AI Decision Intelligence
Put Orcanaut in front of your existing AI provider — one config change, nothing else. Every decision is logged immutably, explained in plain language, and ready for the EU AI Act. Built for risk and compliance teams, not just data scientists.
#DR-8F2A1C
credit-scoring · v2.3.1
Business context · via Enrich API
Why this decision
Application declined primarily due to a high debt-to-income ratio (42%) and a short credit history.
Top contributing factors
For EU teams running AI on decisions about people
Fintech
credit scoring, BNPL
HR tech
hiring, CV screening
Insurance
claims, scoring
Health
triage, eligibility
Point your existing OpenAI, Azure, or Anthropic calls at the Orcanaut proxy and swap your API key — one config change, nothing to install. Prefer to integrate in code? Drop in the SDK instead. Either way, every call is captured automatically — no pipeline rewrite.
Append-only event store with a separate schema per tenant. No UPDATE, no DELETE. Tamper-evident hashes and configurable retention — minimum six months.
Factor importance is captured alongside each decision, then an LLM turns the numbers into a sentence. "Why was this application rejected?" — answered in language a customer or auditor understands.
Record every override, reviewer, and reason alongside the model's call. Article 14 oversight becomes provable — not a policy doc nobody can verify.
Watch decisions drift from real outcomes over time and catch bias across protected groups — before a regulator or a journalist does it for you.
One click produces the audit file ČNB, NBS, and ČTÚ expect. Local regulation built in — not a generic US dashboard you have to translate.
EU AI Act
High-risk AI systems carry hard obligations. Orcanaut covers the three that hurt most to build yourself.
Article 12
An immutable record of every decision, captured automatically over the system's lifetime.
Article 14
Overrides, reviewers, and reasons captured inline — oversight you can actually prove.
Article 86
A clear, human-readable explanation for every decision that affects a person.
Standalone high-risk (Annex III) obligations now apply from December 2027 (under the EU's May 2026 Omnibus agreement), with non-compliance penalties of up to €15M or 3% of global turnover. But the deadline isn't the point: when AI already decides about people, immutable logging, plain-language explanations, and a defensible audit trail are infrastructure you'll need either way. Orcanaut ships exactly that on day one — a drop-in proxy you run as fully-managed SaaS or on-prem in your own infrastructure, with regulator-ready export for the ČNB, NBS, and ČTÚ. SaaS gets you live with one line and GDPR covered by a DPA; on-prem keeps every byte inside your tenant.
Change one environment variable — point your OPENAI_BASE_URL at the Orcanaut proxy. No code changes, no pipeline rewrite. Works with OpenAI, Azure AI Foundry, Anthropic, and any OpenAI-compatible endpoint.
The proxy intercepts each request and response — model version, hashed inputs, outcome, and timestamp — and writes an immutable record to your Decision Ledger. Nothing can be quietly rewritten.
Tag each decision with business context via a single API call — who was affected, what the decision type was, and which factors drove it. Risk officers see plain-language explanations and export a regulator-ready audit file on demand.
We're onboarding a small group of design partners in fintech, HR, and insurance. Join the waitlist and we'll be in touch.