Summary

It is a big step to prepare an existing knowledge base and data sources to make them available within a generative AI solution. It’s likely the most significant step because the hard work of creating the ChatGPT model was done for you. For many enterprise solutions, this can be an overwhelming task. Just start small. Learn from the use cases to prioritize solutions that provide the most significant value with the least cost (recall our scoring discussion in Chapter 4, Scoring Stories). Over time, land grabs can expand into other data sources and, thus, new use cases. All of this has to be done with quality in mind. Measuring and monitoring are critical. Newer doesn’t mean better. Mix and match ChatGPT models to perform specific tasks or optimize cost or performance by using one model over another. Use a collection of third-party resources—possibly even other models tuned to a particular problem space—to refine results, make data available to the...