Robotic process automation is dead.
For more than a decade, robotic process automation (RPA) was the poster child of automation. It delivered solid value: it automated repetitive tasks, reduced manual effort, and bought organisations time while they modernised core systems. But today, a fundamental truth is unavoidable: RPA as a strategic automation paradigm is dead.
This is according to Deon van Niekerk, CTO at Ovations Technologies, who says that while organisations need automation more than ever, RPA is dead because the enterprise has outgrown what scripted bots can do.
He says: “A far more intelligent and scalable model has emerged: agentic AI, powered by existing systems with the help of Model Context Protocol (MCP) interfaces, grounded with policy documents, business rule engines, and governed by headless BPM engines.”
Why RPA has reached the end of its lifecycle
Van Niekerk says recent Gartner research reflects the industry transformation from deterministic bots to intelligent orchestrated automation. “Gartner says RPA has moved into the Plateau of Productivity – it is valuable, but no longer strategic. At the same time, AI agents sit at the Peak of Inflated Expectations and are seen as the next major enterprise automation category. Gartner warns of ‘agent washing’ but confirms a trajectory toward autonomous, goal-driven digital workers, and predicts that by 2028, 15% of enterprise decisions will be made autonomously by agents, up from effectively 0% in 2024,” he says.
He believes RPA has reached the end of its lifecycle for several reasons.
“Firstly, RPA is brittle by design. RPA bots operate through deterministic scripts: click here, copy this, paste that. If you change a UI element, data structure, or process flow, the bot fails. This fragility makes scaling difficult and maintenance expensive,” he says.
“Another reason is that RPA cannot understand or reason. Modern business processes include unstructured documents, e-mails, exceptions, negotiations and choices that require judgement. RPA cannot interpret content, understand context, make reasoning decisions or handle ambiguity. It simply executes pre-programmed instructions.”
Van Niekerk also notes that RPA scales linearly, not exponentially. “Every new bot adds new risk, new maintenance, new exceptions, and new operational overheads. At scale, organisations end up with a hundred points of automation instead of a truly automated enterprise,” he says.
RPA was built as a workaround, emerging because systems lacked APIs, van Niekerk says. “Today, mature platforms, event-driven architectures, and standards like MCP make UI-mimicking automation obsolete,” he says.
He adds: “The modern enterprise needs better outcomes, not just task automation. Modern operations demand intelligent decision-making, dynamic adaptation, unstructured data processing, end-to-end visibility, and governed autonomy. RPA cannot deliver this. RPA isn’t failing because it’s broken – it’s failing because the world changed. Trying to fix RPA tools now will just add even more complexity.”
Another compelling reason for the move from RPA to agentic AI is the resources required for RPA, Van Niekerk says. He explains that RPA is costly, it requires infrastructure, and staff need to be trained. With AI, you can only pay for what you use, using tokens or resource units. In addition, a lot of the AI models are free and open source.
However, he does not believe RPA will disappear overnight. “RPA is no longer the core of enterprise automation. It is a legacy technique inside a far broader, AI-first ecosystem. However, it is still useful when the UI is stable, the task is simple, the process is deterministic, where APIs don’t exist and where reliability is more important than intelligence,” he says.
The new automation stack: Agentic AI ++
Van Niekerk says a modern automation strategy moves away from bots performing tasks toward agents delivering outcomes.
“Agentic AI offers automation that thinks, not follows. Agentic AI systems use LLM-powered reasoning to understand goals, plan sequences of actions, interpret documents and context, navigate systems dynamically, manage exceptions, and learn from feedback,” he says. “You no longer script the steps. You define the outcome and the agent determines how to achieve it.”
As part of this evolution, the Model Context Protocol (MCP) replaces brittle REST glue code and eliminates UI-driven automation.
He explains: “MCP allows agents to safely execute structured actions, read and write data deterministically, discover available tools and capabilities, and interact with enterprise systems without custom integration code. MCP is effectively USB-C for AI, enabling agents to connect seamlessly to any system.”
Agentic AI needs structure and oversight, Van Niekerk notes. “Headless workflow engines provide state management, auditability, exception routing, long-running process control, clean orchestration across humans, systems and agents. If agentic AI is the brain, and MCP is the connectivity, the headless workflow engine is the nervous system that ensures every action is governed, tracked, and compliant,” he says.
The enterprise imperative: Pivot now
Ovations sees this as automation reinvented, not just upgraded RPA, Van Niekerk says.
“We believe that organisations that stay on an RPA-centric path will fall behind, and that South African and Pan-African companies need to be moving fast to build their own AI infrastructure and create their own AI capability. Those that shift to agent-first automation will gain exponential scalability, end-to-end process automation, dramatically lower maintenance costs and true hyper automation,” he says.
This new “AI-first” approach is the foundation of Ovations Technologies’ new agentic AI value proposition and separate from the traditional “process first” Enterprise Hyper Automation value proposition. “We realised that AI – and AI governance – had to be at the heart of solutions, not something you could add in later. This new approach is resonating with our customers. Ovations are helping customers think through the AI use cases, help construct the business case, create their AI strategy, and establish the governance, platforms and techniques to achieve optimal value from Agentic AI,” Van Niekerk says.
