Plenio

The problem

Companies deploying AI inside their products and operations have no defensible record of human oversight. When an AI tool generates code, drafts content, makes a recommendation or ranks a candidate, that output flows straight into a downstream decision with no captured moment of human judgment. Two failures follow: a compliance failure — Article 14 of the EU AI Act, ISO 42001 Control 8.4 and SOC 2 CC2.1 each require evidence of oversight, not policy alone — and an operational failure: no organisational memory of why AI-influenced decisions were made, or who stood behind them.

The mechanic

Interception at the human–AI decision boundary, recorded as a tamper-evident audit record — per event, per human, rather than per system.

Plenio manifests as a sign-off for produced artifacts (PRs, drafts, recommendations) and as a pre-prompt risk assessment plus post-hoc attestation for consumptive AI use — CV vetting, credit decisions, anything where a human reads AI output and then acts on it elsewhere.

Where it sits in the market

Existing tools produce evidence of oversight at the policy level (GRC automation — Vanta, Drata) or the system level (AI-native governance — Credo AI, Holistic AI). Plenio produces evidence at the event level: one human, one event, one signed attestation. The legal wedge is Article 26(2) — the deployer must assign qualified personnel who actually exercise oversight. That is event-level evidence nobody else was producing.

What I did

• Architected the system: an immutable audit ledger, pluggable capture sources and a workflow layer, with two interaction modes (artifact sign-off and consumptive attestation).
• Specified the v1 common-core audit-record schema every capture source conforms to.
• Designed and built the surfaces: a Blazor console, a .NET API and a browser extension that captures the decision moment in-context.
• Deployed machine learning into the platform with deliberate restraint. Because auditable accuracy is a base principle here, ML assists the capture and classification of decision moments — it never becomes the unaccountable judge of them.

Why it’s here

Plenio is evidence of end-to-end product leadership and platform architecture: taking a regulatory shift, finding the defensible wedge, and shipping a real, governable system around it — compliance designed in, not bolted on. It’s also the evidence layer inside Codamor, where every AI decision point emits Plenio’s per-event, named-human oversight records by default.