Why agentic AI is stalling — and why orchestration has become the trust anchor
Graph by Camunda: 10% currently use agent-based orchestration, 85% have not yet reached the required process maturity to implement agent-based orchestration, 81% believe a fully autonomous enterprise is unattainable without agent-based orchestration.
Agentic AI has moved rapidly from research labs into executive agendas. The promise is compelling: software agents that can reason, act and collaborate across systems, people and data to execute complex business processes. Yet as interest accelerates, a familiar enterprise pattern is re-emerging — ambitious AI visions colliding with operational reality.
Camunda’s newly released 2026 State of Agentic Orchestration and Automation Report puts empirical weight behind that tension. Based on a survey of 1,150 senior decision-makers in large enterprises across the US, Germany, the UK and France, the report reveals widespread experimentation with AI agents — and a striking lack of production readiness, trust and organisational maturity.
At a time when claims about autonomous AI are multiplying, the report’s conclusions are notably sober. The constraint is no longer access to AI capabilities. It is the ability to embed those capabilities safely, transparently and compliantly into real, end-to-end business operations.
Why Camunda’s perspective matters
Camunda approaches agentic AI from a position shaped by long experience with enterprise process orchestration. The company has spent more than a decade coordinating complex workflows across systems, people and data in large organisations, particularly in regulated and mission-critical environments.
Its platform is used by more than 700 enterprises, including ING, Atlassian and Vodafone, where auditability, traceability and operational control are foundational requirements rather than optional features. That background informs the report’s framing: agentic AI is evaluated not as a standalone innovation, but as something that must operate within existing governance structures and process architectures.
The research itself was commissioned with Coleman Parkes, a firm specialising in enterprise technology studies, and targeted senior IT leaders, business decision-makers and enterprise architects responsible for automation in organisations with more than 1,000 employees. Conducted between September and October 2025, the survey reflects current enterprise conditions rather than theoretical futures.
Widespread adoption — limited operational impact
On the surface, agentic AI adoption appears robust. Seventy-one percent of surveyed companies report using AI agents. In operational terms, however, progress is limited: only 11 percent of those use cases reached production readiness last year.
This gap echoes earlier phases of enterprise AI adoption, where proof-of-concepts and pilots proliferated but failed to translate into stable, business-critical deployments. What differentiates agentic AI is the level of autonomy involved. Agents are not merely advisory; they can initiate actions, interact across systems and influence outcomes.
That autonomy helps explain why 73 percent of respondents acknowledge a gap between their vision for AI agents and what they can realistically deploy today. The issue is not experimentation — it is confidence.
Trust, not capability, is the dominant barrier
Concerns about risk, transparency and compliance dominate executive thinking around agentic AI:
-
84 percent fear business risks if AI agents operate without adequate control mechanisms
-
80 percent are concerned about a lack of transparency into AI usage
-
66 percent cite compliance concerns
These concerns translate directly into constrained deployment. Eighty percent of organisations report that most of their AI agents function as chatbots or assistants — generating summaries or answering questions, but not handling business-critical cases. At the same time, 48 percent say their agents operate in silos rather than being embedded across end-to-end processes.
The pattern is consistent: enterprises are willing to tolerate AI at the margins, but hesitate to integrate it into the operational core.
Kurt Petersen, Senior Vice President Customer Success at Camunda, captures the dynamic succinctly: “The value proposition of AI agents is undeniable, but trust remains the biggest hurdle to acceptance.” Without transparency and clear guardrails, agents remain isolated tools rather than transformative capabilities.
Automation is delivering value — and amplifying complexity
The caution around agentic AI stands in contrast to the broader trajectory of automation. According to the report, 95 percent of companies experienced increased business growth through process automation last year, up from 87 percent the previous year.
On average, organisations have automated 48 percent of their processes and expect that figure to rise to 64 percent. Nearly 79 percent plan to increase automation spending, with budgets expected to grow by 20 percent over the next two years.
Yet this progress is unfolding within increasingly fragmented technology environments. Seventy-six percent of organisations report that the number and diversity of endpoints involved in each process are growing exponentially. APIs, cloud services, legacy systems and external platforms are multiplying — often faster than governance and integration models can adapt.
As a result, 85 percent of respondents say they need better tools to manage interfaces between processes. Against this backdrop, the introduction of autonomous agents without stronger orchestration risks compounding existing weaknesses. Half of respondents explicitly fear that uncontrolled AI agents could exacerbate problems caused by poorly implemented automation.
Agentic orchestration as an enterprise operating model
Camunda’s report argues that the central challenge is not the intelligence of individual agents, but the absence of an operating model capable of governing them. That model is described as agentic orchestration.
Agentic orchestration combines:
-
Deterministic orchestration, which defines workflows, responsibilities and compliance guardrails
-
Dynamic orchestration, which allows AI-driven reasoning and adaptation within those boundaries
In this framework, agents are not free-floating actors. They operate as governed participants within structured processes, interacting with people and systems under defined rules.
The survey data strongly supports this approach. Eighty-eight percent of respondents say AI must be orchestrated across business processes to realise maximum value. Ninety percent believe AI, like any other endpoint, must be orchestrated to ensure compliance.
At the same time, 85 percent acknowledge that their organisations have not yet reached the process maturity required to implement agentic orchestration effectively — a gap that helps explain why so many initiatives stall at the pilot stage.
Moving from experimentation to enterprise capability
The report positions orchestration as the bridge between AI ambition and operational reality. Petersen is explicit: “Individual agents aren’t the key to bridging the gap between AI vision and reality; agent-based orchestration is.”
This perspective aligns with long-standing patterns in enterprise technology adoption. Technologies that scale successfully do so not because they are intelligent in isolation, but because they integrate cleanly into operational systems, audit frameworks and governance models.
Deterministic orchestration provides accountability. Dynamic orchestration enables adaptability. Together, they create the conditions under which enterprises can trust AI agents to operate in mission-critical workflows — not as experiments, but as reliable components of the operating model.
Governance is no longer a secondary concern
One of the report’s most telling signals is how closely AI adoption is now tied to governance expectations. Ninety percent of respondents believe AI must be orchestrated to ensure compliance.
In regulated industries and jurisdictions such as the EU, this reflects a broader shift. AI systems are increasingly assessed not only on performance, but on explainability, controllability and auditability. In that context, orchestration functions as a governance mechanism as much as a technical one.
A measured outlook on agentic AI
Camunda’s 2026 State of Agentic Orchestration and Automation Report does not frame agentic AI as a shortcut to transformation. Instead, it presents a measured view: enterprises are automating rapidly, investing heavily and experimenting broadly — but remain cautious about autonomy without control.
The gap between vision and reality is clear, but so is the emerging consensus on what needs to change. Trust, transparency and orchestration — not ever more agents — are becoming the decisive factors.
For organisations looking to move agentic AI from pilots into production, the message is direct: intelligence alone is insufficient. Without orchestration, agents remain peripheral experiments. With it, they gain a credible path toward becoming stable, business-critical capabilities.
If you value independent explainers on AI and deep tech, you can support our work via Buy Me a Coffee. It helps keep MoveTheNeedle.news focused on depth, not clicks.