The product
What is g-nosis?
g-nosis makes AI safe to adopt in regulated firms. It is the harness for every decision a person or an AI agent makes: kept with its reason, who answers for it, the authority it ran under and its outcome, and replayable as it stood.
How is this different from model governance?
Model governance tests the model: accuracy, bias, drift, version history. g-nosis keeps the decision: what the AI advised, who challenged it and why, the call and the named person who made it, the rule in force and the outcome.
The model is rarely what goes wrong. The mess around it is.
We already use an AI agent manager. Do we need both?
Yes: they keep different things. Keep Agent 365 for the AI agent; g-nosis reads its registry and keeps the decision. Copilots read documents, AI agent managers track the AI agent, and model tools test the model. The decision your name is on falls between them.
Which decisions does it keep?
The same five steps wherever decisions get made: the question, the options, the challenge, the call and the outcome. A claim referral, an investment committee's call and a pre-trade exception are kept the same way. See the worked decisions.
Does it slow people down?
No. g-nosis captures the reasoning as part of the work, so nobody has to write it down afterwards. Inside its limits an AI agent acts at once, so nobody waits. Anything outside goes to a person.
AI agents
Which AI agents does it work with?
Copilot Studio, Agentforce or home-built: any AI agent, any host. Every AI agent's decisions land in one record. AI agents write theirs through a hook, their traces or a gateway.
What is a mandate?
The limits a named person signs for an AI agent: what it may decide, up to what limit, until when. Inside its mandate an AI agent goes ahead at once. Outside it, a named person signs it off.
What does “tested with NVIDIA OpenShell” mean?
OpenShell holds the boundary; g-nosis holds the decision. A mandate given in g-nosis compiles into OpenShell policy. OpenShell checks the action, not the amount; g-nosis checks the limit. g-nosis also checks its record against OpenShell's audit log, so a missing decision is found, not assumed. The test ran in October 2026.
Can AI change our rules?
Not by itself. AI drafts a rule from an override, or from a pattern your people follow that no rule states. A person adopts it or not. Rules change only when a person adopts them, and old decisions stay judged by the old rules.
Your records and controls
Where do our records stay?
In your systems. g-nosis connects to any system that has a web API and looks records up where they are kept.
The decision record is kept in g-nosis, with a copy of what the person saw when they decided, so it can replay as it stood. Documents you add to the library are kept in g-nosis too.
Which systems does it connect to?
Microsoft 365 and HubSpot, plus Slack, Salesforce, Google Workspace, Jira, Confluence and Workday, and any system that has a web API.
What controls are on every decision?
Four. A reason is required to overrule the advice in the app. Whoever proposed a decision cannot approve it in the app. A signed decision is replaced, not edited. Any decision replays as it stood, under the rules in force that day.
How does it run?
Decisions are made in the g-nosis desktop app for macOS. Admins manage teams and intake keys from a web portal. The AI assistant runs on each person's own Claude plan.
Getting started
What does a pilot involve?
Ninety days on one decision area, starting with claims. Weeks 1 to 2, map: every AI agent, model and spreadsheet in that area, and who answers for each. Weeks 3 to 10, capture: every decision recorded as it is made. Weeks 11 to 13, proof: precedent at work, the evidence pack your supervisor would ask for, an outcome review, and a report for your risk committee. See the founding pilot.
What do we have to change to start?
No new committees, no change to who decides, no re-platforming. Before week one: your security review, DPIA and data-feed approvals, with acceptance tests agreed before you sign.
Who owns the record?
You do. The record is yours, whatever you decide next.
What happens in a walkthrough?
Pick a decision your firm made last quarter, or one an AI agent made that nobody can answer for. In forty-five minutes we will show you the record it would have left, and what you could ask it a year from now. Book a walkthrough.
bring us one decision you have to explain.
A claim referred above authority, a complaint upheld, a trade outside its limit, a committee's call, a decision an AI agent made that nobody can answer for. We will show you the record it would have left.