Leadership
Oracle's Rachelson: measure AI by results, not agent count
Oracle's Natalia Rachelson says AI redefines how work gets done. Her example: a 10,000-step process that could fall to five.
Oracle's senior vice president for applications development, Natalia Rachelson, told PTI that companies should judge AI agents by the improvement they deliver, not by how many they deploy. Business Standard carried her remarks on 15 November 2026, and The Hindu BusinessLine also reported them. "It's important to remember that the number of agents deployed matters less than the improvement they deliver," she said.
Her advice starts before deployment. According to Business Standard and The Hindu BusinessLine, firms should set a measurable goal and a baseline first. They also need reliable data, clear process ownership, appropriate controls and staff who can work with agents. YourStory reported example outcomes, including less time spent on accounting errors and lower transport costs.
Rachelson also rejects the idea that AI simply does today's work faster. Business Standard and The Hindu BusinessLine both reported her view that AI means redefining how work gets done. Her example is a hypothetical process of 10,000 steps that could drop to five, with AI doing everything in the middle. Earlier tools, she said, answered questions, drafted content and made recommendations. Newer agents, she said, can organise what comes next and act within limits they are given.
The use cases she named span several functions. In finance, agents reconcile accounts before books close, and in HR they flag potential payroll issues. In a hospital, AI could schedule nurses by specialisation and shift, and could redraw the schedule when someone calls in sick. Business Standard reported that the quickest results are expected in routine processes heavy with information gathering and follow-ups.
On governance, Rachelson was direct. "Enterprises need to define what an agent is allowed to do, how its work will be checked, and when a person needs to approve or intervene," she said. She has spent about 15 years at Oracle and says AI is a bigger shift than the move to cloud. She compares today's worries with Y2K date-bug fears, whose predicted catastrophe did not happen.
Why it matters
Rachelson's advice is that measurement comes before deployment. A firm without a baseline cannot tell whether an agent has paid off. Her redefinition argument points, in our reading, towards redesigning processes rather than adding tools. The uses she named are in finance, HR, supply chains and healthcare, and her call for defined permissions and human sign-off points is something leaders would need to plan for before rollout.
Sources
Checked against 3 outlets
- Business Standard — Rachelson's PTI interview remarks, quotes, publication date, examples of agent use cases and capability shift
- The Hindu BusinessLine — Corroboration of Rachelson's measurement advice, baseline and readiness conditions, process redefinition example, and governance remarks
- YourStory — Example outcomes (accounting error resolution time, transport costs) and measurement advice