Power Digital runs its AI-powered Nova platform on data.world to empower clients to better engage with and question their data.
In late 2022, Michael Murray became Chief Product Officer at Power Digital, a growth marketing firm where he is focused on accelerating the deployment of data and technology for Power Digital through its Nova cloud platform and portfolio of purpose-built Nova applications.
Under Murray’’s direction, the Nova team developed a blueprint for “data activation” — contextualizing data for clear, actionable business use. Given each client of Power Digital connects their commerce, site and ad platform data, the team started with modernizing Power’s machine learning (ML) as a mechanism for scaling discovery of meaningful business insights. Through that lens, Power Digital partnered with data.world to establish a data catalog built on a knowledge-graph architecture.
As it happened, OpenAI’s ChatGPT launched the same month Murray joined Power Digital, and a missing piece of the puzzle presented itself in the form of a Large Language Model (LLM) that could answer complex questions in real-time – bringing data fluidity to the end user.
“The market changed and offered the potential to automate what was otherwise an arduous process between wanting to know something and getting the answer,” Murray said of ChatGPT. With this new possibility, the Nova team looked to bring AI to a familiar landscape of BI and “fixed” data reporting.
When faced with a report or a dashboard – no matter how comprehensive – a business or marketing leader invariably has questions. As much as BI has advanced, there often isn’t an easy way to answer dashboard questions quickly. More often than not, an analyst has to be engaged, extending the process by hours, days or even weeks. At best, the delay stalls access to insights that could be put into action. At worst the questions are never answered and jeopardizes opportunities to move the business forward.
AI represented a new path, but LLMs presented challenges. On their own, LLMs are prone to inaccuracy and will surface and present misinformation as fact, making them hard to trust. Additionally, they lack explainability and governance to facilitate improvements in accuracy, making them hard to scale. Power Digital’s Nova team knew they needed to address these limitations to deliver on the vision of “fluid interactions.”
“It’s not about delivering all your data to generative AI – it’s about positioning aspects of your data to an LLM in a way that you can sleep at night,” Murray explained. “A data catalog provides semantics for the knowledge graph, and a knowledge graph provides a map of data relationships which is jet fuel for an LLM.”
Generative AI opens new doors
For Murray and Power Digital, data.world’s Knowledge Graph architecture provides the context needed to maximize outputs from the LLM. data.world’s Knowledge Graph models data and metadata in an interlinked, flexible, and open graph format. With the organization’s data and knowledge governed in this flexible format, AI-powered applications have the rich context needed to generate accurate, explainable, governed responses.
data.world contextualizes the broad portfolio of data Power Digital manages, and it provides the company an opportunity to build an industry ontology for advertising and performance marketing. Connecting data from hundreds of clients to create a reliable foundation for powerful, actionable applications fueled with required context for trustworthy generative AI.
“We needed the data catalog to do powerful things with data with a high degree of confidence and a high degree of trust,” Murray said. “You need to know the ingredients and the ingredients of the ingredients—the metadata. Given the data we're working with, we also need to know and map the relationship between and across the data.
“We needed a system of record that could manage the data, map the data, and could be easily used to expand and improve the value of the data. Ultimately, it's about the outcomes, but you need the componentry in order to get to the outcomes.”
With generative AI and data.world, the interactivity and prompt-driven creativity enables Power Digital and its customers to quickly ask and answer questions of their digital marketing data.
Three key pieces enable this:
Knowledge Graph: Provides the mapping of the concepts and data
Data Catalog: Provides the needed context and governance
Snowflake Cloud and Container Services: Virtualizes the data so it remains in place and data isn’t exposed to an open environment
The future: Fluid interactions with data for every customer
With the data foundation in place, Murray is eyeing business outcomes. Primarily, the team wants to extend the power of insights delivery. He pointed out that for his entire career, any time someone looked at a dashboard, they instantly have questions about the context and extended meaning of the data. The combination of data.world and generative AI can help answer those questions and deliver insights in a fast and fluid way – as if you were talking to an expert analyst who knew the ins and outs of your organization and its data.
Murray believes that very soon Power Digital clients will be able to look at a Nova-powered dashboard and engage through a prompted interface to get trusted insights immediately.
“When I can ask questions of your data and gain trusted insights instantly and repeatedly, the speed of innovation, the speed of moving from information to insight to action goes hyperbolic,” Murray said. “And no end user should have to know the intricacies of the data. You don’t need to understand the engine of your car, you just want to drive it.”