The German process mining leader Celonis recently utilized its Process Intelligence Day in London to announce a strategic expansion of its technological ecosystem, headlined by the launch of a new sovereign Artificial Intelligence (AI) platform. Developed in collaboration with deepset, the creators of the Haystack AI framework, this platform is designed to meet the rigorous security and data residency requirements of sensitive sectors, including defense, national infrastructure, and healthcare. The announcement comes at a pivotal moment for the technology industry, as organizations grapple with the complexities of AI orchestration, the rising costs of "tokenomics," and a growing geopolitical emphasis on digital sovereignty. Rupal Karia, Senior Vice President for Northern Europe, Middle East, and Africa at Celonis, emphasized that the move is a direct response to the "orchestration issues" facing Chief Information Officers (CIOs) who are currently navigating a fragmented landscape of AI agents and enterprise data.
Strategic Alliance and the Rise of Sovereign AI
The partnership between Celonis and deepset represents a 12-month development cycle aimed at creating a transparent, auditable, and European-centric alternative for operational AI. By integrating the Celonis Context Model (CCM) for agentic AI with the Haystack platform, the two firms intend to provide a system that remains entirely within a customer’s preferred infrastructure. This "sovereign" approach is specifically tailored for organizations that cannot risk data leaving their domestic borders or being processed by third-party providers subject to foreign jurisdictional overreach.
The timing of the announcement coincided with significant turbulence in the AI sector, following the withdrawal of the Claude Fable 5 model by Anthropic. This event has heightened concerns among digital leaders regarding the reliability of platform providers and the long-term stability of international data agreements. Milos Rusic, CEO and co-founder of deepset, stated that the joint venture is intended to provide a grounded, real-world operational context for AI that is often missing from large language models (LLMs) that rely on generalized data. The platform allows for the unification of scattered data, documents, and workflows, enabling risk detection and data-driven decision-making without vendor lock-in.
Process Intelligence as a Catalyst for AI Transformation
Throughout the London event, Celonis framed Process Intelligence (PI) not merely as a diagnostic tool but as a foundational requirement for successful AI adoption. Karia described process mining as an "outcome" that provides a "digital twin" or an "MRI scan" of a business. This visibility is increasingly seen as the missing link in enterprise AI strategies. While many organizations have experimented with generative AI for personal productivity—such as Microsoft Copilot—the consensus among attendees was that these tools offer only marginal gains, often cited as a 5% improvement in individual efficiency.
The real value of AI, according to Celonis, lies in its ability to transform business operating models. For example, if a company seeks to improve shipment times, the AI must understand the granular components of the supply chain, such as inventory imbalances across different factories. Without the process mining layer to provide this context, AI models remain "blind" to the underlying operational inefficiencies. To bridge this gap, Celonis has solidified partnerships with major technology providers, including Databricks, Microsoft, Oracle, and Workday, positioning itself as an agnostic orchestration layer that can extract and analyze data from any system of record.
The Public Sector Debate: Sovereignty and Supplier Diversity
A significant portion of the discourse in London centered on the UK public sector and the ongoing debate surrounding vendor concentration. Recent reports from the UK Parliament’s Science, Innovation and Technology Committee have suggested that the government should reconsider its reliance on foreign providers like Palantir for critical services such as the NHS Federated Data Platform. The committee’s inquiry argued that the 2027 break clause in the NHS contract should be used to explore domestic or European alternatives to avoid long-term vendor lock-in and protect public services from geopolitical shifts.
Karia noted that while Palantir continues to win major contracts, including the National Firearms Licensing Management System and Ministry of Defence (MoD) agreements, there is a growing appetite for providers that prioritize European data standards. However, he also highlighted the "opportunity cost" of public sector procurement. In the UK, a single government tender can take 18 months and require a dedicated team of 20 people—a resource burden that often leads private sector-focused firms to prioritize commercial clients. Despite these hurdles, the demand for "sovereign" solutions is expected to grow as UK-EU relations evolve and as organizations prepare for potential shifts in US trade and data policies, often referred to colloquially as the "Trump 2.0 effect."
Case Studies in Efficiency: AstraZeneca and Arm
The practical application of Celonis’s technology was demonstrated through detailed case studies from AstraZeneca’s Alexion division and the semiconductor giant Arm. Following AstraZeneca’s acquisition of Alexion, the rare disease specialist adopted Celonis to harmonize its global supply chain. John Aherne, Senior Director of Global Supply Chain Services at Alexion, explained how the firm created a digital twin to gain an end-to-end view of the journey from manufacturing to the patient.
This data-driven approach has allowed Alexion to target a reduction in working capital of between £200 million and £300 million. By utilizing process mining to identify bottlenecks in real-time, the company is now moving toward scenario planning to mitigate global economic shocks. Saurabh Saxena, Head of Ops Data Intelligence & AI Hub at AstraZeneca, noted that the integration of process mining with agentic AI allows subject matter experts to engage directly with data, providing the necessary context for automated decision-making.
At Arm, the focus has shifted from general-purpose AI to specialized use cases. Will Pratt, Senior Director of Enterprise Data at Arm, revealed that the company has moved beyond personal productivity tools to launch dedicated AI applications for contract management. By indexing contracts dating back to 1991, the firm has gained unprecedented insight into its legal and sales history. Pratt emphasized that for AI to be effective in complex environments like sales, it requires "sub-agents" with deep contextual knowledge of product catalogs and regional variations. This complexity necessitates a robust orchestration layer to monitor agent performance and manage the escalating costs of "tokenomics"—the per-token pricing model used by many AI providers.
The Challenge of AI Orchestration and Governance
As organizations deploy increasing numbers of AI agents within SaaS applications, the task of orchestration has become a top priority for CIOs. While many low-code and SaaS vendors claim to offer orchestration layers, Celonis argues that these solutions are ineffective if they lack visibility into the underlying business processes. The "agnostic nature" of Celonis’s platform is intended to serve as a neutral ground where data from various systems—be it an ERP from SAP or a CRM from Salesforce—can be analyzed holistically.
The concept of "agentic AI" involves autonomous systems that can perform tasks, but without governance, these agents risk creating new silos of "shadow AI." Karia warned that orchestration is inherently difficult in large, complex organizations. The Celonis approach is to provide the "process intelligence" that feeds these agents, ensuring they operate based on real-world data rather than hallucinations or incomplete information.
Future Implications and Market Outlook
The move toward sovereign AI and process-led transformation signals a maturation of the enterprise AI market. The initial "hype" phase, characterized by the broad rollout of chatbots, is giving way to a more pragmatic era focused on operational ROI and data security. For Celonis, the challenge lies in maintaining its independence while deepening its integrations with the very tech giants it seeks to orchestrate.
The rise of tokenomics also introduces a new financial metric for businesses. As AI usage scales, the cost of processing data through LLMs becomes a significant line item, requiring firms to be more selective about which processes they automate. Leaders like Arm’s Will Pratt suggest that measuring token efficiency will soon be as fundamental as measuring cloud spend or software licensing.
In summary, the Celonis Process Intelligence Day highlighted a strategic pivot toward a more secure, contextualized, and sovereign AI future. By addressing the dual needs of operational visibility and data residency, Celonis is positioning itself as a critical intermediary in the global race for AI-driven efficiency. As the UK and European markets continue to prioritize digital sovereignty, the partnership with deepset and the focus on "MRI-like" business transparency may provide the blueprint for the next generation of enterprise technology deployment. The road ahead involves navigating complex procurement cycles and the high costs of emerging technology, but for firms like AstraZeneca and Arm, the potential for hundreds of millions in capital savings suggests that the investment in process intelligence is no longer optional.
