The case study “Taming Gen AI Hallucinations: Inspeq AI” explores the strategic challenges faced by Inspeq AI, a Dublin-based start-up founded in 2023 by Kumar Apoorva and Vijayakumar Ramanujam, as they navigate the emerging market for Generative AI (Gen AI) optimization. The company focuses on addressing hallucinations—convincing but fabricated responses—and biases in Gen AI outputs through its suite of products including an SDK, Plugin, and Platform.

As Inspeq prepares for the regional semi-final of the Start-up World Cup at the Dublin Tech Summit in June 2024, the founders must make critical strategic decisions about resource allocation, product focus, and market positioning. They face a rapidly consolidating industry with increasing interest from tech giants, strict regulatory requirements including the EU AI Act, and diverse stakeholder perspectives on the right approach to AI trustworthiness.

The case examines how Inspeq can design innovation processes that embed regulatory requirements while maintaining technological edge, while also considering whether to pursue hyper-specialization in specific industries or develop a meta-layer integration approach. Additional challenges include determining responsibility for residual hallucinations, adapting their sector-agnostic approach to diverse client needs, and addressing disagreements in their two-level feedback system. Through Inspeq’s journey, the case offers insights into strategic decision-making in emerging technology markets characterized by rapid innovation, evolving regulations, and intense competition.

Learning Objective

• Examine and analyze Inspeq’s legal liability and ethical responsibility, particularly regarding residual hallucinations.
• Analyze and evaluate approaches for reducing Gen AI hallucination while maintaining creativity through an integrated approach to Inspeq’s optimization strategy.
• Evaluate and design a competitive strategy for maintaining a technological edge in the Gen AI optimization space against tech giants, considering options for hyper-specialized solutions versus continuing as sector-agnostic solutions versus integrative Meta layer to incumbent AI solutions.
• Critically assess the effectiveness of Inspeq’s sector-agnostic AI solutions across diverse industries such as healthcare, finance, and legal retail, considering each domain’s expected nuanced solutions.
• Design and justify a protocol for resolving conflicts in Inspeq’s two-level expert-in-the-loop feedback system, addressing scenarios where there is a lack of consensus among the experts regarding Inspeq’s output.