When it comes to generative AI, or genAI as we call it, technology is a cornerstone for success, right alongside people and processes. Bryan Kirschner, who’s the Vice President of Strategy at DataStax, stresses that getting a handle on this trio is crucial for adopting genAI effectively. We’ve already talked about the human side and the processes involved, so now it’s time to focus on technology.
Think of deploying technology as a tool to achieve your goals. Kirschner talks about fostering a growth mindset towards genAI, much like you would with any new tool. The aim here is twofold: to enhance your operations and to excel at using these enhancements. The processes in your organization should shift into what we might call a “new normal,” where a cognitive value chain seamlessly integrates knowledge into workflows, drastically cutting down on errors. This is similar to how digital value chains have improved data-driven experiences in businesses.
The goal is to steer organizations toward technology that consistently supports, rather than hinders, their progress. Having access to the right data is key. Teresa Heitsenrether, JPMorgan’s chief data and analytics officer, points out that genAI can make preparing for client discussions much easier, a task that usually takes a lot of manual work. Tools like ChatGPT Enterprise can help users craft prompts and integrate documents smoothly.
It’s important for genAI app developers to ensure their apps have access to relevant data sources. This way, the enriched context can lead to better outcomes. GenAI brings an added layer of complexity compared to traditional apps, which usually stick to predefined data sources and queries. GenAI, on the other hand, can use tools and APIs beyond its initial setup, requiring an integrated orchestration layer in developer tools. This allows genAI to effectively use both internal data and external sources. For example, a genAI app could easily move a support ticket conversation to Slack or reconcile conflicting internal data with verified external information.
Even though genAI can mimic human-like behavior, at its core, it relies on mathematical models to find relevant contexts. While vector search is a basic part of this, combining it with lexical search can improve results by balancing semantic understanding with precise keyword matching. That’s why a knowledge layer is essential, enabling comprehensive multi-modal search capabilities beyond traditional SQL queries.
For AI to be successful, there are three critical changes to consider: recognizing unstructured data as a key part of the data layer, improving orchestration and data access in developer tools, and building a strong knowledge layer. These elements will form the backbone of effective genAI processes, setting up both users and developers for success.
