Praev Information

Praev is the governed execution layer for AI-driven business processes. Most organisations automating with AI face the same problem twice over. Rule-based workflow automation is reliable but brittle: it does exactly what it was told, and breaks when reality changes. AI agents adapt, but they are probabilistic, and a process you cannot predict is a process you cannot assure. Teams end up running two disconnected stacks, and auditing neither properly. Praev runs both on one execution model. A business process is built as a flow : triggers, integrations, branching, human approval and any step in it can be handed to an AI agent that reasons, deliberates with other agents, or pursues a goal over time. Deterministic where the work is predictable, agentic where it isn't. The part that makes it deployable is the assurance layer underneath. Every action, deterministic or agentic, passes through policy middleware that evaluates it before it runs and can warn, require human approval, or block it outright. Controls cover prompt-injection detection, output filtering, capability allowlists, blast-radius limits, goal-drift scoring, confidence thresholds and violation circuit breakers. Every evaluation writes to a single immutable, tamper-evident evidence ledger, threaded by a correlation ID : so one query reconstructs an entire process end to end, across both deterministic and AI steps. Controls are mapped to UK GDPR, ISO 27001, ISO 42001, SOC 2, the EU AI Act, DORA and the OWASP LLM Top 10, and every compliance figure decomposes to the individual enforcement events behind it. Nothing is self-reported. Praev deploys as managed cloud, fully self-hosted, or air-gapped : so regulated organisations can adopt AI without data leaving their jurisdiction or their network, and small businesses can utilise automation and AI without huge cost. Built in the UK, for teams in regulated industries who need AI that is useful in production and provable afterwards.