Shashank Garg, Founder and CEO at Infocepts, speaks with Robert Lutton, VP, Sandhill Consultant and Editorial Vice Chair, CDO Magazine, in a video interview about enhancing business with generative AI, the process to capitalize on the GenAI opportunities, and building full-stack data products for business use.

Infocepts enables improved business results through more effective use of data, AI, and user-friendly analytics.

Garg begins by stating that with any hype cycle, the concerns are overrated in the short term and underrated in the long term. Therefore, when it comes to putting generative AI to work for business, he lists out three aspects:

Better business operations when it comes to effectiveness
Business outcomes for clients’ businesses
Employee or developer productivity

For instance, Garg mentions a situation where banks monitor a thousand loan portfolios, and it becomes challenging to do it manually with such massive data. This is where generative AI can effectively be leveraged for smooth business operations.

Speaking of better customer service, Garg refers to using a chatbot for shopping recommendations or making it easy for healthcare providers to help patients understand medical bills.

From the developer side, advancements in generative AI have evolved data applications. Garg maintains that traditional data analytics involves warehousing, cleansing, mastering, securing, and reporting data using BI tools.

Whereas, now developers can create highly immersive UX on the tools using the same curated data sets. This enables the rapid development of data apps, which can be completed in about a week instead of the previous six-month cycle.

To capitalize on the opportunity to use AI for business, Garg presents two approaches:

Renting existing apps for the use case offers the shortest and quickest path to value at a low cost. However, it does not provide a unique business advantage.
Organizations can bring open-source models and train them on organizational proprietary data. While it may take weeks or months and the cost would be higher, the potential benefits are significant.

Elaborating, Garg gives an example of working with a hedge fund that was working on creating an M&A detection tool. Currently, they rely on news articles and manual analysis by a group of analysts, which is cumbersome and costly. By training a model on external data sources, the hedge fund can create its own IP and automate the process, making it more efficient and valuable.

In addition, Garg states that the process to do this should involve taking input from not just the leaders but also from employees. He mentions conducting frequent hackathons that generate valuable ideas from clients as well.

Then, Garg suggests starting by aiming for a substantial gain target and then refining that. He also states that cross-functional collaboration between business users and technical experts is a must when organizations embark on this journey.

Moving forward, Garg discusses how clients could develop a full stack of data products for their businesses to produce the results. Traditionally, he says, it starts with a data team that produces some data products that analysts consume to produce insights.

Based on the insights, the business users then propose actions to be taken by operational users, asserts Garg. He continues that the cycle has been going on for two decades because of the state of technology and systems.

However, in the present state of technology with pervasive cloud and data, many business professionals are mastering the data language, affirms Garg. He defines a data product as something that starts with the personas it addresses and focuses on the end users instead of the ones responsible for generating them.

Furthermore, a data product must address the domain-relevant business use cases with consistent design patterns to achieve the business outcomes, which is the missing link. Also, it must have a very intuitive application interface, which may be in the form of a BI tool.

According to Garg, adoption problems with data products have been happening for two decades because users did not think like that, but adoption is a given based on the new paradigm.

There must be a mindset shift for better adoption of data products, says Garg. He reiterates that it must start with the personas, understand their decisions and actions, work backward into decisions, and have an intuitive UX on top of those.

Another crucial aspect that Garg points out is that it is critical to have supportive metadata. He states that the entire stack is managed using versions of the product built iteratively and the best practices of software development.

In conclusion, Garg mentions releasing a managed data product called Employee 360 that addresses various use cases such as workforce planning, retention, performance, DEI, and compensation analysis. He notes that this data product can be considered an example of how to apply a full-stack data approach to multiple organizational areas.

CDO Magazine appreciates Shashank Garg for sharing his insights with our global data community.

PODCAST | AI Models Can Pick Needles Out of a Haystack — Expedient SVP for AI