When Marco Lübbers, Managing Director for KYC Operations in Corporate and Investment Banking at Deutsche Bank, first overlaid the Celonis process map onto his KYC workflows, what had remained hidden for years became visible: systematic bottlenecks, redundant client outreach and process loops that no audit and no workshop had ever uncovered. "Celonis helps us spot opportunities to improve the KYC process and reduce our onboarding times, which is super important to the client experience," Lübbers summarises the outcome. This insight – both a banal foundation and a painful catalyst – stands at the beginning of every operational transformation in the financial sector.
In an industry where some institutions spend up to 80 per cent of their IT budget maintaining legacy systems (a widely cited industry estimate whose precise origin is debated), process transparency becomes a strategic necessity. Celonis, founded in 2011 by three students at the Technical University of Munich (TU München), has become the world's leading provider of process intelligence – with an Annual Recurring Revenue (ARR) exceeding 770 million dollars (as of 2023), more than 5,000 deployments and a client list that ranges from Deutsche Bank and HSBC to the Brazilian stock exchange B3.
The operational dilemma of banks
The problem is well known, but its magnitude is underestimated: approximately 70 per cent of all global financial transactions – from ATM withdrawals to credit card payments and securities settlement – pass through mainframe systems at some point. Over 40 per cent of all banking systems worldwide are based on COBOL, a programming language dating from 1959. At the same time, the COBOL programmers who keep these systems alive are retiring – and the next generation has little interest in outdated technology.
For operational excellence, this creates a fundamental paradox: banks invest heavily in technology, yet costs do not fall. McKinsey speaks of the "productivity paradox" – technology spending among asset managers is growing at a compound annual growth rate (CAGR) of 8.9 per cent, yet costs as a share of assets under management remain stubbornly flat. The reason: individual processes are being automated, but the overall operational picture remains opaque.
These figures are further exacerbated by the current agentic AI hype. BCG data shows: 95 per cent of organisations achieve no measurable return from Generative AI (GenAI); only 26 per cent progress beyond the proof of concept. Gartner confirms: fewer than 30 per cent of AI leaders report that their CEO is satisfied with AI ROI. The reason is almost always the same: AI initiatives are layered onto broken processes that nobody fully understands.
Legacy systems: Rigid mainframe architectures prevent rapid adaptation. 53 per cent of institutions cannot scale their operations effectively.
Regulatory pressure: DORA, T+1 settlement, MiCA, FiDA and Anti-Money-Laundering (AML) requirements absorb enormous resources while demanding both stability and agility.
AI scaling: Fragmented processes, unstructured data and a lack of real-time integration prevent banks from scaling AI beyond pilot projects. 57 per cent of organisations rate their data as not AI-ready (Gartner, 2025).
Talent shortage: Knowledge loss from retiring legacy experts meets a lack of willingness among young talent to work with outdated systems.
Cultural resistance: Hierarchical structures and risk-averse cultures in large financial institutions slow the necessary cultural change. HFS Research notes: more than two-thirds of process intelligence implementations deliver disappointing results – frequently due to a lack of organisational anchoring.
Competitive pressure: Neobanks and fintechs operate with cost-to-serve models up to ten times more efficient. McKinsey projects that 170 billion dollars in global banking profits are at stake for institutions that fail to adapt.
It is precisely within this constellation that Celonis positions itself. Not as yet another automation tool, but as the intelligence layer that first enables banks to understand their problems – before attempting to solve them.
How Celonis works: from event log to digital twin
Celonis answers a single but fundamental question: How do our business processes actually run – as opposed to what we designed or documented? The Celonis Process Intelligence Platform delivers the answer in four steps.
1. Data extraction: Connectors to Enterprise Resource Planning (ERP) systems, core banking, Customer Relationship Management (CRM) and other systems extract event logs – timestamped records of every process step.
2. Process Intelligence Graph: Millions of data points are reconstructed into end-to-end process flows and enriched with business context: a living digital twin.
3. Analysis & impact: Process deviations, bottlenecks and compliance violations are visualised and their financial impact quantified.
4. Orchestration & AI: The Orchestration Engine – available as a core platform feature since November 2025 – coordinates tasks across all tools, systems and departments, including AI agents, Robotic Process Automation (RPA) bots and manual activities.
At its core, Celonis creates a Process Intelligence Graph – a living digital twin of business operations that brings together process data and business context. Unlike static process models, this graph shows not the target state but the actual operational reality, across systems and in real time. Constellation Research put it succinctly at Celosphere 2025: "Celosphere 2025 isn't about process mining anymore, it's about decision mining: finding, understanding and executing where AI should act."
The platform is deliberately system-agnostic: it connects to SAP, Oracle, Salesforce, Avaloq, Temenos or proprietary core banking systems equally well. In practice, however, "system-agnostic" does not mean "effortless": banks with heterogeneous, non-SAP-based IT landscapes – often grown through mergers – require upstream consolidation of event log formats. The neutrality is an advantage, but not self-executing.
Object-Centric Process Mining (OCPM): Since 2023, Celonis analyses processes not in isolated single threads but maps the interrelation of multiple business objects (orders, invoices, customers) in a single model. New analysis tools such as Performance Spectrum and Instance Explorer make the complexity navigable.
Task Mining & AI-driven Task Discovery: Desktop actions such as keystrokes, mouse clicks and screen scrolling are captured and linked to business processes – for complete transparency even beyond system boundaries. Deutsche Bank plans to deploy task mining to capture process steps outside its core systems.
Celonis Data Core: Bi-directional zero-copy integration with data lakes such as Databricks and Microsoft Fabric – no data duplication required. At Celosphere 2025, the Databricks integration was announced as a complement to the existing Microsoft Fabric connection.
AgentC Suite: Process Intelligence API for AI platforms such as Microsoft Copilot Studio, Amazon Bedrock and Salesforce Agentforce. Plus the world's first Model Context Protocol (MCP) server for process intelligence – a universal connector through which any AI agent can access the Process Intelligence Graph.
Process Copilot: Generally available since May 2025. A GenAI chatbot with natural-language queries on process data, integrated into Slack and Microsoft Teams. Suggested questions, KPI monitoring and direct actions (CSV export, email generation) from within the chat.
Orchestration Engine: Coordinates the increasingly complex network of RPA bots, AI agents, workflow automations and manual activities in a single control layer – generally available as a core feature since Celosphere 2025.
This thesis now has academic backing: Prof. Wil van der Aalst (RWTH Aachen, Chief Scientist at Celonis) published the paper "No AI Without PI!" (arXiv:2508.00116) in July 2025, in which he describes Object-Centric Process Mining as the "missing link" between data and processes. And Celonis' own survey of 1,620 executives (companies with over 500 million dollars in revenue) shows: 89 per cent say AI needs the context of how their business operates to deliver effective results. 58 per cent worry that process deficiencies limit AI value.
An important caveat: this survey was commissioned by Celonis, and the thesis is inherently self-referential – if "No AI without PI" holds true, Celonis as the market leader in PI is the natural beneficiary. Regardless of the source, however, the findings converge with observations from BCG, Gartner and McKinsey: AI initiatives launched without process understanding fail disproportionately often.
Celonis in banking: where the platform delivers concrete results
The Celonis Banking Solutions are used by prominent institutions – including Wells Fargo, HSBC, ABN AMRO, Standard Bank, Deutsche Bank and Atruvia, the IT service provider for Germany's cooperative banking sector (Volks- und Raiffeisenbanken). The application areas in the financial sector span the entire operational value chain – with increasingly concrete, measurable results.
KYC and client onboarding
Know Your Customer (KYC) consumes up to 33 per cent of the compliance budget at large financial institutions (Fenergo study). Deutsche Bank deploys Celonis in Corporate and Investment Banking to optimise KYC workflows – achieving a significant reduction in onboarding times and fewer redundant client outreach instances. The next step: machine-learning components for predictive SLA adherence.
In July 2025, Celonis and Bright Cape launched the Smart KYC Control Tower – a specialised solution for KYC compliance optimisation that combines end-to-end process monitoring, real-time insights and SLA breach prevention. ABN AMRO has been using Celonis for over four years with five realised use cases across the entire E2E process chain.
Cross-border payments and CIB operations
The most impressive banking case study comes from South Africa: Standard Bank, the continent's largest bank, has transformed its CIB operations (Corporate and Investment Banking) with Celonis. The result: a straight-through-processing rate (STP) exceeding 90 per cent for cross-border payments and a reduction in processing time from up to 55 hours to just a few hours. Richard de Roos, Head of CIB Operations: "Celonis is part of how we run the business. It allows us to leapfrog into the future and to compete with best-practice fintechs globally." The bank is now expanding deployment to trade products and investment banking.
STP rate: >90% for cross-border payments (previously significantly lower)
Processing time: From up to 55 hours to just a few hours
Volume: 9–11 trillion ZAR in monthly cross-border payment volume
Method: "Intelligent Processing Model" (IPM) with Celonis as single vendor
Expansion: Trade products and investment banking as next domains
Credit processes and regulatory compliance
In the lending business, Celonis has reduced waiting times for credit approvals by 50 per cent and saved 100 days in credit recovery, according to the joint BCG white paper. PostFinance uses the platform for account opening and credit decisions. Degussa Bank – with 6.1 billion euros in total assets, one of the larger German private banks – has been using Celonis since 2020 for compliance with MiFID II and WpHG recording obligations (Securities Trading Act) and achieved a 12 per cent improvement in the recording consolidation rate.
Trade settlement and capital markets
In the context of the T+1 settlement regime – in force in the US since May 2024, in the EU from October 2027 – real-time process transparency in the post-trade space moves from a convenience feature to an operational necessity. Celonis addresses SEC requirements and the Central Securities Depository Regulation (CSDR) directly: process mining in the trade lifecycle enables automatic mapping of even high-volume products such as FX, quantifies the impact on liquidity and capital at risk, and provides root-cause analyses for failed or misbooked trades. Accenture offers a dedicated post-trade settlement app with an early-warning system for delays in the Celonis Marketplace.
The Brazilian stock exchange B3 deploys Celonis in procure-to-pay and procurement – with a 75 per cent reduction in procurement cycle times and an AI agent for automated contract renewals. This demonstrates that capital market infrastructure itself benefits from process intelligence.
An honest assessment: in the narrower sense – derivatives operations, collateral management, specific reconciliation processes in securities – publicly documented use cases remain limited. Celonis' strength in the capital markets space lies primarily in post-trade settlement, cross-border payments and KYC onboarding for investment banking clients. Expansion across the full trade lifecycle chain is the logical next step but is not yet broadly evidenced.
Fraud and anti-money laundering
In partnership with Doculabs, Celonis is transforming the processes behind financial crime investigations. Process mining uncovers systematic inefficiencies in AML investigation processes – such as redundant manual reviews, lack of risk-based prioritisation and inconsistent escalation paths. EY studies show: agentic-AI-powered AML investigations can reduce processing time per investigation by 50 per cent.
Cross-border payments: >90% STP rate, 30% reduction in cycle time (Standard Bank)
KYC onboarding: Significant reduction in onboarding times, fewer redundant client outreach instances (Deutsche Bank)
Credit approval: 50% reduction in waiting time; 100 days saved in credit recovery (BCG/Celonis)
Recording compliance: 12% improvement in recording consolidation rate for MiFID II (Degussa Bank)
Document matching: 25,000 hours saved in manual document review (BCG/Celonis)
Procurement (B3): 75% reduction in cycle times, 5 weeks to payback
Conversion rate: 15% improvement; at one digital bank approximately 50% (BCG/Celonis)
Customer satisfaction: 10–15 point improvement in NPS (BCG/Celonis)
Market overview: Celonis and its competitors
The process mining software market is growing strongly, but estimates vary considerably depending on definition: from 18.6 per cent CAGR (Compound Annual Growth Rate, Mordor Intelligence, narrowest definition) to 59.4 per cent (Grand View Research). The consensus lies at an annual growth rate of 44 to 45 per cent. The Gartner Magic Quadrant 2025 evaluates 16 providers for the first time – a sign of market maturity. Celonis leads for the third consecutive time as Leader with the highest placement for "Completeness of Vision".
| Provider | Positioning | Banking relevance |
|---|---|---|
| Celonis (Leader) | Process Intelligence Platform. ~50% market share. Gartner Leader 2025 (third consecutive time). Forrester Leader. System-agnostic, strongest AI capabilities (PI Graph, AgentC, MCP server). | Broadest banking references (Deutsche Bank, Standard Bank, HSBC, Wells Fargo, ABN AMRO, Atruvia). Smart KYC Control Tower. BCG and McKinsey partnerships. 383% ROI over three years (Forrester TEI, Celonis-sponsored). |
| SAP Signavio (Leader) | Process mining + modelling + simulation. Gartner Leader. Deepest SAP integration on the market. Joule AI copilot with 40 skills since January 2026. AI Agent Excellence for agentic AI governance. | Ideal for SAP-centric institutions. Dedicated DORA compliance offering and Financial Services & Insurance industry solution. Weaker with non-SAP systems. |
| IBM (Leader) | Process mining with watsonx integration and OCPM compliance. Hybrid cloud via Red Hat OpenShift. Focus on regulated industries. | Strong position with banks running IBM infrastructure and hybrid cloud requirements. OCPM capability for complex banking processes. |
| UiPath (Leader) | Process mining + task mining + communications mining in one platform. Closed loop from discovery to bot automation. Multi-process mining since May 2025. | Attractive for institutions that view mining primarily as a precursor to bot automation. Only provider with seamless discovery-to-execution. Weaker in real-time analysis. |
| Pegasystems (Leader) | Workflow intelligence + process mining. Integration with the Pega platform for case management and low-code automation. | Dedicated financial services focus. Strong with institutions already using Pega for case management. |
| Software AG (ARIS) (Leader) | Established process modelling heritage. Prescriptive analytics and Decision Model and Notation (DMN) rule checks. Strong in governance and compliance documentation. | Widespread in regulated environments. Lags behind on AI features. Requires more technical expertise. |
| Microsoft (Challenger) | Power Automate process mining with Copilot integration. Low-code. High visibility through the M365 ecosystem. | Entry scenario for smaller institutions. Growing relevance, but not a standalone enterprise process intelligence platform. Strategic partnership with Celonis. |
| ServiceNow (Challenger) | Native Now Platform integration. Acquisition of UltimateSuite for task mining. GenAI features. | Relevant for institutions using ServiceNow as their IT Service Management (ITSM) platform. Growing rapidly, but still without deep banking specialisation. |
The Celonis–McKinsey–BCG dimension
Noteworthy is the strategic interconnection of Celonis with the leading management consultancies. BCG maintains a dedicated partnership with Celonis for AI-driven process intelligence in banking and has jointly published a white paper on "Banking (Ops) Excellence" that describes a three-stage approach: Assessment of Potential Value, North Star & Execution Blueprint, MVP Launch. The quantified results – 50 per cent efficiency improvement, 30 per cent reduction in cycle time for cross-border payments, 25,000 hours saved in document matching – stem from this context. Important caveat: BCG and McKinsey have a financial interest in selling Celonis projects. The results in the white paper have not been independently verified.
McKinsey is a Platinum Partner of Celonis and combines strategic advisory expertise with process mining technology for transformation programmes. Accenture has maintained a Strategic Global Alliance since 2022 with over 2,000 AI engagements. For financial institutions, this means: Celonis is not merely a software vendor but increasingly part of the strategic advisory ecosystem – and the consultancies use Celonis as an "evidence engine" for their transformation mandates.
The thesis of both consultancies converges: operational excellence in banking requires first process transparency, then process redesign, and only then technology deployment. Celonis provides the infrastructure for step one and increasingly for step three – while the consultancies take responsibility for step two.
A strategically relevant personnel signal: in July 2025, Dilipkumar Khandelwal – formerly a top executive at SAP and Deutsche Bank – was appointed Chief Customer Officer (CCO) of Celonis and Chairman of the India Advisory Board. This underscores the banking focus and the planned expansion into India.
Agentic AI and process intelligence: the convergence
At Celosphere 2025 in Munich (4–5 November 2025, over 3,000 attendees), Celonis articulated its vision of "agentic process intelligence". The core message: AI agents need process context – "just like a GPS needs a map" (Alex Rinke). The platform architecture consists of three layers: Data Core (high-performance data infrastructure), Process Intelligence Graph (living digital twin) and Build Experience (analysis, design, operation of AI-driven processes).
The Model Context Protocol (MCP) server is the strategically most significant product in this context. MCP – launched by Anthropic as an open-source standard in November 2024 – standardises how AI models access enterprise data. Celonis' MCP server for process intelligence enables any AI agent, regardless of the Large Language Model (LLM) provider, to access the Process Intelligence Graph. It is the "USB standard for AI integrations" in the enterprise context.
However, Thoughtworks' Technology Radar (Vol. 33) warns against "naïve API-to-MCP conversion": security is the primary concern of the MCP ecosystem – tool poisoning, cross-server shadowing and missing authentication are real risks that must be addressed in enterprise adoption.
The market figures for agentic AI in banking are impressive but should be treated with caution: Accenture calls 2026 the year of "scaled transformation through agentic AI". McKinsey projects that AI could reduce certain banking costs by up to 70 per cent. IDC expects 1.3 billion AI agents in business workflows by 2028. But KPMG soberly notes: only 11 per cent of companies have agentic AI productively deployed. 48 per cent cite governance concerns as the primary obstacle.
What does not work: lessons learned and critical assessment
The documented practical experiences reveal clear patterns for successful and failed implementations. A balanced assessment must also name the limitations.
Scope deliberately limited: Process mining is only suited to high-volume, highly automated processes. Not every banking process benefits – expectation management vis-à-vis the board is critical.
Data quality as a precondition: The greatest initial challenge is identifying, collecting and cleansing event log data. Business unit owners must be actively involved. The average Centre of Excellence (CoE) comprises 15 full-time employees (Deloitte).
Define a clear use case: Whether process transparency, optimisation, monitoring or compliance – the purpose must be clear before the start and supported by stakeholders.
Actually implement the optimisation: The value from process mining is only realised through the implementation of identified measures. Mining alone does not generate ROI. Prof. van der Aalst warns in his arXiv paper against "post-mortem analysis" without real-time actionability.
Scale in stages: Successful institutions start with a single process or sub-process and then systematically scale to further domains. Implementations not infrequently take 12 to 24 months.
1. Data quality is a threshold, not a feature: Process mining depends on event logs. If data is fragmented, locked or erroneous, false conclusions arise – particularly at banks without SAP as their core system that must consolidate heterogeneous event log formats.
2. Analysis paralysis is real: The massive volume of insights can be paralysing if no clear governance exists regarding who may act on which findings. HFS Research notes: more than two-thirds of process intelligence solutions deliver disappointing results.
3. ROI figures originate from Celonis: The Forrester TEI study (383% ROI) was sponsored by Celonis. The 89 per cent survey was commissioned by Celonis. Independent ROI studies for the banking sector not funded by Celonis are scarce.
4. Agentic AI also works without process mining: For certain use cases – code generation, KYC document extraction, customer service chatbots – agentic AI is successful without process intelligence being a prerequisite. The causal direction of the "No AI without PI" thesis is simplistic.
5. GDPR and works councils are showstoppers in Germany: Celonis extracts event logs from core systems; task mining captures employees' desktop actions. In German organisations with works councils (Betriebsrat) and strict GDPR interpretation, this is a highly sensitive topic that must be coordinated early with employee representatives and data protection officers – otherwise the implementation fails on legal, not technical, grounds.
6. Costs and lock-in are not addressed: Celonis does not publish list prices. The average Centre of Excellence (CoE) comprises 15 full-time equivalents (Deloitte), and implementations take 12 to 24 months. If the Process Intelligence Graph becomes the central orchestration layer, significant vendor lock-in emerges. Data portability and exit strategies are topics that should be clarified before procurement.
Celonis platform evolution · 2011–2026
Recommendations for German financial institutions
Based on the analysed market data, consulting approaches, case studies and critical assessments, seven concrete recommendations emerge for institutions evaluating process intelligence as a lever for operational excellence.
Before investing in RPA, AI or new core banking systems, clarity about actual process flows must be established. A process mining pilot project on a clearly defined core process – KYC onboarding, loan application or cross-border payments – creates the evidence base for all subsequent investment decisions. Deutsche Bank's experience shows: between assumed and actual workflows lie systematic deviations that remain invisible without data-driven analysis.
Process mining delivers results only as good as the underlying event logs. A dedicated data quality assessment of the relevant source systems should precede every implementation – including clear responsibilities assigned to business units. Institutions with heterogeneous, non-SAP-based IT landscapes require additional consolidation work. Building a Centre of Excellence (CoE) with an average of 15 full-time employees should be planned from the outset.
The DORA requirements (Digital Operational Resilience Act) for process resilience and documentation and the T+1 transition (EU: October 2027) offer natural entry points for process intelligence. Rather than treating compliance as a pure cost factor, the digital process twin should simultaneously serve as a basis for operational optimisation. SAP Signavio already offers a dedicated DORA solution; Celonis addresses T+1 and CSDR directly.
SAP-centric institutions should evaluate SAP Signavio. Institutions with a strong automation focus should consider UiPath. Banks with IBM infrastructure and hybrid cloud requirements should assess the IBM approach. For system-agnostic, end-to-end process intelligence with strategic AI ambitions, Celonis remains the benchmark – particularly through its banking case studies and BCG/McKinsey partnerships. Gartner and Forrester confirm the leadership position, but the competitive landscape with 16 evaluated providers now offers genuine alternatives.
Process mining alone does not generate ROI – that is the most uncomfortable truth in this market. Successful institutions like Standard Bank define in advance what resources are available for implementing identified measures and link process intelligence with concrete automation and change management initiatives. The Celonis Orchestration Engine and dedicated AI agents (such as B3's contract renewal agent) close this gap technically – but organisational anchoring remains a management responsibility.
The convergence of process intelligence and agentic AI is real – Celonis' MCP server, AgentC and Process Copilot are production-ready. But market reality is sobering: only 11 per cent have agentic AI productively deployed (KPMG). Institutions should evaluate today how a process intelligence layer can serve as a context layer for AI agents in KYC, AML, credit risk and procurement – but with realistic expectations regarding implementation timelines and governance requirements.
Process transparency can feel threatening – it reveals where inefficiencies lie and who is responsible for them. Accompanying change management that positions transparency as an opportunity rather than a control mechanism is essential for organisational acceptance. McKinsey's data shows: first movers achieve a 4 per cent advantage in Return on Tangible Equity (ROTE). Those who fail to act risk structural competitive disadvantages – but those who act too quickly without bringing the organisation along also fail.
Christian Schablitzki
Strategy & Management Consultant · Agentic AI Expert for Financial Services
Over 20 years in investment banking and derivatives trading, followed by more than 10 years advising financial institutions. Currently Partner at Infosys Consulting in Germany. Certified in Google AI, Generative AI Leader (Google Cloud) and IBM RAG and Agentic AI.
LinkedIn profile →Keep reading – in your inbox every two weeks.
Capital markets insights, regulatory updates and AI trends. Concise, well-founded, free of charge.
GDPR-compliant. Unsubscribe at any time.