InformationWeek Podcast: Rightsizing AI frameworks to avoid failure modes

AI has introduced a wide range of tools and models that enterprises can explore, but these frameworks are not one-size-fits-all. For example, C-suite tech leaders might use retrieval-augmented generation for more user-friendly results. Meanwhile, a long context model -- designed to process very large inputs in a single pass -- may be more suited for analyzing broad data sets and information resources without relying on retrieval pipelines. The trick is determining which approach makes the most sense for a specific challenge. They may also work in tandem to combine retrieval with broader context analysis.

In this episode of the InformationWeek Podcast, Gabe Goodhart, chief architect of AI open innovation at IBM, and Robin Gordon, chief data officer at Hippo Insurance, shared their experiences of selecting the right available AI resources for their enterprise use cases.

They discussed how they make the first determination about the models they use, whether they let the scope of the data or the outcome they want dictate their choices, and how they address mismatches between organizational needs and the capabilities of the AI resources they deploy.

Related:How CIOs can climb out of the AI governance chasm

In our tabletop exercise, Questionable Ideas, Gordon and Goodhart put their knowledge to work as interim executives to save the fictional company from the latest misuses of technology by the resident gremlins, kobolds and goblins.