Camunda is pitching ProcessOS as the missing intelligence layer for companies that want to move beyond AI demos and actually rewire how their businesses run day to day. Instead of yet another chatbot or copilot, this is positioned as an “agentic operating system” that discovers how your organization really works, redesigns those processes for an AI-first world, and then keeps tuning them over time.

Camunda unveiled ProcessOS on stage at CamundaCon in Amsterdam, in front of more than a thousand enterprise leaders from 25 countries – a pretty clear signal that this is not a side project but a strategic bet for the company. At a high level, ProcessOS sits on top of Camunda’s existing agentic orchestration platform, which already coordinates millions of workflow instances for large organizations, and adds a layer of AI that can both design and optimize processes rather than just execute them.

The backdrop for this launch is what Camunda calls “the great process re-engineering.” Their argument is that most enterprises today are stuck using AI as task-level assistance – think recommendations in customer support tools, AI-powered search, or friendly chatbots that ultimately still hand off to a human or a legacy system when something serious needs to happen. That model, they say, only layers AI on top of old workflows, adding complexity and technical debt instead of unlocking new ways of working. Jakob Freund, Camunda’s CEO, frames it bluntly: every process in an enterprise is effectively legacy because none of them were originally designed for a world where powerful AI agents could own parts of the workload.

ProcessOS is meant to be the engine room for that rethink. Instead of starting from the way things are done today, the system encourages teams to define the outcomes they care about – faster onboarding, fewer claims errors, higher NPS – in plain language and then lets the platform generate the process to get there. Under the hood, ProcessOS uses agentic software development techniques: AI-generated process models, connectors and integrations, data mappings, agent prompts, decision tables, and even UI forms. Camunda already has a marketplace of extensions and connectors, and ProcessOS taps into that catalog to assemble end-to-end solutions that tie together AI models, existing systems, and human tasks.

One of the more interesting aspects is discovery. Rather than expecting enterprises to know and document all their processes perfectly, ProcessOS is built to infer them from existing knowledge and operational data. That might include things like BPMN diagrams, SOPs, logs from ERP or CRM systems, ticketing data, and other digital breadcrumbs that describe how work actually flows through an organization. The system then proposes a re-engineered version of that process that explicitly decides what should remain human-driven and where AI agents can take over or assist.

Camunda stresses that this is not about letting a black box rewrite how your company operates. ProcessOS keeps a growing knowledge base of patterns, policies, and approved integrations, and it learns what works over time as more processes are brought onto the platform. That means a bank, for example, can encode its compliance patterns and risk rules once and then reuse them as the AI proposes new processes in other business units. The idea is that the platform becomes more tailored to that specific enterprise the more it is used, rather than behaving like a generic off-the-shelf automation tool.

Governance is a big theme here, partly because any serious AI platform pitched at enterprises in 2026 has to address it head-on. ProcessOS enforces a “verification by design” approach, where every process is modeled visually and clearly indicates which steps involve AI and under what conditions. That gives operations teams, risk managers, and auditors a way to see exactly where AI is allowed to act and where humans remain in the loop. Camunda is also explicit that any process change proposed by ProcessOS has to be reviewed and approved by humans before it hits production, turning the platform into a kind of AI-powered co-designer rather than an autonomous rule-maker.

This is aligned with a broader shift analysts are tracking. Forrester has been talking about “adaptive process orchestration” – platforms that blend AI agents and non-deterministic flows with traditional workflows – as the next step beyond classic automation. In a recent landscape report on adaptive process orchestration, Forrester’s Bernhard Schaffrik describes how markets are moving toward AI-first approaches that still emphasize governance, observability, and hybrid human-in-the-loop execution. ProcessOS fits neatly into that framing, effectively giving Camunda a named product story in a category the analysts say is just now coalescing.

Camunda also makes it clear that ProcessOS is designed to work with, not replace, the systems enterprises already rely on. The platform plugs into existing stacks like ERP, CRM, core banking, or claims systems and acts as an open orchestration layer over them. That’s consistent with how Camunda historically positioned itself: a neutral process engine and orchestration layer that can sit between people, APIs, and tools, coordinating how work moves without forcing a rip-and-replace of core systems. For enterprises, that matters because most AI projects fail not for lack of models but because of integration and change management headaches.

On the infrastructure side, ProcessOS is very much aligned with AWS. The platform runs natively on AWS and integrates deeply with Amazon Bedrock and Bedrock AgentCore for foundation models, agent memory, identity, and gateway services. That means customers can tap into a range of large language models through Bedrock while using Camunda as the orchestration and governance layer that decides when and how those models act in real business processes. Camunda is already listed on AWS Marketplace with reference architectures for Amazon EKS, ECS, and EC2, and AWS recently named the company a Rising Star Technology Partner of the Year, which underlines that this partnership is more than just a marketing logo swap.

From a customer perspective, Camunda is highlighting early interest from major enterprises. Barclays, for example, is quoted saying ProcessOS tackles the real reason AI adoption stalls: companies try to build tomorrow’s processes using only what they know today. The bank’s leadership frames ProcessOS as a way to turn a “bold vision of the future” into a practical roadmap, which neatly captures the positioning: it’s not selling magic; it’s selling a structured way to move from pilot AI experiments to transformed operations. For organizations wary of AI hype but under pressure to “do something with AI,” that kind of pragmatic testimony carries weight.

ProcessOS is launching initially in closed beta, which is fairly typical for a platform that can touch critical business workflows. Selected enterprises can get early access, while others can register their interest to join future cohorts. That controlled rollout should give Camunda time to refine how ProcessOS handles different industries, from financial services and insurance to manufacturing and telecom, where process complexity and regulatory expectations differ significantly.

Zooming out, ProcessOS is a statement about where Camunda thinks process automation is heading. Over the last couple of years, the company has invested heavily in what it calls agentic orchestration – layering AI agents into BPMN-based workflows so they can decide which actions to take in real time based on context. Instead of scripting every path up front, organizations can define goals, guardrails, and toolboxes, and let AI agents choose the next step within those constraints, whether that means calling an API, escalating to a human, or kicking off another sub-process. ProcessOS adds a meta layer on top of that: not just running those agentic workflows, but helping design and improve them continuously.

For enterprises, the promise is that they get a way to evolve from fixed, deterministic workflows to adaptive, AI-augmented operations without losing control. In practice, that might look like customer support flows where AI agents handle most routine tickets but escalate edge cases intelligently, claims processes that adjust to new patterns of fraud in near real time, or onboarding journeys that adapt based on user behavior while still complying with strict regulations. If Camunda can deliver on the discovery and continuous improvement claims, ProcessOS could become the brain that keeps all those flows aligned with business outcomes as markets, regulations, and customer expectations shift.

For now, the launch of ProcessOS signals that AI in the enterprise is entering a more serious phase. The experimentation era of scattered pilots and one-off chatbots is giving way to a push for platforms that can embed AI into the core of how organizations operate, with strong governance and clear outcomes. Camunda is betting that an “agentic operating system” focused on process re-engineering, not just automation, is what enterprises will need to actually make that leap.