Lancaster
Lancaster
Lancaster
LA! 4XQ
Relative Insight is a deterministic AI company that turns organisations' language data (surveys, reviews, contact centre transcripts, AI agents, chat bots, social and CRM content) into trusted, auditable Voice of Customer insight.
The company grew out of a ten-year research partnership between Lancaster University's linguistics and cyber security departments, where the underlying technology was originally developed to help law enforcement identify adults posing as children online. Relative Insight has spent the years since applying the same comparative language analysis techniques to a different problem: helping brands understand what their customers are really saying, and why it's changing.
Unlike generative AI tools, which can produce different outputs from the same inputs, Relative Insight's Deterministic Comparative AI is repeatable and fully auditable, giving CX, insight and analytics leaders analysis they can trust, defend and act on. The platform connects to existing survey, contact centre and review sources, tracks how customer language shifts week on week, and pushes insight automatically into the tools teams already use, including Tableau, Power BI and Salesforce.
That same deterministic approach also powers Agent Optimizer, Relative Insight's newest product, which helps enterprises analyse and improve the chatbots and AI agents already handling their customer conversations. Rather than relying on subjective spot-checks, Agent Optimizer applies the same auditable analysis to agent transcripts, giving teams a clear, evidence-based view of where bots are succeeding, where they are escalating unnecessarily, and where deflection rates and customer experience can be improved.
Customers span telecoms, financial services, higher education, healthcare, sports and entertainment, travel and logistics, and defence and government (including Comcast, Medtronic, BAE Systems, McMaster University, New York Life and the NHL), drawn to Relative Insight because it doesn't replace their existing VoC stack, it makes it work harder.
