Governed AI-assisted workflows

AI → Human
→ Evidence

Make AI-assisted decisions reviewable.

TruncAI-OTiX is an experimental reference architecture for governed, evidence-backed AI-assisted workflows. It preserves the chain between human intent, information sources, AI analysis, recommendation, human decision and evidence — without giving the AI operational control.

Explore the demonstrator Technical overview
Truncated octahedron
◇
TraceableFrom intent to decision, with structured evidence.
✓
GovernedPolicies, limits and human oversight by design.
◎
Evidence-backedReconstruct relevant workflow events without exposing sensitive data.
◉
Human-centricAI recommends. Humans decide.
The problem

AI can recommend. Can you reconstruct why?

As AI systems and agents become part of real workflows, a plausible answer is not enough. Teams may need to reconstruct what was asked, which sources and tools were used, what the AI was allowed to do, where uncertainty was recorded, and what the human ultimately decided.

From intent to evidence

Govern the workflow, not just the answer.

The current baseline is deliberately consultative: observe, analyse and recommend within an authorised scope. It does not autonomously issue operational changes or control physical systems.

IntentPurpose and authorised scope
SourcesContext and provenance
AI analysisProcessing inside declared limits
RecommendationConsultative output and uncertainty
Human decisionReview, accept, reject or escalate
EvidenceRelevant records for reconstruction
Reference demonstrator

One workflow. One recommendation. One evidence trail.

The public demonstrator is designed to use local, synthetic or sanitised information. The energy domain is used only as a reference workflow; there is no operational control over energy equipment.

Example recommendation

“Review the observed anomaly and compare it with recent authorised context before taking any operational action.”

CONSULTATIVEHUMAN REVIEW REQUIREDUNCERTAINTY RECORDED
> show_evidence()

purpose........ anomaly review
sources........ authorised / sanitised
capability..... observe · analyse · recommend
policy......... no operational change
uncertainty.... explicit
recommendation. consultative
human_decision. pending / required
outcome........ not asserted
Current status

Internally verified. Explicitly limited.

TruncAI-OTiX is presented as an internally verified reference baseline within a controlled, non-operational scope. External technical, legal and sector-specific evaluation remains necessary.

Not claimed
  • Regulatory certification
  • Demonstrated legal compliance
  • Industrial validation
  • Production readiness
  • Autonomous operational control
  • Proven superiority over alternatives
External evaluation

Evaluate it. Challenge it. Take it further.

The next stage is external evaluation: technical assessment, integration assessment, licensing or technology transfer / acquisition. Public materials will be released progressively while restricted evidence, real datasets and transfer-sensitive assets remain private.

Request technical information Public GitHub — planned