Applied systems, not slideware.
These projects are working environments where methods, architectures and tools are developed before they become client work. They show the same pattern: human control, auditability, local sovereignty and practical delivery.
Negotiation Agents
Decision training under uncertainty.
AI-powered diplomatic simulations with 41 behavioural archetypes, multi-channel integration and full audit trail. Negotiators train against realistic counterparts — without real consequences.
- 41 behavioural archetypes
- Multi-channel negotiation (text, voice, video)
- Full audit trail and replay
- Scenario editor for custom scenarios
- Real-time coaching and debriefing
PolicyMonitor Studio
From policy signal to technical impact — automated.
Automated source monitoring, evidence-based triage and report workflows for policy teams. From source capture to finished briefing — in minutes, not days. PolicyMonitor also offers a free Gold-Plating analysis for an initial deviation scan and a paid technical screening for source-verified reporting.
- Automated source monitoring (EU institutions, news agencies)
- Evidence-based triage and relevance filtering
- Free Gold-Plating analysis for an initial deviation scan
- Paid technical screening with source-linked traceability
- Report workflows with templates
- Stakeholder mapping and impact analysis
- Export to common formats (PDF, DOCX, HTML)
ShellAgents
AI-driven automation without vendor lock-in.
Modular workflow automation with file-based traceability and auditable pipeline. Every step is transparent, reproducible and under your control.
- Modular agent architecture in pure Bash
- File-based traceability (every step documented)
- Auditable pipeline with complete log
- AI integration (local models and API)
- No vendor lock-in, no proprietary dependencies
More about ShellAgents
ShellAgents is an AI orchestration system that breaks complex tasks into small, well-defined steps and assigns each step to a specialised AI agent. Rather than producing answers from a single black box, ShellAgents coordinates a team of agents — each with a clear role and mandate — to deliver results collaboratively. The system works across domains: research, analysis, document preparation, or administrative workflows.
The core design principle is auditability. Every step of every workflow is documented in plain, human-readable files: which agent did what, what sources were consulted, what quality checks were applied, and whether any issues were flagged. A structured task sheet (the “Laufzettel”) records responsibilities, dependencies, and review outcomes — directly comparable to the audit trail requirements in Articles 9, 12, 13 and 14 of the EU AI Act. Nothing happens in the background without a record.
This matters for risk and compliance. The EU AI Act imposes logging, traceability and human oversight requirements on AI used in high-risk contexts — HR decisions, credit assessments, public procurement, critical infrastructure. Violations carry fines up to €15 million or 3% of global turnover. ShellAgents was designed for accountability before these rules existed: the architecture that satisfies the Act’s requirements is not a compliance patch, it is how the system works.
Beyond the AI Act, organisations in financial services, healthcare and public administration face additional sector-specific requirements (MiFID II, EBA guidelines, MDR, procurement rules) that demand explainable, documented AI outputs. ShellAgents provides this by design. AI handles the workload, but the audit trail stays with you.
→ Read the full analysis: ShellAgents and the EU AI Act
Typical work patterns
Operations Stabilisation
Problem: IT runs but is fragile — every change creates side effects.
Result: Documented dependencies, monitoring, structured maintenance cycles.
Security Resilience
Problem: Security is based on assumptions, not on tests.
Result: Validated backups, tested recovery, prioritised measures.
Architecture Clarity
Problem: Grown landscape without a big picture — migration or consolidation is blocked.
Result: Target architecture, migration path, executive decision brief.