology, AI voice agents are gradually taking over an increasing number of phone communications. However, ensuring the reliability of these AI voice agents has become a significant challenge for the industry. Recently, Hamming.ai, a platform focused on the reliability of AI voice agents, announced the completion of a $3.8 million seed funding round, led by Mischief, with participation from Y Combinator, AI Grant, and several angel investors.
The emergence of Hamming.ai is aimed at addressing the current pain points in testing and managing AI voice agents. It is understood that billions of phone calls are made every day, and as AI advances, most of these calls will ultimately be handled by AI. However, even minor prompts or changes from model providers can lead to significant variations in the responses of AI voice agents. Currently, engineers spend a considerable amount of time each day manually testing their AI voice agents, a method that is not only inefficient but often lacks comprehensiveness. Even after the AI voice agents go live, operational teams still need to listen to thousands of calls to identify edge cases that manual testing might have missed. These issues not only lead to high setup costs for AI voice systems but can also trigger liability issues and negative public opinion due to insufficient testing infrastructure.
Hamming.ai effectively addresses these issues through automated testing, monitoring, and management of AI voice agents. The company has deployed its self-developed AI voice agents, which can make thousands of calls simultaneously to test clients' AI voice agents just like a real person. In addition, Hamming.ai provides B2B teams with LLM prompt management solutions, automated red team testing for AI voice agents (to detect vulnerabilities), and call analysis solutions to track user interactions with AI voice agents and flag cases that require attention. It is claimed that Hamming.ai's approach is 20 times faster than manual testing and reduces costs by a factor of 10. The company plans to continue enhancing these advantages through further product development and refinement.
Hamming.ai's co-founder and CEO Sumanyu Sharma met co-founder and CTO Marius Buleandra at Citizen, a personal safety network company backed by Founders Fund. They both have extensive experience in building trust and safety infrastructure at Citizen. Sharma previously served as the data chief at Citizen, helping the company triple its user base. Before that, he was responsible for an AI-driven sales project at Tesla that increased its annual revenue to hundreds of millions of dollars. Buleandra has accumulated rich experience in data infrastructure, AI, and complex systems engineering at companies like Anduril, Square, and Microsoft. He was also a founding engineer at Spell, a machine learning observability and infrastructure startup that was later acquired by Reddit.
Lauren Farleigh, co-founder and general partner at Mischief, stated: "Conversational AI is rapidly evolving, but most testing and management tools have not kept pace with the needs and compliance requirements of developers. Hamming.ai will become a cornerstone in ensuring the safe development of this technology, as AI reshapes the interactions between businesses and their customers."
Hamming.ai believes that if 2024 is the year of prototypes, then 2025 will be the year of reliability. Industries such as healthcare, legal, insurance, and real estate are bound by compliance frameworks that govern how businesses interact with customers and share information. As these regulations gradually adapt to the realities of AI, automated testing will become the cornerstone of every long-term, trust-centered AI strategy.
