As artificial intelligence (AI) technology continues to evolve, a new form known as “Agentic AI” is changing the landscape of AI companions and task management.
In an interview, Kitman Cheung, technical and pre-sales engineering leader at IBM Asean, shed light on the transformative potential of agentic AI, which enhances the traditional generative AI by incorporating autonomy and decision-making capabilities.
While generative AI relies on prompts to produce relevant responses, agentic AI takes this a step further by utilizing logic, mathematical reasoning, and machine learning to extrapolate multiple options from a single command. Cheung explained that this advanced AI could effectively help the workforce streamline tasks and orchestrate workflows across various systems and platforms.
One of the features of agentic AI is its ability to comprehend complex objectives, break them down into manageable tasks, and execute them efficiently. Cheung illustrated this with an example of managing an appointment. Instead of just setting a single reminder, the AI agent would assess travel times based on the user’s current location, available modes of transport, necessary preparations, and other logistical considerations. It can book arrangements and set reminders based on the set task, functioning much like a human assistant.
In business environments, agentic AI promises to automate a multitude of tasks, from salary calculations to banking transactions. Cheung noted that while the integration of such technology would indeed transform company workflows, the human element will still play a crucial role. The aim is to handle more complex tasks before requiring human intervention, enabling employees to focus on higher-level responsibilities.
However, Cheung cautioned that the implementation of agentic AI should be approached with careful consideration, especially in sectors like banking where investment strategies can carry significant risks if managed solely by AI.
Conversely, automating tasks such as account applications, anomalous transactions detection, or monitoring credit scores presents numerous opportunities for effective AI application.
On the cybersecurity front, Cheung highlighted that agentic AI faces similar vulnerabilities as other AI models. To mitigate these risks, ongoing research and governance are essential.
He emphasized the advantages of using smaller agentic AI models with limited parameters, which help reduce the likelihood of data breaches and cyberattacks while also offering greater sustainability in terms of cost, efficiency, and environmental impact.
Overall, Cheung said technology enables humans to work faster and smarter, and the same goes for agentic AI. By offloading repetitive or menial tasks, employees at all levels can enhance their productivity.
He urged industry leaders to be judicious in selecting use cases and evaluating potential risks when integrating AI into their operations.
As part of IBM, Cheung states that the company treat its own organization as “client zero,” testing AI developments internally before rollout. The company has already embraced AI-assisted workflow orchestration and various other applications within its structure.
Looking ahead, the executive said agentic AI is anticipated to grant individuals greater freedom in their daily activities while providing enterprises with innovative options to grow and adapt in a rapidly changing business landscape.
