From workflows to AI agents. With the same discipline.

WildcardTech has been automating business processes since 2018: first with a workflow platform of our own, then in projects in the legal and insurance sectors, later with services for document processing. Since 2024 we have been building AI agents into these processes, with the control and traceability that regulated industries demand.

Düsseldorf · Since 2018

1. 2018

   Start and Process Core

   WildcardTech starts out as a freelance business with one goal: workflow management for companies that need process automation but not the complexity and license costs of the large BPMN platforms. Our first product is Process Core, a workflow and case management system with a graphical editor in the browser: model, run and monitor processes in one application.

2. 2019–2020

   Law and insurance

   Our first larger projects come from two industries that live on processes, and we deliver them on enterprise BPMN platforms. In the legal sector we digitize contract management, deadline monitoring and case files — processes in which mistakes are expensive and traceability is mandatory. For insurers we automate applications, claims and policy issuance, and connect core systems that partly date back to the 1990s.

3. 2020

   WildcardTech GmbH

   The freelance business becomes WildcardTech GmbH, based in Düsseldorf.

4. 2020–2023

   Documents

   At the start and end of almost every business process there is a document: invoices, contracts, applications, official notices, correspondence. We build standalone microservices for document-based processes: text recognition with extraction rules per document type, PDF compression, merging and splitting documents, stamping exhibit bundles with page numbers and case references, and generating documents from templates. Each service is stateless, accessible via REST, orchestrated through BPMN workflows and usable as a building block in enterprise BPMN platforms.

5. 2023

   The turning point

   Two things happen at once. Our rule-based systems reach their limits: every new customer and every new document format means more work. And language models become good enough to understand context, prepare decisions and use tools. We start internal experiments: document classification, data extraction with language models, and agent prototypes that carry out simple process steps on their own.

6. 2024

   The pivot

   We move from rule-based process digitization to AI agents embedded in business processes. The agents research, check and prepare; when they are uncertain, they hand over to a human. Our first projects use LLM-based agents for document review and data preparation. When Anthropic publishes the Model Context Protocol as an open standard, we start building MCP servers for enterprise systems.

   The same year, the EU AI Act enters into force. It classifies AI systems by risk and, for high-risk systems, requires human oversight, documentation and transparency, among other things. Human approval steps and audit trails have been part of our workflows since our first projects in regulated industries. We also build up expertise in operating AI agents: monitoring, drift detection and documentation, so that agents do not just launch but run in production for the long term.

7. Today

   AI process engineering for critical operations

   We connect AI agents, humans and systems in business processes that stay controlled, traceable and independent of any single vendor. Our umbrella is digital resilience: cloud exit, AI on-premise and on-edge, open-source strategy. Our products Process Core, Edge AI Fabric, INTAKE and TCKT are open source under Apache 2.0. Process Core is a complete rewrite of the 2018 version. And we operate digital-resilience.cloud ↗, an independent directory of European providers.

1. 2018

   How processes really work.

   With every exception, special case and human factor that no process diagram foresees.

2. 2019–2020

   Control is the foundation.

   Regulated industries allow no shortcuts. Audit trails, permission concepts and human approvals come first; this is where our human-in-the-loop approach comes from.

3. 2020–2023

   Where rules reach their limits.

   Every new document layout meant new rules. Rule-based systems do not grow with the variety of the real world.

4. 2023–2024

   AI needs the same discipline.

   Systems that learn and adapt need the same control, traceability and human oversight we have built into workflows since 2018.

03 — Talk to us

We start with your process.

Not with the question of which agent we could build, but with the business process you want to change.

Talk to an engineer ↗