(US & Canada) | AI Experimentation and Implementation Demand Separate Systems — Infocepts Founder and CEO
Shashank Garg, Founder and CEO at Infocepts, speaks with Robert Lutton, VP at Sandhill Consultants, and Editorial Vice Chair at CDO Magazine, in a video interview about focusing on business value, his recently published whitepaper on , the critical pieces for establishing AI labs for businesses, the DiscoverYai solution, and the delineation between systems of implementation and experimentation for building AI labs.
Infocepts enables improved business results through more effective use of data, AI, and user-friendly analytics.
Garg begins with a shout-out to the global Infocepts team for working tirelessly on cutting-edge problems in the data and AI space for around two decades. The success that followed was built on experience and expertise while working closely with clients.
Focusing on business value has driven the success of Infocepts, says Garg. He adds that while one can lose sight of the value after getting submerged in data complexities, the organization approaches it backward.
Elaborating, Garg says that Infocepts works backward by first focusing on articulating the business value and asking relevant questions about the problem being solved, business outcomes, cost efficiency, and revenue.
Continuing, he maintains that if anyone invests in data and AI initiatives, they must get the ROI in cost savings of 2-3x or through generated revenue of 10 to 20x.
When asked about the context of his recent whitepaper on the “Data and AI Initiatives to Drive Business Growth in 2024,” Garg states that the concept came from witnessing misplaced project priorities.
The whitepaper focuses on things a CDO or data leader must get right in the context of technology disruptions with AI and generative AI in particular. He lists out the critical aspects to be considered:
Prioritizing data enablement for AI
Reinforce the AI council and establish an AI lab
Simplify modernization
Expanding on prioritizing data enablement for AI, Garg emphasizes that data helps in differentiating, not the models. Even the latest AI model would not work without data.
Next, he urges data leaders to establish or reestablish an AI lab by focusing on business value. He recommends leaders think through the generative AI disruption and make it work for their business. Further, he asks to simplify the modernization frenzy.
The other topic that demands data leaders’ attention is assessing the definition of data products in businesses today and extending them to full-stack data products.
Moving forward, Garg states that there are two important pieces for establishing AI labs for businesses. Assuming most organizations now have an AI console, he asks to revisit whether it works or not.
Garg affirms that there must be a good balance of stakeholders on the business side, driving decisions and demand for AI, and a few people thinking outside the box on the tech side. The combination of these two skills drives a successful AI console, he notes.
The first task is enabling data by putting in the right data quality measures and metrics, says Garg. However, while that happens regardless of the effectiveness quotient, understanding the highest-value data map does not happen across organizations.
By data map, Garg refers to the highest-value use cases and the data needed to recognize the success of those use cases. He urges organizations to be methodical about prioritizing based on complexity and value gaps and then having a plan to drive that.
Next, Garg highlights the need for delineation between systems of AI experimentation and AI implementation when it comes to AI labs. According to him, people confuse the two, as there is a fair amount of overlap, but organizations must separate the two.
Adding on, Garg mentions working on DiscoverYai, a fully managed AI lab solution that could be taken as an organizational system of experimentation. The platform made the entire cycle auditable and made it clear for the business stakeholders to understand.
From things like data availability, having meaningful signals or patterns, solving the right problems, trust, explainability, and compliance with a responsible AI framework, the platform can contain all AI experiments.
After successful experimentation, organizations can take the steps into the implementation system with all the observability to ensure that it is put into production and realize its value.
Commenting on the delineation of the experimentation and implementation systems, Garg reveals that numerous AI use cases get built in different stages of implementation or experimentation. The separation is necessary because putting the traditional IT governance systems into the implementation system kills experimentation.
Furthermore, without a system for experimentation, people will continue working in silos and experiment without any results. To curb the issues arising out of that, Infocepts believes in separating the two activities while having systems for both, Gard concludes.
CDO Magazine appreciates Shashank Garg for sharing his insights with our global data community.