Celonis has published survey findings suggesting a widening gap between companies plans for agentic AI and the day-to-day reality of their operations, with many leaders saying existing processes are slowing progress.

In its annual 2026 Optimisation Report, Celonis found that 85% of organisations want to become an agentic enterprise within three years. Yet 76% said their current processes are holding them back.

The research also points to concern about the economics of AI programmes that lack operational grounding. Some 82% of decision makers said AI will fail to deliver return on investment if it does not understand how the business runs.

The report is based on a survey of 1,649 business leaders across supply chain, finance, operations and IT. Respondents were split evenly across regions, with 20% each from APAC, DACH, Europe, India and the US.

Most participants were in a similar size bracket, with 80% reporting revenue between $2 billion and $10 billion.

Respondents represented nearly 20 industries, including manufacturing (18%), banking (15%), automotive (13%) and tech and software (11%).

Multi-agent interest

Interest in more complex AI set-ups appears widespread. The report found that 90% of organisations are already using or exploring multi-agent systems to automate complex decision-making.

The data also suggests companies see AI as central to competitive strategy. A total of 89% of leaders described AI as their single biggest opportunity to compete in the market.

Despite that urgency, the research highlights practical constraints inside organisations. The top barriers were limited internal expertise (47%) and difficulty getting AI to understand business context (45%).

Process friction

Cross-functional working remains a persistent challenge. Among process and operations leaders, 58% said their departments still do not operate seamlessly together.

That disconnect can limit end-to-end visibility across workflows and systems, which is often required for automation that spans multiple teams. It can also introduce delays and inconsistencies, making it harder to move from narrow proofs of concept to broader AI deployment.

Celonis positions its core technology, Process Intelligence, as a way to provide operational context for AI systems. It argues that AI agents need optimised processes, process data and operational context to act autonomously and effectively.

Process Intelligence refers to technologies that analyse how work is executed across systems and teams, using process data to map flows and identify bottlenecks. Celonis argues this context is needed if AI is expected to do more than complete isolated tasks.

The survey results suggest decision makers are already alert to the issue. If most leaders believe AI needs to understand how the business runs to deliver ROI, investments in data, governance and process standardisation may become more closely tied to AI strategies over the next few years.

Companies aiming to deploy agentic AI at scale may face two parallel requirements: the technical ability to deploy AI across workflows, and the operational discipline of clear ownership, common definitions and integrated processes across departments.

Executive view

Many organisations are still struggling to turn AI ambition into measurable business results, Celonis said.

“While business leaders are leaning boldly into an agentic AI future, the reality is that many are struggling to translate that ambition into tangible ROI right now,” said Carsten Thoma, President and Board Director, Celonis.

Thoma linked stronger operational context with more consistent outcomes from AI deployments.

“For AI to truly work for the enterprise, it needs more than just data. It needs operational context. By using Process Intelligence to give AI a shared understanding of how a business actually runs and how to improve it, were finally turning that ambition into continuous, measurable value,” Thoma said.

The findings add to growing research that frames AI transformation as an organisational change challenge as much as a technology project. Over the next three years, many firms are likely to increase spending on governance, process visibility and internal skills as they move from multi-agent experimentation to broader deployment across supply chain, finance, operations and IT.