A Copilot for Every Silo Is Still a Silo — Just Faster

Chapter 4 of 7 in the joint white paper "BPM: Where Are We Headed? The Reinvention of Process Management in the Age of Agentic AI," written by BPM&O and bpExperts.

Key Message: BPM initiatives always serve several goals at once—and today are mostly decoupled from one another. AI makes the overarching meta-model economically viable for the first time, but scattered AI capabilities merely reproduce the old silos. What's needed is a shared meta-model, an agentic platform (the Business Transformation Brain), and an actively shaped cultural shift.

4.1 Starting Point: Not an End in Itself—But Fragmented

Process management initiatives never exist for their own sake. They always serve several objectives at once: process documentation in support of process excellence and the continuous improvement process (CIP), support for major IT transformations (ERP/CRM), compliance and audits (GRC), data-driven optimization through process mining, and managing the IT architecture within enterprise architecture. They also share a cultural goal: making end-to-end workflows visible ('Get out of the silos!'), establishing accountability for end-to-end business processes, and embedding process-oriented thinking and action—a fundamental shift from traditional management approaches. In the past, the sheer effort of making end-to-end processes transparent at all held this shift back; AI's capabilities are now removing that brake (see Chapter 3).

The ambition of process management has always been to bring all these perspectives together—the process as the shared point of reference for the business, IT, compliance, and analytics. Reality, however, often looks different: the initiatives are decoupled. Different people work on the same processes with different tools, following different methods—the process-excellence community maintains its own modeling repository, the ERP program documents in its implementation tool, the GRC team keeps its own control catalogs, the mining analysts work on event data, and the enterprise architects model in yet other tools.

Although the need for convergence keeps growing, efforts are often reduced to simple toolchains: interfaces that copy data from A to B without a methodological 'big picture'—that is, without an overarching meta-model that semantically connects the perspectives. A pipe between two tools is not yet an integration of knowledge. Nor does value automatically arise from linked data. Value arises when the shared view enables better prioritization.

4.2 The Technological Dimension: The BT Brain—Meta-Model Plus Agentic Platform

AI fundamentally changes this starting point, because it makes the overarching meta-model economically viable for the first time: enriching, linking, and maintaining a shared knowledge base across all initiatives—previously defeated by effort and tool limits—can now largely be automated. But this is where the next fragmentation trap lurks: if every initiative introduces its own AI assistants, copilots, and agents, the fragmentation isn't removed but reproduced at the AI level—only faster. Rolling out scattered AI capabilities is not the answer. Alongside the pure content layer, an agentic platform is needed that ensures hybrid teams of humans and agents work on the same knowledge base, follow the same rules, and produce traceable results.

Our answer to this is the Business Transformation Brain (BT Brain). It connects three architectural layers: 1) Content—a central knowledge graph acting as an overarching meta-model that links reference processes, end-to-end scenarios, IT-transformation scope items, compliance obligations, validated AI use cases, and architecture information through explicit, typed relationships; 2) Interaction—platforms and assistants through which people query, visualize, and maintain this knowledge in real time; 3) Agents—specialized AI agents that actively work on the basis of this knowledge.

The knowledge graph is more than a 'single source of truth' for documentation: it is exactly the overarching meta-model that decoupled initiatives previously lacked. An AI use case is not merely 'relevant to financial planning'—it is linked via a typed relationship to the specific end-to-end scenario it accelerates. A compliance obligation—say, GDPR Article 22 on automated individual decisions—is linked to precisely the scenarios and agents it constrains, including the mandatory controls. The process-excellence, transformation, GRC, mining, and EA perspectives thereby meet for the first time within a shared semantic frame of reference.

The principle of the structured process debate shows how the agentic platform prevents fragmentation in action: instead of a single AI assistant prone to diplomatic vagueness, specialized agents with deliberately different perspectives work on the same question—a process analyst checks against the reference repository, a value analyst assesses the economic benefit, a compliance critic tests every proposal against the regulatory obligations stored in the graph (GDPR, EU AI Act, SOX, GxP), and a solution broker weighs the options in a transparent scoring matrix. An orchestrator only lets the debate converge once critical risks have been addressed. The tensions that, in decoupled initiatives, lead to friction losses and downstream compliance corrections are thus resolved in a structured way before a decision is made—with fully traceable evidence cited in the underlying graph. The synthesis automatically produces the deliverables that used to require manual translation work: validated BPMN 2.0 process models with human-in-the-loop checkpoints modeled in from the start, traceable management presentations, and interactive process visualizations. Human experts retain the role that should never be delegated: judging whether the right question was asked, and what the recommendation means for the organization and its people.

The BT Brain integrates into the existing tool landscape rather than replacing it: through open interfaces—such as the APIs of ARIS or SAP Signavio, SAP LeanIX, and SAP Cloud ALM—established systems turn from a mere 'system of record' into a 'system of connection'; their process, project, and landscape data become readable and writable by agents. And the Brain gets better with every use: every debate, every generated model, every structured compliance obligation is written back into the graph. It isn't the language model that gets trained—it is the organization's knowledge that accumulates. If the process owner changes a policy in the model, that change immediately affects all connected agents and initiatives.

The BT Brain is a conceptual design—but by no means only theory. We have already implemented it and are using it in real client engagements: with Claude as the AI foundation, MCP servers as an open integration layer, a Neo4j knowledge graph as the content layer, and connections to clients' established BPM and EA tools—ARIS, SAP Signavio, SAP LeanIX, and SAP Cloud ALM. That the concept is formulated platform-neutrally is a deliberate choice: no company starts on a greenfield. We therefore build the BT Brain individually for each client as a tailor-made, integrated platform—the knowledge graph and the agentic orchestration form the stable core, while the connected tools follow the client's respective as-is landscape. How viable this approach is becomes especially clear where a vendor itself brings together the building blocks of such a platform—the subject of the following chapter.

4.3 The Organizational Dimension: Fusion Teams and New Responsibilities

Opening up process development to the business side carries the risk of an uncontrolled spread of undocumented, insecure solutions ('shadow IT'). To prevent this, IT and governance leaders need to put guardrails in place—organizationally anchored in Fusion Teams, where business experts, IT developers, and security specialists work closely together and put the new division of labor with IT described in Chapter 3.4 into practice.

The Process Owner evolves into both an agile architect of the rulebook, designing the guardrails for human and digital employees, and a driver of value creation for the end-to-end process—the organizational counterpart to the role evolution of the Process Owner described in Chapter 3.4.

The Process Expert/Process Manager takes on a greater share of curating and quality-assuring process knowledge. Key Users become co-designers and trainers of digital colleagues. The Process Management Consultant in the CoE evolves into an enabler of organization-wide process and AI governance.

This role shift has to be actively supported. A differentiated, role-specific yet coordinated enablement program creates role clarity and builds competencies. Across all roles, a shared qualification foundation is needed: understanding of agile, value-oriented steering of end-to-end processes, data literacy, a solid grounding in AI and agents, and change-management competence.

Enablement is a key building block for culturally embedding the shift from rigid hierarchies to an agile, connected ecosystem—a process we support with learning journeys and community formats.

Want the rest of the argument now instead of waiting for next week's chapter? The full white paper — written jointly by BPM&O and bpExperts — is available on request. Write to sales@bpexperts.de and we'll send it straight over.