The Databricks Financial Services forum in NYC last week was a great event to learn how industry leaders are utilizing Data, Analytics and AI. It was also a chance to preview new Databricks capabilities that help drive value through Data and AI. Below are some key takeaways from the event.
AI is Accelerating New Business Models and a Threat to Incumbents
AI technologies are a democratizing force that is accelerating the pace of change in all industries. Broad access to this new wave of technological capabilities is a threat to incumbents and a great opportunity for those that can harness these technologies. As Junta Nakai, Global VP at Databricks, pointed out, this is reminiscent of the first internet boom that changed everything in the early 1990s. Who remembers when Japanese banks were the leading companies in terms of market capitalization in the late 1980s, later to be replaced by the Tech giants? AI’s usage is creating new business models and challenging existing businesses that don’t quickly adjust. One example is Nvidia, the manufacturer of chips that power AI models, going from a $1T to a $2T company in less than a year and surpassing Google, Amazon and Meta for market capitalization.
How can your company move up on the Data and AI maturity curve and be on the growth side of this tsunami of change? The big themes discussed in the forum were Democratizing Data, Data Governance, Monitoring Cost and Technology Platform strategy.
Democratizing Data
Democratizing data for broad use within organizations to make it available for AI application is key to success. This can be difficult due to silos of data in various systems, complexities and costs of moving data, lack of visibility of data lineage and generally a lack of trust in data. A pre-requisite to democratizing data is modernizing the tech stack to more of a cloud-based solution that allows for scaling of resources quickly, a strong platform focused approach, and governance of data across platforms.
Data Governance
The Financial Services industry has been successful in using data to drive decisions largely because of the government oversight that has led to them be very good at data governance. Strong governance can lead to trust in data and ease concerns with sharing across departments. We heard a good description of what good and bad governance looks like from Robin Sutara, Field CTO at Databricks, and how critical it is for organizations to become more mature on the Data and AI maturity curve. Governance can mean different things to different people so it’s important to define what that means in your organization. Categories should include governing access, data quality, security, usage, residency, and sovereignty of data.
Cost Observability
Training and executing generative AI models with large amounts of data is expensive, and sometimes the true costs are buried in the financial accounting of Tech spend that isn’t always visible to users and their leadership. As companies embark on their AI journey, building in cost observability and monitoring are key. There were numerous examples given of high Tech spend for low value AI use cases where the costs weren’t fully understood and were not part of the planning process. As companies are looking to build their AI capabilities, it’s important that any platform used has good cost visibility, monitoring, and controls.
Platform Strategy
For a while, Tech-forward companies have been experimenting with and using machine learning techniques for common problems of prediction and classification. A prediction example is forecasting which customers will default on loans. A classification example is determining which transactions are fraudulent. However, with generative AI there is a whole new world of possible uses because the models are generating something new and helping discover new insights.
Generative AI models at most companies are still largely experimental, and one of the hurdles to putting these models into production is that there isn’t a good platform approach for how to graduate these to production quality solutions. Trying to embark on this journey without a platform strategy will limit the speed of experimentation to production.
Databricks Data Intelligence Platform
Throughout the day we saw demos of how the Databricks Data Intelligence platform is helping with all the above and why it’s a great solution regardless of where a company is on the Data and AI maturity curve. Databricks is an accelerator that can help companies quickly progress into the new world of AI-enabled data capabilities.
We were excited to see new capabilities such as how AI can enhance data governance and the quality of metadata as well as help users find the data they need. A very common problem is poor metadata quality due to the manual effort needed to create and maintain metadata, and difficulties finding the data needed. Imagine a product that can generate metadata through AI considering factors, such as the content of the data and its usage and evolve over time. The Databricks Data Intelligence solution aims to be that product, and the pace at which they are innovating on the solution is impressive.