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

Businesses adopting artificial intelligence (AI) should start with a measurable goal and judge success by the improvement it delivers rather than the number of AI agents deployed, as the technology moves from helping employees to doing work for them, a senior Oracle executive said.

"It’s important to remember that the number of agents deployed matters less than the improvement they deliver," Natalia Rachelson, Senior Vice President, Applications Development at Oracle, told PTI.

Business leaders should pick an outcome they can measure, such as reducing the time spent resolving accounting errors or lowering transport costs, she said.

"It’s important to establish a baseline first so you can assess whether AI is making a meaningful difference," she noted.

Companies then need to get the basics right: reliable data, clear ownership of processes, appropriate controls, and employees who understand how to work with AI agents. The biggest gains come from parts of a process where work gets delayed or needs repeated manual intervention, she said.

The biggest misconception business leaders have, she said, is that AI is about doing the same work faster.

"The biggest misconception is when people think that AI is about doing the same things faster, better, and cheaper. It is not. It is really about redefining how work gets done...If instead of 10,000 steps, you’re now down to five steps because AI does everything in the middle, that's a redefinition of a business process," she said.

Earlier AI tools focused largely on answering questions, drafting content and making recommendations, Rachelson said. Newer AI agents can coordinate the next steps and carry out authorised actions within a business process.

A company aiming to raise its customer renewal rate by 2 percent, for instance, could set it as a goal, and a team of specialised AI agents would analyse the data, come up with a plan and carry it out, she said.

In a hospital system, AI could schedule nurses with different specialisations and shifts. If a nurse calls in sick, it can redraw the schedule and alert the people who will fill in, she said.

The quickest results are likely in routine processes where teams spend a lot of time gathering information and chasing follow-ups, she said. In finance, AI agents can help reconcile accounts before the books are closed. In human resources, they can flag potential payroll issues. In supply chains, they can compare shipping options or work out how to allocate scarce materials, with teams reviewing and approving the plans.

People remain in charge throughout, Rachelson said.

"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."

Rachelson, who has been with Oracle for about 15 years, said AI was a bigger shift than the move to cloud computing, but the questions it raises are familiar.

"AI is bigger than the cloud, but it’s no different. Every technology transition raises a lot of questions and concerns," she said.

She recalled the fears around Y2K, the computer date bug at the turn of the millennium, when "the world was coming to an end and then absolutely nothing happened".