An operating model aligned with data, IP and risk
NyxAI · Clarity before it becomes obvious.
AI agents for your business processes — controlled, measurable, local if you need it.
AI agents and decision systems for demanding business processes and sensitive data. Start with a defined Agent Pilot — local, private or hybrid.
Typical pilot scope: €20,000–40,000 excluding VAT · 4–6 weeks · one defined workflow
Response within 2 business days.
Approvals, fallbacks and traceable handoffs
Test cases, quality criteria and audit trails
A bounded scope instead of an uncontrolled platform bet
Starting points
Where you can begin.
Agent Pilot
A controlled pilot for one concrete business workflow — local, private or hybrid.
Explore the Agent PilotAI Workflows
One clear process. Your data. A measurable pilot. 27 workflow starting points and 8 use cases in depth.
See the AI workflowsDecision intelligence
Our architecture approach connects operational data, domain knowledge and simulation within clear guardrails.
See the platformAI workflows
One workflow. One pilot. One sound decision.
Invoice review
Surface discrepancies before an invoice moves forward.
Discuss this workflowOrders from email
Reduce retyping between customer email and order entry.
Discuss this workflowInternal knowledge
Get internal answers with sources instead of searching many repositories.
Discuss this workflowMaster data changes
Review critical changes before they affect downstream work.
Discuss this workflowHow it works
Four steps. One clear decision.
- 01
Fit call
Review the workflow, data, accountability and expected value.
- 02
Scope
Define pilot boundaries, metrics, operating model and safety controls.
- 03
Build & test
Build, evaluate and document a controlled agent.
- 04
Readout
Receive evidence, risks, cost shape and a clear recommendation for the next step.
Who is behind it
Lars Heppert
Founder and Managing Director of NyxAI GmbH
Technology operator. AI builder. Long-horizon investor. Author of “Coding for Fun mit Python” (2010).
DeepSquare · Research by NyxAI
NyxAI builds DeepSquare.
DeepSquare is NyxAI’s engine and research platform. Chess provides a precise test field for search, evaluation and decisions under time pressure.
- A position is evaluated in twelve understandable blocks — king safety, pawn structure, piece activity and more.
- The Lichess bot DeepSquare-ai explains its evaluation on request: type !explain in the chat.
- Applications beyond chess require their own data, assumptions and tests.
NyxAI / Insights
Think clearly. See further.
Perspectives on AI agents, sovereign systems and decisions under uncertainty.
What does an AI agent pilot cost?
Data access, integrations, evaluation and handover shape a pilot budget. A defined scope makes the commercial decision easier to assess.
Read the analysisAI proof of concept (PoC): what it should prove before you invest
A PoC is useful when it answers a named question with representative data and agreed criteria. What counts as success, and what counts as a reason to stop, matters more than the demo.
Read the analysisLocal AI or cloud: Which architecture fits?
Model location answers only part of the deployment question. Data sources, identities, logs and backups belong in the same decision.
Read the analysisNext step
Which workflow deserves a controlled agent?
An initial idea is enough. Describe the task and the outcome you want. Together, we can assess whether an Agent Pilot or a different starting point makes sense.