AI Drives Transition from Support Role to Strategic Leadership

Executives in IT and business sectors recognize the treasure trove of data they possess. While conventional tools like business intelligence platforms and statistical software can uncover insights within this data, achieving this quickly, in real-time, and at large scale still presents significant challenges.

When implemented effectively and responsibly, Enterprise AI can transform these challenges into prospects. This technology excels at rapid data processing, even in real-time scenarios like during customer interactions. AI's scalability allows it to manage vast information from diverse sources effortlessly, similar to handling a simple one-page spreadsheet.

However, integrating AI into modern businesses is no easy feat. It requires structure, trust, and skilled personnel. Beyond practical difficulties, AI brings its own challenges, including data governance, setting boundaries for AI outputs and training data, and ongoing staffing issues.

We spoke with Rani Radhakrishnan, PwC Principal for Technology Managed Services in AI, Data Analytics, and Insights, about what is working and what is hindering CIOs in their AI initiatives. Our conversation took place before her keynote at the TechEx AI & Big Data Expo North America, scheduled for June 4 and 5 at the Santa Clara Convention Center.

Rani is particularly sensitive to governance, data privacy, and sovereignty challenges faced by enterprises. Her extensive experience with clients in the health sector highlights the critical nature of issues like privacy, data oversight, and accuracy in successful technology deployments.

"Having just a prompt engineer or a Python developer is insufficient. It's essential to have human input to curate appropriate training datasets and address any biases," Rani Radhakrishnan of PwC remarked.

Rani noted that there is a rising demand from PwC's clients for AI-driven managed services that deliver business insights across sectors and for AI's proactive use in roles where technologies autonomously act on data and user input while interacting with humans and data resources.

For instance, PwC's Agent OS is a modular AI system that integrates systems and escalates intelligent agents into workflows with remarkable speed, surpassing traditional computing methods. This exemplifies PwC's response to clients' growing interest in AI while acknowledging the lack of in-house expertise to address these needs.

Interest in AI comes from various organizational sectors. Examples include proactive monitoring of physical or digital systems, predictive maintenance in manufacturing, or automation-led cost efficiencies in complex, customer-facing environments.

Despite AI's potential value, most companies lack the diverse range of skills and personnel required for successful AI deployment—at least, deployments that yield ROI without significant risk.

"A prompt engineer or a Python developer alone isn't enough," Rani stated. "A structured approach is needed, with human involvement to curate appropriate training datasets and address bias issues."

Successful AI application requires a blend of technical skills—data engineering, data science, prompt engineering—alongside domain expertise. Domain experts define optimal outcomes, while technical teams ensure responsible AI practices, proper data handling, and adherence to company guidelines.

"To derive maximum value from AI, organizations must ensure accurate underlying data," Rani emphasized. "I know of no company that claims its data is perfect... Properly structuring and normalizing data is crucial for meaningful querying, analysis, annotation, and trend identification."

Enterprises must actively monitor and rectify bias within AI systems and the analysis of training and operational data to effectively leverage AI.

A stringent data cleaning and annotation process within AI system architecture is crucial. This requires significant human effort, Rani noted, with emerging professionals skilled in these tasks becoming increasingly important.

When data and personnel challenges are addressed, feedback loops enhance the value of generative AI, Rani explained. "AI prompts offer the chance to refine responses, training the model to deliver desired answers, making it uniquely valuable."

For CIOs, the transition involves not only technical integration but harmonizing AI with enterprise architecture, aligning with strategic goals, and managing associated governance risks. CIOs are evolving into AI stewards, architecting trust and transformation within systems.

AI's rise from academic research to practical application is recent, so it's natural for organizations to navigate their way toward realizing AI's full potential.

A new framework is emerging, enabling CIOs to unlock the potential of their data across business strategy, operational enhancements, customer interactions, and numerous other domains.