Who may act, and who must ask.
NobleCloak builds Approvals for AI agents. This page is the idea behind it, and the two people who stand behind it.
Every organization that has ever existed runs on one question: who may do this alone, and who has to ask first.
Nobody wrote it down as software; people learned it by being people. You join a company and within a week you know what you can just do and what you walk down the hall to ask about.
AI agents are the first workers who never learn that.
They arrive with the keys and no sense of the room, and they move at a speed the walk down the hall was never built for.
Approvals: the first place that question becomes a transaction.
What the agent may do on its own, what it must escalate, who gets asked, and how that handshake happens at the speed of AI without the person being left out of it.
Approvals are the first way an organization hands authority to an agent, not the last.
Delegation, limits, escalation, quorum, review. Every way an organization hands authority to a person is a way it will have to hand it to an agent.
Who you're working with.
NobleCloak is built by two people, and both of them will be in the room when you need them.
David McNelis
LinkedInOver nearly 13 years at Emerging Threats and Proofpoint — one of the world's largest email-processing platforms, screening billions of messages a day — David led engineering on threat-intelligence pipelines and reputation systems behind a platform that serves more than half of the Fortune 1000. At NobleCloak he builds the isolated, audited substrate that lets customers see and prove exactly what their AI is doing with their data.
Aaron Nicely
LinkedInAaron spent years of his product career in third-party risk at Venminder, building the tools banks and credit unions use to vet the vendors they depend on, and served as Vice President of Product at Driven Data, where he led product teams. His background runs through data, analytics, and business intelligence — and NobleCloak began with the question his risk work kept raising: when the vendor is an AI, who checks its work? He leads product and design.
We state what we can prove.
No SOC 2 report yet; preparing toward our audit. The full claim set, including what we cannot yet claim, is published at our trust center.
No demo maze. No SDR queue. Just us.
Write to us and one of the two people above answers. If you are examined and want a first answer before you talk to anyone, start with the AI Trust Report.