January 10, 2023Data Analytics is The Foundation of an Enterprise AI Strategy
Curated advice from 50+ data leaders, industry experts & customers
Over the last 12 months, we’ve hosted 18 webinars with 50+ industry experts, data leaders, and customers reaching an audience of 25k+ data and analytics community members/professionals.
I’ve decided to do a 3 part series recapping some of the best actionable insights I’ve learned from these experts. These posts will center around three core themes that have emerged over the past year; aligning AI & BI teams to business outcomes, scaling self-service analytics, and optimizing cloud analytics for scale.
Let’s dive into the third post:
How Do You Measure the Success of AI Programs?
In our webinar focusing on How to Align AI & BI to Business Outcomes, I asked Aruna Pattam, the head of AI & Data Science at HCL technologies, about the best ways to measure the success of an AI investments. At HCL technologies, it starts with value-driven experimentation and concrete ROI estimates. She estimates that an ROI of 10x the initial investment is a good starting point for a successful AI program. The key takeaway is that companies must always view AI investment through the prism of business value and better decision-making outcomes. If we can’t come up with tangible ROI targets, then it’s a good sign that we need to understand the strategic objectives that AI can enable.
What Team Owns the Semantic Layer?
We spoke with Harendra Kanuru, the Engineering Manager for Big Data and Analytics at Home Depot, about how their company leverages the semantic layer to bridge data science and business intelligence to enhance their data science capabilities. One question that came up was who owns the semantic layer? The answer isn’t so simple because business intelligence stakeholders exist across the entire organization, from business leaders to data scientists and architects. While it’s a team effort, The Home Depot relies on a data governance team to act as the central touchpoint for data quality and ensure that the single source of language and truth that the semantic layer provides is actionable and value-driving for stakeholders across the company.
What Distinguishes BI and AI Analysis?
One theme that comes up consistently in our panels discussing analytics in organizations is how different BI and AI analysis are in terms of the approach to the data. The most significant difference is that the business analyst community usually analyzes data that’s already there for production workloads to empower specific decisions. Conversely, the data science teams leveraging AI analysis look at existing data to create new, potentially actionable data points. So while BI style analysis benefits the most from consistency and well-manicured data by reliable governance to meet the needs of the job, data science is more interested in experimentation through investigation, data mining, and hypothesis testing.
How Can BI and Data Science Work Together to Improve Business Outcomes?
As we worked through our various panel discussions, the industry experts consistently agreed that the semantic layer is the ideal bridge between the efforts of BI and data science teams to deliver value back to the business. In this model, BI teams define the standard semantics used by the company and the relevant dimensionalities like time, geography, and product. They also focus on understanding and generating the metrics needed by the business to understand the definition of success in their analytics efforts. Data Science teams develop domain-specific features and build predictive models based on their characteristics. In essence, the data science team is creating new data points to feed into the semantic layer that the entire organization can use. The semantic layer ensures that this new data speaks the same language the BI teams define, acting as a unique bridge between these disparate efforts.
How Do You Deliver AI and BI at Scale?
85% of all AI projects fail. One of the reasons AI and BI transformation efforts fail is how hard it is to bridge the gap between the intelligent data scientists who develop the models and the business leaders and analysts who need actionable insights to perform their jobs more effectively. Without these teams working in concert, it’s impossible to deliver AI and BI at scale. In our webinar on How to Make AI & BI Work At Scale, we addressed this question with Ben Taylor, the Chief AI Evangelist at DataRobot.
The critical takeaway from Ben’s presentation is how important it is to empower business SMEs to contribute to AI and BI projects at every stage of the life cycle. Tools like the semantic layer help drive this by turning every BI stakeholder into a potential contributor to scale AI and BI across the organization.
How Can Data Science Teams Do Less Data Wrangling and More Prediction?
While companies generate more individual data points today than ever before, volume isn’t the main driver for data science teams’ time spent wrangling data rather than building predictive models that benefit the business. Instead, the variety and complexity of today’s business and systems data creates challenges for Big Data, BI, and Data Analytics. These data sources are diverse and often distributed across many complex silos. Data science teams have to spend a great deal of time figuring out ways to integrate multiple data sets so that BI teams get all the information they need to empower decision-making and improve business outcomes.
Data lakes help bridge the gap, but on their own, they still struggle with BI performance because of the unstructured data prevalent in these types of platforms. Combining data lakes with a semantic layer eliminates the need for costly overhead and performance bottlenecks prevalent in data lakes. It helps data science teams focus on predictive analytics rather than data integration.
Stay Tuned for More Actionable Insights from Our Webinar Experts.
That wraps up my last recap of the best insights from our 18 webinar panels. Review our webinar section to see all of the webinar topics and featured guests. The follow-on post will focus on industry expert insights on scaling self-service analytics, and strategies for data literacy across their organizations.