Measure AI by results, not number of agents deployed: Oracle senior executive

Businesses implementing artificial intelligence (AI) should begin with a quantifiable objective and evaluate success based on the enhancement it generates rather than the count of AI agents employed, according to Natalia Rachelson, Senior Vice President of Applications Development at Oracle. Rachelson stressed that the number of agents deployed holds less significance than the improvement they generate.

Business leaders must select a measurable outcome, such as decreasing the time taken to address accounting errors or diminishing transportation expenses, she advised. Establishing a starting point is crucial to determine if AI is making a significant impact, she noted. Companies must also ensure the basics are in place: dependable data, clear process ownership, suitable controls, and employees with the ability to collaborate with AI agents.

The most substantial advancements arise from process segments where work experiences delays or necessitates repeated manual intervention, Rachelson explained. A prevalent misunderstanding among business leaders is that AI entails performing the same tasks more rapidly. The actual misconception is that AI pertains to redefining how work is executed, not merely completing tasks faster, better, and more cost-effectively.

Earlier AI tools predominantly concentrated on answering inquiries, creating content, and suggesting recommendations. Contemporary AI agents, however, can oversee subsequent actions and execute sanctioned activities within a business process. For example, a firm aiming to boost its customer renewal rate by 2 percent could establish this as a target, and a team of specialized AI agents would analyze data, devise a strategy, and execute it.

In a healthcare institution, AI could schedule nurses with diverse specializations and shifts. If a nurse calls in sick, the system can redesign the schedule and notify the individuals who will substitute. The most rapid outcomes are typically observed in routine processes wherein teams spend considerable time collecting information and following up.

In finance, AI agents can assist in reconciling accounts prior to the books being closed. In human resources, they can flag potential payroll concerns. In supply chains, they can compare shipping options or determine how to allocate limited resources, with teams reviewing and approving the plans. Human intervention remains essential throughout, Rachelson emphasized.

Enterprises must delineate the tasks an agent can perform, how its work will be verified, and when human approval or intervention is required. Rachelson, who has been with Oracle for approximately 15 years, compared the shift to AI with the transition to cloud computing. While AI represents a more substantial change, the inquiries it raises are familiar.