RPA is not dying, it is becoming the infrastructure layer
Why the automation of the future is neither pure RPA nor pure AI – and what the new platform generations from UiPath and Microsoft mean for enterprises.
In few technology debates are so many false claims made so confidently as on the question of whether RPA is dead. The short answer: no. The longer one: RPA is not disappearing; it is moving one layer deeper. What comes on top are AI agents that think, while the bots continue to click. 2026 is the year in which this thesis becomes platform reality.
| 60 % | 5,2 Mrd. $ | 40 % |
| ENTERPRISE APPS STILL EXCL. RPA Gartner, through end of 2026 | RPA MARKET VOLUME 2025 Forecast: $35bn by 2030 | LESS TCO Deloitte, when switching to hybrid architecture |
The wrong question: RPA or AI?
Since the availability of powerful language models, predictions about the end of robotic process automation have been piling up. The logic sounds compelling: if an agent can understand text, make decisions and operate systems, why would you still need a bot that rigidly executes a script?
The answer comes from practice. Peter Hecker, author of the German specialist blog Hecker Consulting, analyses in a comprehensive comparison from January 2026 why the either-or question is misleading. His central finding: LLM-based UI agents achieve success rates of up to 90 percent in the lab – in production environments without RPA, they drop below 50 percent. At the same time, Gartner warns that over 40 percent of projects with autonomous AI agents could fail by 2027.
| “AI is the brain and RPA is the hand. The future belongs to hybrid approaches.” — Peter Hecker, Hecker Consulting, January 2026 |
The implication is clear: RPA remains indispensable for deterministic, business-critical processes in regulated environments, wherever auditability, reliability and traceability are non-negotiable. AI agents take over where interpretation, contextual understanding and adaptive decisions are required. The future lies not in replacement but in layering.
Platform reality 2026: UiPath and Microsoft
What was previously a theory becomes a platform reality in spring 2026. Two of the world’s most influential automation platforms define the architecture of the next wave of automation. The two providers approach the same goal from very different directions.
UiPath introduced Maestro, an orchestration engine that coordinates bots, AI agents and human approvals in an end-to-end workflow. Maestro is deliberately designed to be endpoint-agnostic, it also orchestrates systems built outside of UiPath – and supports so-called Durable Execution: when a process is interrupted, the platform resumes exactly where it left off.
Microsoft counters with Agent 365, which is being expanded into the central orchestration platform for agentic automation with the 2026 Release Wave 1 (from April 2026). Multi-agent orchestration becomes generally available: a parent agent delegates tasks to specialised child agents, passes context between them and consolidates the results. Via the open Agent-to-Agent Protocol (A2A) and the Microsoft 365 Agents SDK, Copilot Studio agents communicate directly with agents from other systems, including cross-vendor. Power Automate continues to provide the RPA execution layer with new self-healing capabilities for Desktop Flows, Microsoft Fabric the data foundation, and a new Model Context Protocol (MCP) Server connects agents, flows and tools across the entire Microsoft ecosystem.
The differences matter. UiPath starts from an existing RPA landscape and extends it with an agentic layer – ideal for enterprises that already have hundreds of bots in production. Microsoft focuses on integration into the business applications already in use: Teams, Outlook, Dynamics, Power Apps. Where employees do their work, that is where the agents should live.
| Dimension | UiPath Maestro | Microsoft Copilot Studio |
| Launch | 2025 | 2026 Release Wave 1, ab April 2026 |
| Core architecture | Endpoint-agnostic orchestration with Durable Execution | Copilot Studio as agent platform, Power Automate as RPA engine, integrated via MCP |
| Agent focus | Coding agents natively embedded, Claude Code and Codex supported | Multi-agent orchestration in Copilot Studio, self-healing for Desktop Flows |
| Governance | Agent guardrails, content moderation, EU AI Act compliant | Real-time risk assessment, governance agents, agent feed for human oversight |
| Typical target audience | Enterprises with existing RPA landscape looking to scale agentically | Organisations with M365 stack embedding automation where work happens |
The real paradigm shift: agents that build automations
The strategically most important announcement came on 6 April 2026. In a virtual fireside chat, UiPath founder Daniel Dines and Chief Product Officer Raghu Malpani outlined a directional decision that goes far beyond a product update: UiPath has aligned its entire engineering roadmap to make the platform primarily usable through coding agents.
What this means in practice was explained by Malpani in the specialist publication Diginomica: coding agents will capture requirements from documents, create production-ready workflows, diagnose errors in running automations and suggest fixes – all within the UiPath platform. UiPath intends to remain deliberately agnostic and support coding agents such as Claude Code and Codex.
Microsoft pursues the same idea from a different direction. With Agent Authoring in Copilot Studio, developers can describe agents in natural language, which are then automatically configured, tested and equipped with the right plugins. AI-powered governance agents additionally handle monitoring and remediation across the entire tenant – agents that oversee other agents.
| “They author and deploy, meaning they go from natural language to production-ready agentic workflows with guardrails.” — Raghu Malpani, Chief Product and Technology Officer, UiPath |
The implications are significant. Until now, humans built automations and bots executed them. In the future, agents build the automations – and other agents execute them. The role of the RPA developer shifts further towards architecture: They define guardrails, review results and manage governance. Execution is handled by the machine.
What hybrid automation actually means
Hybrid sounds like compromise. In practice, it is the opposite: an architecture that combines the strengths of both worlds and compensates for their weaknesses.
| HYBRID AUTOMATION IN PRACTICE RPA handles execution. Deterministic, rule-based tasks in structured systems: transferring data, filling forms, operating interfaces. Reliable, auditable, fast. AI agents handle the thinking. Interpreting unstructured data, preparing decisions, detecting exceptions and choosing escalation paths. Adaptive, context-aware, capable of learning. The orchestration layer coordinates both. Whether UiPath Maestro or Microsoft Copilot Studio – one layer coordinates people, bots and agents in an end-to-end workflow, with governance, audit trail and Durable Execution. |
The numbers support this approach. According to an analysis by Deloitte, companies that have migrated their core processes to hybrid AI architectures see an average 40 percent reduction in total cost of ownership (TCO) for their automation programmes within 24 months. At the same time, Forrester estimates that over the past three years, nearly 45 percent of automation budgets have gone into maintaining existing bot infrastructure – rather than building new efficiency gains. Hybrid architectures break this cycle because agents adapt to changing interfaces and APIs instead of failing with every update.
What companies should do now – and what Lunatec delivers
Gartner forecasts that by the end of 2026, around 40 percent of all enterprise applications will have task-specific AI agents embedded – up from under five percent the previous year. Those who wait now risk not only a competitive disadvantage but also rising maintenance costs for a bot landscape that becomes increasingly fragile without agentic augmentation.
The entry into hybrid automation follows a clear sequence. First: inventory the existing RPA landscape and identify the processes that currently generate the most maintenance effort – these are the candidates for agentic augmentation. Second: select the orchestration layer, whether UiPath Maestro, Microsoft Copilot Studio or a combined architecture. Third: set up a concrete pilot that delivers measurable results within three months.
Lunatec supports this transition as a UiPath Diamond Partner and Microsoft Partner. We combine deep platform expertise across both ecosystems with the ability to design hybrid architectures that are compliance-ready from the start, including the requirements of the EU AI Act, which takes effect for high-risk systems from August 2026. Experience from dozens of automation projects shows: companies that start with the right architecture today save themselves the costly retrofit tomorrow.
RPA is not dying. It is becoming the foundation on which the next generation of automation stands. The question is no longer whether you need RPA or AI. The question is who brings both together first – securely, scalably and compliantly.
ABOUT LUNATEC
Lunatec, headquartered in Frankfurt am Main, is a UiPath Diamond Partner and Microsoft Partner. We support companies in the transition from traditional RPA to hybrid, agentic automation – from process analysis through platform architecture to ongoing operations.