Enterprise Leaders Transition from Generative AI Experiments to Scaled Applications

Enterprise leaders have spent significant time exploring the possibilities of generative AI. Balkrishan “BK” Kalra, Genpact’s president and CEO, expressed during the AI Impact Forum on September 17 that the era of experimentation is concluding. Kalra stated, “Time of proof of concepts is gone. Time of experiments is gone. It is now scale use cases.”

Moving from testing phases to scaled applications pushes companies to address challenges left unresolved by proof of concept tests. This includes ensuring data usability, maintaining processes across different business units, achieving employee fluency with AI tools, and ensuring accountability when AI agents act.

Kalra emphasized evaluating scale use cases by how they impact business performance, focusing on whether enterprises can “grow faster, operate leaner or convert more to cash.” These objectives become more challenging when AI goes beyond controlled tests and integrates into everyday systems, data, and governance rules.

Dr. Ranjit Tinaikar cited a Genpact and HFS Research survey of 2,002 enterprise executives across 16 industries and 14 functions. The study found only 6 percent of organizations qualified as effective debt re-mediators. Tinaikar asked Kalra what management teams should address before advancing further into agentic operations.

Kalra discussed technology debt, describing it as the accumulation of outdated systems and infrastructure. He noted that beneath this lie data debt, process debt, and talent debt, which can limit AI capabilities. AI agents expose these weaknesses since they rely on usable data and company-specific context.

Business processes add complexity, as what appears to be a single process often contains different rules and practices across regions or business units. Kalra illustrated this using global operations, where requirements vary by jurisdiction. Such knowledge rests within the enterprise, sometimes undocumented in employee handling of exceptions.

“There is no artificial intelligence, no gains from artificial intelligence, if it is not coupled with process intelligence,” Kalra asserted.

IT and governance discussions need to start early. Kalra emphasized involving key players from the outset, suggesting companies bring in CIOs or CDOs early to secure their buy-in.

Security concerns must be addressed alongside a responsible AI framework, as AI agents execute tasks in sensitive areas like finance, posing potential financial or regulatory risks.

Kalra explained agentic operations progress from human-processed and human-validated work to increasingly machine-processed and human-validated processes. Humans remain accountable for exceptions even as agents undertake more execution tasks.

Workforce readiness is vital. Employees need exposure to AI tools to grasp capabilities, aiding meaningful involvement in redesigning work processes. At Genpact, thousands of workers now access these tools.

Tinaikar linked workforce development to the timespan allowed for honing skills. Traditionally, training competed with daily priorities. He stressed training as “not a privilege, it is an imperative.”

Kalra discussed two categories for skill development within Genpact. AI builders merge technical skills with business domain knowledge. AI practitioners bring expertise in sectors such as finance or banking while developing enough command of AI and data to engage directly with technology.

Task and role transformation will follow machines undertaking more execution work. Tinaikar referenced the smartphone era to illustrate how anticipated displacement doesn’t always occur; new ecosystems can develop around emerging technologies.

“When the mobile phones came, they thought the laptops and the PCs would become obsolete,” Tinaikar stated. “Instead, it created a whole new ecosystem of apps on mobile phones.”

Kalra acknowledged that new business models and roles arise with new technologies, referencing Jevons paradox, where improved efficiency increases overall usage, sparking additional demand.

Kalra’s immediate caution for workers was clear. He stated, “Your job will not be taken by AI, but your job can be taken by somebody who knows AI better.” Expanded automation remains contingent on usable data, functional processes, robust security measures, and employees ready for different operational dynamics.

“Aspirations are really high,” Kalra concluded, “Readiness is low.”

The next AI Impact Forum webinar featuring Dr. Ranjit Tinaikar will occur on October 22. He will engage with Firdaus Bhathena of S&P Global about enterprise technology, AI, and organizational changes needed to translate capabilities into business results. Register today to attend for free.