Agentic AI: People and organisations at the crossroads

Ten years ago, I worked as the Chief Commercial Officer of a mid-size software company developing solutions for the BFSI sector. The word fintech was not so popular back then. RPA or Robotic Process Automation software was introduced in Sri Lanka during this time, and we opted to be a distributor for one of the three pioneering brands entered the market namely Blue Prism, UI Path and Automation Anywhere. RPA was ideal to automate processes of banks, insurance firms, and state authorities such as passport and border control handling large volumes of repetitive and rule-based tasks.  Fast forward, we now have AI enabled RPA which is an evolutionary leap beyond RPA and popularly known as Agentic AI that is going to make a big revolution in every industry across the world. This article aims to explain benefits, challenges and strategic directions for managers and business firms who will adopt Agentic AI in automating their workflows. 

The profound use of AI 

AI has now become an everyday tool for most of the digitally literate individuals and tech savvy firms. Students, researchers, and employees of business firms use AI in their work with varying degrees of consent, competence and confusion. Many business firms allow employees to use AI as it can substantially improve productivity. Hours of coding work done by software engineers now happen in minutes. Teachers caution poor analytical skills and creativity among children due to heavy reliance on AI. Discussions between service providers and customers such as doctors and patients are becoming short and precise. Similarly, people are afraid of losing jobs as AI has become explosive filled with promises and fears.

Convergence of automation and AI 

RPAs were good in doing tasks but lacked thinking capability. AI particularly the LLMs or Large Language Models provide the ability to understand, logical reasoning, and plan. In other words, AI provides the brain. RPAs provide the ability to execute functions such as clicking a button, copying and pasting etc. The convergence of RPA and AI has given us real digital workers known as AI agents who can work like human employees. Agentic AI which comprises of multiple agents, can interact with different applications, manipulate data, manage and control hardware and execute tasks to achieve specific goals. Some argue that humans will work for machines in future. Humans losing control on fully autonomous systems can bring unprecedented threats to the world. Think of an era where power generation, irrigation management, air traffic control, border control etc. happen with no human intervention. Given these capabilities of Agentic AI, you may now think of the fate of BPO and KPO operations, service-oriented organisations and the back-office functions of top brands outsourced from developing countries. 

Five levels of automation

The popular 2025 book by Pascal Bornet and his team on Agentic AI highlights its progress with five levels of automation. Level 1 describes rule-based automation, and this is what we had as RPA. Level 2 refers to intelligent automation. Levels 3 to 5 describe agentic workflows, semi-autonomous and fully autonomous agentic systems respectively. They emphasise the importance of starting automation at a low level in your organisation as it will enable you to be familiar with it, have control and governance mechanisms in place. This also shows the low control managers will have at higher levels of agentic automation and less instructions humans need to give.  The Cybercab by Tesla that has no steering wheel, pedals and mirrors is probably the latest and popular example for an autonomous agentic system.

Innovations in the era of Agentic AI

Unlike inventions, innovations need not to be grand discoveries and brand new to the world. If there is any sensible change, we make to a product or a process that can add value to stakeholders we can explain it as an innovation. It is stunning to see what Agentic AI can do in the wide array of innovation types and processes popularly in use such as incremental, radical, architectural, modular, disruptive, reverse, frugal, outcome driven and open innovation. As Agentic AI learns and improves what it does in each task, what we can expect is a dramatic push in all types of innovations. It is true that creativity is central to invent and innovate and we often tend to question the ability of machines to think creatively. Despite the algorithms and lines of codes or instructions they follow, machines are now capable of thinking. This flexibility in controls may lead to continuous incremental innovations in products and processes. We as human beings can have some protection until AI powered machines explore the exciting and uncertain world of disruptive innovations.  Some already highlight the enormous threats the world may face with super intelligence.

Dynamic capability embedded to DNA 

David Teece coined the term dynamic capability to remind organisations of the need to upgrade and update their capabilities in response to changes in the business environment. His latest articles and videos published in 2025 reemphasise the need of reconfiguring the capabilities of an organisation to remain competitive. Managers of all levels of a business firm, especially at the middle level, have this challenge of managing dynamic capabilities to ensure the organisation has an edge over competition and to accelerate innovations. Agentic AI being a practice that necessarily relies on learning from mistakes and constant learning for improvements, the future firms that are empowered with Agentic AI will have dynamic capabilities embedded to its DNA. This is no different to how culture is infused into the blood of an organisation.  

Organisational inertia to fade away

People and organisations generally resist change. The preference to maintain the current equilibrium in the organisation known as the status-quo and rigid processes and procedures lead to organisational inertia that prevents change or makes it challenging. System integration is identified as the most challenging area business firms face when introducing AI Agents. Significant resistance may come from the technical people as the remedial measure involves ‘API first’ architecture. However, with more AI agents in use by business firms, managers may expect low inertia in future. One can argue that the level 3 Agentic workflows which is right now the most seen level of agentic automation will certainly result in low resistance to change.  

Sustainability 

AI and Agentic AI require large amounts of resources. As highlighted in the AI Index Report 2026 by Stanford University, this is an industry which is scaling faster than surrounding systems around can adopt. Research and practices that aim to optimise resource utilisation include and are not limited to edge computing, fog computing, AI accelerators etc. Bio-inspired swarm intelligence and socially inspired intelligence are also popular areas for research. Increasing energy and water footprints, especially at data centers have become a major concern for governments and other interest groups. Given the speed of growth of AI, it is becoming important to monitor and regulate its purpose of use, the scope and cost before it is too late and irreversible. 

Need for policy imperatives 

We often see in local media multiple initiatives by state authorities and think tanks on the way forward to a digital economy. Given that as a key goal in the national agenda, artificial intelligence particularly Agentic AI should necessarily be a focus area by policy makers and managers. Attempting to be competitive in regional or global markets without leveraging on agentic AI as a value driver will certainly be a costly lesson for everyone. 

Once again, we prove that we are quite slow in technology adoption as a nation in this massive transformation as well. We are behind many other nations, as highlighted in the AI diffusion report 2025 by Microsoft AI Economy Institute. State authorities, universities and business firms have a major role to play in increasing AI adoption for productive purposes.

Ethical perspectives in Agentic AI 

Top managers will face extremely challenging situations in adopting Agentic AI. The popular utilitarian way of deciding by looking at the consequences of a decision where pleasure for majority and pain for minority is the accepted norm cannot be easily applied here. One might argue that Agentic AI can enhance productivity and the broader society will benefit. Yet, at what cost is the question one needs to answer by looking at multiple stakeholders. A regular question I asked in my MBA classes for many years is who will need driverless cars and at what cost. Agentic AI systems are no different. Cost here is not only the cost of ownership and operations but also the social cost, impact on the environment, use of depleting resources and hidden losses in upstream and downstream activities. Surely, the excitement of a new technology cannot be at the cost of losing your job, water to drink and electricity for daily use. Hypercompetition may prompt business firms to resort to Agentic AI to get ahead. Managers need to think carefully about the quality of the decisions they make and the consequences, especially in the long run.

(The author is a Senior Academic Fellow attached to SLIIT Business School. He is a Fellow of the British Computer Society, and holds chartered/corporate memberships of nine other professional institutions in mechanical engineering, computing, psychology, marketing, logistics and innovation.  He is an ex-banker with several decades of experience in diverse industry sectors. He has been teaching organisational behavior, strategy and innovation management for over 25 years in postgraduate programs of 25 local and international universities. He could be reached via email [email protected])