Boosting team productivity and decision-making with GenAI

Slowed by costly and frustrating manual processes

Throughout their 80 years in business, HP has been at the forefront of innovation. Within their Big Data Platform and Solutions organization, this team provides data ingestion, platform support and customer data products for all HP business units. However, due to their large volume of data from PCs, printers and web and mobile apps, and the sheer number of partners and customers on the platform, there is a significant burden and bottleneck for nontechnical users to discover, access and gain insights from the data. As a result, making high-quality data available to thousands of data users on their data platform required manual interventions that significantly slowed productivity, added latency and contributed to overall costs.

HP’s data engineering teams, who were used to being regularly bombarded with questions about specific data models and issues, were overwhelmed with inquiries and troubleshooting requests. Whether users needed access to restricted data or wondered how to use platform features or new employee onboarding, the teams were constantly distracted by customer support requests. William Ma, Data Science Manager at HP, explained, “With all those inquiries coming in, our teams were spending 20–30% of their time trying to whip up SQL queries, dig into the data and cross-reference in different areas to figure out what was going on. With a five-person team, this equates to an additional full-time employee.”

Data teams were also required to help build usage and share cost and budget analysis dashboards used by HP’s leadership for strategic decision-making. To produce these insights, teams had to access and explore data manually. Additionally, data teams didn’t always have the bandwidth to support incoming requests, creating latency that impeded effective decision-making in real time. William stated, “When you need to move your data out of the warehouse to a workspace where you can apply AI tools, you typically need to make copies of the data. There are additional security and privacy governance concerns, such as DSRs, which means you must delete data when customers send in their DSR requests. All those complexities add up to tying up data teams and driving up costs.” 

With the desire to move fast and reduce the load on their engineering team, HP looked to leverage generative AI (GenAI) to alleviate the data barriers to innovation and intelligent decision-making.

Data team enablement with Databricks

HP began using the Databricks Data Intelligence Platform on AWS while moving from their data center, leveraging a lakehouse architecture to unify data, analytics and AI. William elaborated, “Databricks has a data-centric approach that brings AI, analytics and the GenAI tools we need on a single platform on top of the lakehouse architecture. Serverless Databricks SQL allows us to provision ephemeral compute clusters on demand, and Unity Catalog is used for data management and governance, making it very easy to control fine-grained access for our 600+ Databricks users and enabling them to access information via a knowledge base chatbot. Unity Catalog also provides data lineage and auditing capabilities. We don’t have to reinvent the wheel and everything we need to build data and GenAI solutions comes with it. It’s all built in the Databricks Platform, and we just have to leverage its framework and tools.”

HP migrated from AWS Redshift to Databricks SQL for better performance/cost ratio, resource isolation, attribution and unified data cataloging, access control and governance. Using the Mosaic AI Playground, HP experimented with different large language models (LLMs) — ultimately selecting DBRX as the most appropriate and cost-effective model for their chatbot needs. With GenAI at the forefront of the data team’s ambitions, they quickly implemented their first use case to help onboard their data users. Information was scattered among internal wiki pages, SharePoint files and team support channels, making it difficult for partners to find answers and get started quickly. To disseminate and manage the knowledge more efficiently, HP implemented a retrieval augmented generation (RAG) solution with a Vector Search database back end for sourcing all relevant information. 

Using Databricks Mosaic AI, an intern on the data team quickly developed a functional assistant chatbot that provided accurate answers in real time. William explained, “This solution has quite a few pieces: a web front end and a back-end agent that essentially parses the input from the user, searches and grabs relevant data, then sends it to the GenAI endpoint to generate answers. Another component is a web crawler that crawls and tokenizes the internal information on various sites and gets them populated into the Vector Search database. The college intern implemented the end-to-end solution in less than three weeks on Databricks Mosaic AI, which is amazingly fast compared with other teams who spent months building similar solutions on other platforms with experienced staff engineers.” The Vector Search database entries and answers generated also contain URLs for reference so users have the opportunity to validate or explore further.