On April 21, 2026, Robin Gordon, chief data officer at Hippo Insurance, and Gabe Goodhart, chief architect of AI open innovation at IBM, appeared on the InformationWeek Podcast to discuss strategies for selecting and deploying AI frameworks that align with specific enterprise needs. Their conversation centered on the concept of “rightsizing” AI tools—matching model capabilities to data context and business outcomes to avoid common failure modes in enterprise AI adoption.
The podcast episode, hosted by InformationWeek senior editor Joao-Pierre S. Ruth, explored how technology leaders navigate the growing complexity of AI options. Gordon and Goodhart emphasized that no single AI framework is universally applicable, and that success depends on understanding whether a retrieval-augmented generation approach or a long context model better serves a given use case. They also discussed scenarios where combining both methods could yield stronger results.
According to the discussion, retrieval-augmented generation is often favored by C-suite technology leaders seeking more user-friendly outputs, particularly when integrating external knowledge sources with internal data. In contrast, long context models—designed to process very large inputs in a single pass without relying on retrieval pipelines—are better suited for analyzing broad datasets and information resources where maintaining full context is critical.
Gordon and Goodhart explained that the first step in model selection involves determining whether the scope of the data or the desired outcome should drive the decision. They noted that mismatches between organizational needs and AI capabilities frequently arise when teams adopt tools based on trends rather than fit-for-purpose analysis. Addressing these gaps requires ongoing evaluation and, at times, adjusting deployment strategies mid-project.
The executives also highlighted the importance of organizational alignment in AI initiatives, pointing out that technical decisions must account for workflow integration, team readiness, and change management. They referenced an internal tabletop exercise called “Questionable Ideas,” in which they assumed interim executive roles at a fictional company to simulate responses to emerging technology risks, including misuse by fictional entities labeled as gremlins, kobolds, and goblins—a metaphorical way of discussing unforeseen operational and ethical challenges in AI systems.
Both leaders stressed that rightsizing AI frameworks is not a one-time decision but an iterative process requiring continuous feedback between data teams, business units, and IT leadership. They advocated for establishing clear success metrics early in the selection process and revisiting them as models are tested and scaled.
The podcast underscored a growing trend in enterprise AI: moving beyond experimentation toward disciplined, context-aware implementation. As AI tools proliferate, the ability to match the right model to the right task is becoming a key differentiator for organizations aiming to scale AI responsibly and effectively.
For technology leaders seeking to improve their AI framework selection process, the insights shared by Gordon and Goodhart offer a practical starting point: begin with a clear understanding of the data context and desired outcome, evaluate trade-offs between retrieval-augmented and long context approaches, and remain open to hybrid solutions that leverage the strengths of both.