Scaling Generative AI and Overcoming Operational Challenges: Insights from the AI Impact Forum

Enterprise leaders are moving beyond testing generative AI and are now focusing on scaling it. Balkrishan “BK” Kalra, president and CEO of Genpact, asserts that this shift marks the end of the experimental phase. Speaking at Newsweek’s “AI Impact Forum” webinar, he emphasized the need for companies to transition from proof of concepts to scale use cases.

Scaling efforts bring unresolved challenges to the forefront. Data must be usable across various business units, processes need strengthening, employees require fluency with AI tools, and accountability must be maintained when AI agents act autonomously. Kalra underlines the importance of measuring scale against business performance, focusing on growth, efficiency, and cash conversion.

As AI moves from testing to operational systems, the complexity of integration increases. Research by Genpact and HFS Research, surveying 2,002 executives across multiple industries, indicates that only 6 percent qualify as effective debt re-mediators, able to address and measure outcomes of technology improvements.

Beneath technology debt, Kalra identifies data debt, process debt, and talent debt as hindrances to AI efficiency. Agents rely on specific data and context to function, while business processes vary widely and require enterprise-specific knowledge to navigate.

Kalra illustrates these challenges with global operations, where jurisdictional requirements vary. Much of the needed knowledge resides within enterprises, often undocumented, in workflows and processes. Thus, artificial intelligence needs to integrate with process intelligence to be effective.

IT and governance discussions should start early. Kalra advises involving CIOs and CDOs from the outset to garner buy-in for AI solutions. Security needs, alongside a responsible AI framework, are crucial as agents take on roles within finance, supply chains, and operations, where errors can have regulatory implications.

Agentic operations are evolving from human-processed and validated work to machine-processed tasks with human oversight. While agents execute more functions, humans remain responsible for exceptions. Workforce readiness is vital, requiring workers to engage with AI tools to redesign work processes effectively.

Tinaikar stresses the importance of allowing employees time to develop new skills, competing with daily responsibilities. Kalra categorizes skills at Genpact into two groups: AI builders and AI practitioners, each combining domain expertise with technical knowledge.

With machines handling more execution, roles will shift. Tinaikar draws parallels with the smartphone era, highlighting how technology creates new ecosystems and opportunities, not just displacement. Kalra agrees that new technologies foster new business models and roles.

He invokes Jevons paradox, suggesting that efficiency enhancements can lead to increased activity and demand. His immediate warning is for workers: AI won’t take jobs, but those unskilled in AI might. Greater automation requires data, processes, security, and skilled employees.

Kalra concludes with a straightforward observation about aspirations versus readiness in adopting AI technologies. Upcoming AI Impact Forums will continue exploring these topics, as noted by Dr. Ranjit Tinaikar, who will host a webinar with Firdaus Bhathena on October 22.