Using Generative AI for real-world KM solutions
Digital agents
Model fine-tuning, by which GenAI systems become well-versed in the domain-specific knowledge for a particular application, is essential for the use of digital agents. “For customer support, even to answer simple questions, the models have to be trained on all of the documentation,” Patel stipulated. “So, we are preparing documents and feeding them so the embeddings can be created for our documentation.” The embeddings Patel mentioned are respective vectors of content to store and search via a vector data store. Gu likened this vital step to training a new employee who’s a recent
graduate. “Each vertical has its own specific domain knowledge—sales, marketing, and branding, for example—and they all have their specialties. Those are things models need to acquire from a specialized training process. Think about a college graduate or high school student; how much do you expect them to know?”
Vendors can assist with this facet of fine-tuning models so organizations can get the results they expect from digital agents for interactive GenAI
deployments. “Some of the basic documents, the PDFs, and other materials, they can share with us and we’ll do this heavy lifting for them,” Gu said. “We also have an interface where they can easily upload those documents and be able to customize and train the agent very quickly.”
Once the statistical cognitive com- puting models are fine-tuned in the domain-specific knowledge organizations provide, they’re not only able to answer questions about it and retrieve information on demand, the models can also provide timely recommendations to users based on specific situa- tions. Galal detailed an application in which an educational institution was using the records management solution for providing academic advisors with suggestions for responding to student questions about scheduling, courses, and their workloads.
Right now
These examples showcase the capac- ity for generative models to enlarge the scale and speed at which KM practi- tioners can access the information they need to meet their business objectives. With many investments in this technol- ogy still getting underway, its overall worth is inestimable.