Advisory Hour
per hour
For a founder or team with one sharp question.
- 60-minute video call
- Written notes and next steps within a day
Practice
AI orchestration
From AI agents to deterministic workflows.
We build AI agents, pipelines, automation and custom software, from autonomous agents where judgement helps to deterministic workflows where every run must match. Every step is versioned, every output checked, every decision traceable. Then we map them against the EU AI Act and GDPR, so they can be defended as well as used.
AI that holds up under scrutiny
An agent or a workflow holds up under scrutiny when it behaves predictably within the limits you set, stops itself when an output fails a check, and leaves a record of why. Where you need the same answer to the same input every time, we make it deterministic. That is engineering and governance together, and it is what this practice delivers.
What this covers
- AI agents, pipelines and automationAI agents, LLM workflows, n8n orchestration and integrations
- Custom software and dataPortals, data infrastructure and hosted digital assets
- Evaluation and quality assuranceRepeat-run, regression and adversarial testing of AI outputs
- AI governanceEU AI Act classification, Art. 25 boundaries, auditability and provenance
- Data protection by designGDPR-aware retention, deletion and access
- Adoption and trainingWorkshops, talks and capability building for teams
Fixed-scope offers
Talk or Training Session
per session
For teams, conferences and programmes.
- Up to 90 minutes
- Slides and materials to keep
AI Workflow Audit
per workflow
For teams already running an AI or n8n workflow.
- Repeat-run consistency test
- Failure-mode map
- A prioritised fix list
AI Adoption Workshop
half day or full day
For a team of up to 15.
- Where AI helps
- What to automate first
- What not to trust
- A 30-day plan
AI Governance Review
per system
For teams putting an AI system or use case into production.
- Risk classification, including Annex III
- Your role and where Art. 25 could change it
- Data protection by design
- A gap list and what to fix first
AI Orchestration Sprint
2 weeks
For one production AI workflow, built end to end.
- Versioned prompts, so every change can be traced
- Output checked against a schema
- A quality gate that fails closed: an output that fails a check is not sent
- Documentation and handover
External Project Management
monthly retainer
For teams that want the work steered, not just delivered.
- Weekly planning and check-ins
- Vendor and freelancer coordination
- Progress reporting to founders or the board
Ready-made AI Automations
licensed per company
For companies with recurring back-office paperwork.
- Set-up in your tools
- Checks before anything is issued
- Licence per company
Every engagement is scoped and quoted in writing before work starts. ZORRI is not a VAT payer, so quotes are final, with no VAT added.
Questions
What is AI orchestration?
AI orchestration is the design of a workflow in which one or more AI models, data sources and tools run in a defined order, with checks between the steps. The goal is output you can trust and trace, not a single clever prompt.
What does “fails closed” mean?
If an output does not pass its checks, the workflow stops instead of passing it on. A wrong answer is held back rather than shipped.
Is the AI Governance Review legal advice?
No. It is a technical and operational review: risk classification, your role under the EU AI Act, documentation gaps and design fixes. Your lawyer can work from it.
Which tools do you build with?
Large language models (including the Claude API), n8n for orchestration, and integrations with the systems you already use. The choice is agreed during scoping.
How is an engagement quoted?
Every engagement is scoped and quoted in writing before work starts. ZORRI is not a VAT payer, so quotes are final, with no VAT added.
Have an AI system in mind?
Describe what it should do, where its inputs come from and who relies on its output. You'll get questions back, or a scope.
Email us