Companies need clear goals, reliable data and human oversight as AI moves from assisting employees to executing business tasks.
Highlights:
- Oracle urges businesses to set measurable goals before adopting AI.
- A performance baseline is essential to assess improvements.
- AI agents can execute authorised tasks across business processes.
- Finance, HR and supply chains offer opportunities for early gains.
- Clear controls and human approval remain essential.
New Delhi, Nov 15: Businesses adopting artificial intelligence should measure success by the improvements it delivers rather than the number of AI agents deployed, according to Natalia Rachelson, Senior Vice President, Applications Development at Oracle.
Companies should begin with a measurable business goal, such as reducing the time needed to resolve accounting errors or lowering transportation costs. Establishing a baseline before deployment is essential to determine whether AI is making a meaningful difference, she said.
Rachelson said successful adoption also requires reliable data, clear ownership of business processes, appropriate controls and employees who understand how to work with AI agents. Processes repeatedly delayed by manual intervention offer some of the biggest opportunities for improvement.
She argued that business leaders often misunderstand AI as simply a tool for completing existing tasks faster and more cheaply. Its larger potential lies in changing how work is organised, including reducing lengthy workflows to a small number of steps by handling intermediate tasks.
Earlier AI tools primarily answered questions, drafted content and offered recommendations. Newer AI agents can coordinate subsequent steps and carry out authorised actions within a business process, she said.
For example, a company targeting a two-percentage-point improvement in customer renewals could use specialised agents to analyse information, develop a plan and execute it. In a hospital, AI could help schedule nurses across specialisations and shifts, revise rosters when someone calls in sick and notify replacement staff.
The quickest results are likely in routine processes where employees spend substantial time collecting information and chasing follow-ups. In finance, agents can support account reconciliation before books are closed. In human resources, they can identify potential payroll problems.
Supply-chain teams can use AI to compare shipping options or plan the allocation of scarce materials, while retaining responsibility for reviewing and approving proposed actions.
Rachelson emphasised that people must remain in control. Enterprises need to specify what agents can do, how their work will be checked and when human approval or intervention is required.
Describing AI as a bigger transition than cloud computing, she said the questions and concerns surrounding it are familiar features of major technology shifts. The priority for businesses, she indicated, should remain demonstrable improvements backed by clear accountability.
