According to a survey by Salesforce, generative AI is being used by 73% of the Indian population, 49% of the Australian population, 45% of the US population, and 29% of the UK population. So, what comes next? It’s agentic AI.
This blog defines agentic AI, explains how it is different from generative AI, examines current adoption trends, and charts the challenges that may hamper the adoption of agentic AI. Still, there is much more to discover in our recent report: 'Agentic AI: next generation AI that works autonomously'.
Agentic AI refers to systems that can independently plan, reason, organise, and make decisions with little to no human input to achieve a user's goal. Unlike generative AI, which relies only on training data, agentic AI uses advanced tools and techniques to access real-time information. This allows it to handle complex tasks, break them into smaller subtasks, and adapt its approach as needed to reach the desired outcome.
AI agents are context-aware, meaning they can understand their environment and choose the necessary steps to complete a task effectively. They follow a modular architecture, with each module performing a specific function, allowing the system to work efficiently and intelligently.
Wonder how it works? Users simply define a goal and the agentic AI system breaks it down into manageable steps. Each module helps the agent perceive its environment, analyse data, make decisions, and learn continuously. Over time, this enables the system to improve its performance and expand its knowledge.
Digitalisation and operational transformation continue to accelerate, and businesses are increasingly focusing on improving efficiency by automating workflows and speeding up decision-making with agentic AI. In fact, a UiPath survey published in January 2025 revealed that 37% of respondents are already using agentic AI and about 45% plan to invest in it this year. This growing interest is driven by the tangible benefits it offers, such as better oversight of workflows, improved integration between systems, automation of complex tasks, and faster response times for customers.
Agentic AI plays a crucial role in these processes. It can support a wide range of applications across industries and the companies that are adopting this technology often use it to streamline routine tasks and boost productivity. For instance, in legal and human resource contexts it can be used to summarise and analyse documents. It can also be offered to end-user adopters as an 'Agent as a Service' model, where businesses pay based on results rather than fixed subscription fees.
For example, Taiwan-based Super 8 Studio has provided its Agentic AI-as-a-Service (AaaS) platform to Shengsheng Youdong, a clinic which was facing issues with manually answering patients' appointment calls, flipping through calendars, and confirming appointments. The clinic started using used Super 8 Studio’s MessageHero platform, which can determine customer intent, provide responses that are close to the brand tone, and support interactive content types such as picture and text cards, product recommendations, and reservations, optimising the interactive experience and improving conversion efficiency and customer value. After deploying the platform, Shengshen Youdong achieved lower appointment error rates, and increased rates of patients revisiting the clinic.
Agentic AI is being increasingly adopted by companies across a range of functional areas, including sales and marketing, customer support, cybersecurity, research and development, and finance. For instance, Aqua is using an agentic AI solution to eliminate Tier-1 and Tier-2 cyberthreats and to offer better cybersecurity visibility and improved threat investigation and Owkin is developing a network of AI agents that can autonomously access scientific literature and large-scale biomedical datasets while using its proprietary discovery engines. These agents analyse data and plan and execute experiments in an automated environment. Moreover, Finland-based VTT is using Salesforce’s Agentforce to drive its sales outreach activities by managing sales prospect’s queries and OpenTable is using an AI agent solution to provide real-time customer support to diners in restaurants by handling their customer queries; and Bud Financial uses agentic AI to deliver a personalised customer experience with autonomous finance management.
Despite the numerous benefits of agentic AI, there are certain challenges which may hamper its adoption including in terms of regulations, transparency and error handling.
Current and emerging AI regulations have primarily been developed for traditional and generative AI systems and often fall short when applied to self-learning agentic AI systems. Therefore, regulators, developers, and users face new and complex challenges, emphasising the need for more flexible and responsive regulatory frameworks.
Agentic AI is still in its early stages and businesses are cautious due to the potential risks posed by autonomous systems. The decisions taken by an agentic AI are often influenced by its environment, making the process less transparent. This raises the question as to who is responsible if such a system makes a poor decision – the developer or the user?
A major concern with many agentic AI systems is their inability to deal with incorrect or illogical outputs, especially in multi-agent setups where one agent’s output error can spread to others as an input error, which in turn can lead to poor decisions and significant business risks.
Agentic AI represents a significant leap forward from traditional AI systems, offering greater autonomy, adaptability, and real-time decision-making capabilities. As adoption grows across industries, it promises to drive efficiency and innovation. However, addressing challenges like regulation, transparency, and reliability will be key to unlocking its full potential.
