The hype surrounding generative AI is palpable, with businesses of all sizes rushing to harness its transformative potential or, at the very least, its marketing potential. A lot of the hype really is just hype, but at Workday, we’ve been using large language models (LLMs), like those which power generative AI, for years. We’ve already been able to deliver tangible benefits for our customers leveraging these technologies, and we’re investing even more, looking beyond the hype cycle to deliver meaningful business value and tools that will redefine the way we work.
We’re currently building capabilities that leverage generative AI for various language and image-related tasks, including natural language generation, document understanding, and content search, summarization, and augmentation. These new capabilities will enable our customers to unlock increased productivity through streamlined tasks and processes, increased efficiency, and better decision-making. These are not far off in the future—in fact, our customers can expect access to these cutting-edge features within the next 6-12 months.
So let’s take a deeper look at how Workday is leading the enterprise generative AI revolution.
What We’re Doing Differently
Our approach to generative AI is different in several ways—most notably because of our unrivaled dataset. We firmly believe that the effectiveness of generative AI hinges upon the quantity and quality of the data it is built on. As evidenced by the many stories highlighting how generative AI chatbots have provided biased or incorrect responses, LLMs are only as good as the data that feeds them.
The foundational models that have dominated headlines recently have been deliberately trained to solve a broad class of problems from the broadest set of data available. That vast set of data is not all of equal quality or of well understood provenance—resulting in well-documented erratic, incorrect, unsafe behavior, or infringement of intellectual property. We have also seen that the safeguards on responses put in place to deal with poor training data do not hold up over time. To address this for our critical use cases, we focus on targeted, domain-specific models and high data quality above all else to provide outputs customers can have confidence in.
One of our key differentiators is that all customers are running on the same version of Workday, including the same data model. At Workday, we have over 60 million users who contribute to nearly 450 billion transactions processed by the system every year—and growing. With our customers’ permission, we utilize that data as the fuel for our generative AI capabilities. This massive, high quality dataset allows us to build models that consistently generate accurate, meaningful, trustworthy results.