BI Modernization in the AI Era with Swapnil Srivastava

Business Intelligence (BI) modernization in the AI era represents a transformative shift in how organizations leverage data to drive insights and decision-making. With the advent of artificial intelligence (AI) technologies, BI tools are evolving rapidly to incorporate advanced analytics, machine learning, and natural language processing capabilities. This convergence enables businesses to extract deeper insights from their data, automate processes, and enhance the overall intelligence of their operations. In this dynamic landscape, BI modernization encompasses integrating AI-driven features, adopting agile methodologies, and cultivating data-driven cultures to stay competitive in today’s digital age.

To provide more insights into BI modernization, we spoke to Swapnil Srivastava, the global head of Evalueserve’s analytics practice. Swapnil has over a decade of experience leading the design and development of analytics solutions across various industries. Since joining Evalueserve in 2015, he has significantly expanded the practice in terms of team size, revenue, and global presence. Swapnil has cultivated a diverse team of domain experts, architects, engineers, data scientists, and technologists, cementing Evalueserve as a comprehensive analytics partner for clients. He holds a bachelor’s degree in electrical engineering from Pt. Ravishankar Shukla University, a post-graduate diploma in business finance from the Indian Institute of Finance, and an MBA in international business from Amity University.

Our interview with Swapnil assesses the effectiveness of BI dashboards in 2024 and explores their ability to meet expectations. Other topics of discussion include the integration of AI into BI tools, organizational strategies for adopting generative AI, future trends in BI modernization, workforce readiness, and leadership advice for navigating BI modernization in the AI era.

AIM: Could you summarize the evolution of the data analytics industry from its inception to today, focusing on advancements in BI technology and modern trends like BI modernization?

Swapnil Srivastava: I started my analytics journey around 2013, and we were partnering a lot with companies just starting their BI journeys. Since then, I’ve followed the topic closely and have seen the BI modernization space evolve significantly.

Analytics used to involve a lot of large-scale reporting. Initially, it was heavy on Excel spreadsheets. Analysts would create sophisticated Excel reports with macros for automation and similar tasks. Then, IBM, SAP, and other ERP (enterprise resource planners) providers created their own BI platforms.

In the next stage of BI modernization, business users wanted dynamic dashboards. Self-service BI and analytics tools like Tableau, Qlik, and Spotfire emerged, ruling the BI market for quite some time. Microsoft’s Power BI quickly gained popularity, as many companies were already using Microsoft products, and they were able to seamlessly and easily integrate Power BI with their existing Microsoft systems.

Subsequently, with the emphasis on machine learning and AI, newer technologies like ThoughtSpot started emerging. With Gen AI and tools like ThoughtSpot, a new paradigm emerged where users could chat with their dashboard directly, redefining the self-service concept. This feature became incredibly popular, and Gen AI has proven to be a transformative trend among existing players like Tableau and Qlik as well.

Overall, the evolution in BI has been toward minimizing time-to-value and time-to-insights. This minimal time-to-value and insights can only happen if the technology is largely self-service and automated, making insights more accessible and quickly available.

AIM: What makes a BI dashboard successful in everyday business use, and how can its value be maximized? What characteristics differentiate a highly usable BI dashboard with a high consumption rate? Lastly, how do you distinguish between a successful BI project and one that is not as effective?

Swapnil Srivastava: In my client organizations, most of which are in the Fortune 500, I have observed that they usually have too many dashboards. Typically, only about 15 to 25 percent of these dashboards are actually utilized, which means many dashboards sit idle, creating a burden on the servers and technical debt for the company. So, organizations should pare back the number of dashboards to only those they really need and use, ensuring these dashboards are providing actionable insights.

The second issue I see with clients is having too many KPIs, which defies the definition of a KPI. It’s a key performance indicator, so you should not have more than five or six for a particular line of business or division. Everything else is a metric, not a KPI. You can also have too many metrics. Businesses should define which KPIs are critical. Once you’ve done that, classify your KPIs – output-based KPIs, leading and lagging KPIs, etc. – and weave them in a way that tells a story.

Nowadays, we use the term “BI transformation” instead of “BI modernization.” “BI transformation” reflects a continuous process of adapting and evolving the BI discipline, whereas “BI modernization” can suggest a one-time modernization effort. To ensure successful, actionable dashboards, businesses need BI transformation – rethinking the entire BI function, not just implementing new tools.

As we see it, there are three aspects to BI transformation: modernization, optimization, and governance.

1. Modernization 

Many clients move from one platform to another for a variety of reasons. Some are still upgrading from Excel to self-serve platforms like Tableau or Power BI. Others are migrating between platforms, perhaps due to other systems migrations. For example, we have clients who have migrated from AWS to Azure and chose to change from Tableau to Power BI because it integrates better. We’re also seeing more advanced BI teams take the open-source route with the likes of Node, React, and Angular to create highly custom dashboards. These dashboards can be more flexible in areas where Tableau and Power BI are limited. They tend to perform better on hand-held devices when executives are on the go, and you can be more creative. 

The second aspect of BI modernization is embedding new technology like Gen AI. There are ways to embed Gen AI in dashboards and platforms where it’s not natively available. Augmenting dashboards with natural language capabilities makes them even more self-serve, helping users find insights faster and navigate complicated dashboards. 

2. Optimization

The next element of a successful BI dashboard is optimizing the architecture to minimize latency because you want a real-time user experience. 

The goal is to increase adoption and promote more data-driven actions across the organization. The UI/UX must be optimized, keeping the end user’s needs in mind. An improved UI/UX will see an increased adoption rate. Dashboards that aren’t user-friendly and easy to navigate will see a decrease in adoption, negatively affecting the ROI of BI efforts. 

Secondly, the number of dashboards and the KPIs and metrics they track must be optimized and rationalized. Are the dashboards meaningful and popular with users? Are the KPIs truly critical and revealing? 

3. Governance

The third step of BI transformation is governance. There are two aspects to governance.  

First, the quality of your insights is only as good as the quality of your data. As such, it’s crucial to place a heavy emphasis on data quality.

Second, companies must have proper governance regarding who develops and maintains dashboards. One issue I’ve seen is that, in many client organizations, a new dashboard is created every time there’s a new requirement, whereas the business need could have been met just by creating an additional view to an existing dashboard. Proper governance will help you monitor utilization and the number of dashboards being developed.

Each of these three aspects is individually important to BI success, and combined, we call it BI transformation.

AIM: How are companies like Evalueserve leveraging the latest advancements and technologies in backend infrastructure to enable the development of ideal BI solutions?

Swapnil Srivastava: A lot of these platforms, such as Power BI and Tableau, are already using the latest technologies in backend infrastructure. Power BI is now integrated with Microsoft Fabric, and Fabric offers various types of copilots. Now, because all these technologies are very new, a lot of experimentation is happening to figure out ways to automate backend activities like ETL (extract, transform, load). We work with clients to determine what’s feasible and which backend activities are most important to automate. For example, reports have different categories. There are some reports that you would only need access to periodically – weekly, daily, monthly, quarterly – but then there are other types of reports for use cases like predictive maintenance that need to be refreshed almost in real-time. The goal is to make the frontend as near real-time as possible.

I’m also seeing on implications on the talent front. At Evalueserve, we have a data engineering team and a BI team. The work is intrinsically linked, as integration is required in the background of BI efforts. Now, many emerging technologies have made it possible for data engineers to become savvy in BI and for the BI folks to become very savvy in data engineering. Those two worlds are converging.

Additionally, as more activities on the backend are getting automated, it frees up time for people to add more value to dashboards. The ultimate goal is to share insights with end users that they can act on to attain a tangible ROI. I was reading an article about business intelligence in a leading journal. They used a fantastic phrase: “KPIs should inspire, not just inform.” But KPIs will only inspire if they are presented in a particular fashion. When you have multiple KPIs, they should be viewed together in the form of a story so you can drill down, look at the root cause of operational inefficiencies, and have immediate actions lined up. You can then delegate those immediate actions.

AIM: How do you justify the impact of BI solutions to clients, considering that many of the outcomes may not be immediately tangible, and it’s challenging to quantify the ROI based on metrics consumed? Specifically, how do you address the dilemma of showcasing the impact made by BI tools or solutions on an organization?

Swapnil Srivastava: There are two ways in which you should think about value when it comes to BI organizations. One is the productivity aspect, and the second is the actual value of the delivered insights. 

On the productivity side, we look at how much manual effort we save. For instance, if there was no existing report for certain metrics, and now you have a dashboard available, you get insights in a couple of hours instead of running around for a few days. We consolidate reporting or rationalize KPIs, and now you have gone from an overwhelming number of reports to just a few, providing significant cost savings on BI.

However, the real ROI is from the KPIs and metrics presented and your actions based on them. To guide BI leaders on how to think about the value holistically, we created the PROFIT Framework.

P stands for your positioning in the industry. If your market share goes up because of certain actions you take, or your brand gets elevated, that’s one type of value.

R stands for risk management. So, suppose you can identify, assess, and mitigate different kinds of financial risks for your company. In that case, that’s a type of value you generate from BI. 

O stands for operational excellence, which is anything that helps a company fulfill its promise to its end customers and keep them satisfied. 

F stands for financial performance. Suppose we present you with insights that help you improve your balance sheet, income statement, cash flow, valuation, shareholder returns, stock price, or anything else that is financial in nature. In that case, that’s a type of value. 

I stands for innovation. If we are providing consumer, market, or competitive insights that help you innovate, that’s also a type of value. 

T stands for trust building. Anything that helps you build trust with your customers, regulators, Wall Street, suppliers, or employees is also a type of value. 

Our belief is that the insights you get from your BI platform will fall under one of the six buckets outlined in the PROFIT Framework. When we create these reports, we look at the primary KPIs and try to put them into these buckets. Some will fit one bucket, whereas others will fit into multiple buckets. Those are the categories under which we would quantify the ROI for our end users. 

The more difficult question is how to calculate ROI. One interesting way I’ve seen companies capture BI value is by doing a heavy test control. Conduct an A/B test, just like in marketing, where you roll out a new dashboard to a test population and measure KPIs before and after. You can then calculate the RoI based on a combination of how well the test population performs vs the control and the lift in performance.

AIM: What are some of the future trends you see in Gen AI and other emerging technologies defining BI tools? Specifically, how do you foresee the relationship between Gen AI and the development of visualizations evolving, and what overall impact do you anticipate emerging technologies will have on these visualizations?

Swapnil Srivastava: Generative AI is completely transforming how end users interact with data and visualizations, largely thanks to Gen AI’s natural language capabilities. Sometimes, users may want the answer in a simple numerical or natural language response, and other times, they might want to see a new visualization.

A use case that has been successfully implemented in several organizations is “chat with your documents.” In a way, this is business intelligence, as it involves extracting insights from various documents, often containing unstructured data. So, whether it’s a repository of documents or structured data in Excel or CSV format, the ability to chat with this data is gaining traction among companies, given the heavy reliance on files and documents.

Another evolving trend is “chat with your dashboards,” which extends beyond traditional dashboard interfaces to include websites or any user interface where a chatbot can assist in navigating and extracting insights.

However, what I see as the future, with potentially high ROI and disruptive potential, is “chat with your data.” This concept envisions a scenario where the BI layer becomes obsolete, replaced by purpose-driven data lakes tailored for specific domains like marketing, finance, or supply chain. Imagine a blank canvas atop these data lakes where you can pose holistic questions. These questions would then be processed into queries, traversing the available data, and transforming the answers into natural language, along with the most suitable chart or representation.

At Evalueserve, we’re experimenting with these concepts with several clients. We’ve built our own Gen AI accelerators, where the idea is to present a blank canvas where you can ask anything, and it responds with the answer and the most optimal data visualization. However, the challenge lies in that the current technology has low accuracy when converting natural language to SQL or generating graphs and charts. But this science will evolve further.

AIM: How do you ensure responsible usage and accuracy in AI applications for modern BI, especially considering the potential challenges when opening up data systems for non-SQL users to interact using English language queries, which may lead to misinterpretation and inaccurate data retrieval?

Swapnil Srivastava: The current accuracy levels of AI-to-SQL technology is low. Hence, there must always be a human to validate not just the quality of the queries being produced but also the quality of the output over a period of time. In English, depending on context, words can mean different things. That’s when linguistics becomes very important. So, various disciplines, like art and science, must collaborate to improve these technologies and reduce biases and inaccuracies.

Previously, the customer analytics world relied solely on algorithms. However, there is now a broader understanding involving ethnography, psychology, and neuroscience, all contributing to the advancement of consumer behavior tracking. Similarly, in this space, linguistics and many other disciplines will come together to advance the field.