Frequently Asked Questions
Duologs FAQ
Duologs is execution authorisation infrastructure for machine actions — the control layer between machine decisions and real-world actions.
What is Duologs?
Duologs is execution authorisation infrastructure for machine actions — the control layer between machine decisions and real-world actions. In a Duologs-controlled flow, no high-impact action executes without valid authorisation.
What problem does Duologs solve?
As AI systems move from generating outputs to executing real-world actions, most governance focuses on the model. The real risk begins at execution. Duologs addresses the gap before execution — at the commit moment — where authorisation, policy and context checks must happen.
What is the commit moment?
The commit moment is the point where a machine decision becomes a real-world action. Intent is cheap. Execution is irreversible. The commit moment is where trust must exist.
What is a consequential machine action?
Any action taken by an AI system that produces a real-world effect that is difficult or impossible to reverse — such as initiating a payment, modifying production infrastructure, granting identity or access, sending external communications, submitting a regulatory filing, or triggering an external workflow.
How is Duologs different from monitoring or observability?
Monitoring tells you what happened after execution. Duologs operates before execution — at the point where machine intent becomes real-world action. It is a control layer, not a visibility layer.
Is Duologs an approval tool?
No. Duologs is infrastructure designed specifically for the authorisation of machine-initiated actions — with policy enforcement, context verification and audit-grade accountability at the execution boundary.
Who is Duologs for?
Teams building, deploying or governing consequential AI systems — CTOs, platform and infrastructure leaders, AI product and deployment leaders, security, governance and enterprise architecture teams, and operators of agentic AI workflows.
How can I assess my organisation's exposure?
Use the AI Action Risk Assessment to evaluate your agentic AI authority maturity, or explore the Observatory to see real-world signals of what can go wrong when machine actions execute without authorisation.
How does Duologs work technically?
Duologs sits at the execution boundary of an agentic system. When an AI agent attempts a consequential action, Duologs evaluates the action against defined authority boundaries, policy and context, requires the appropriate authorisation, and produces an audit-grade record of what was authorised, by whom and under what conditions.
What kinds of machine actions does Duologs authorise?
High-impact actions where the outcome is difficult or impossible to reverse: payments and financial transactions, production infrastructure changes, identity and access grants, outbound communications to customers, regulators or partners, regulatory and legal filings, and triggers into external systems and workflows.
Does Duologs slow down AI systems?
Duologs is designed to intervene only at the commit moment for consequential actions, not on every model call. Low-risk actions can continue to execute freely; only actions that cross defined authority thresholds require authorisation. The aim is to make consequential AI safer to deploy, not slower to use.
When will Duologs be available?
Duologs is currently working with a small early access cohort of teams building, deploying or governing consequential AI systems. Organisations interested in joining can request early access to be considered for the next round.