Resources · Evaluating Cogniveil
Evaluating Cogniveil
If someone sent you this link, they think Cogniveil can carry part of your regulated workload and wanted you to see the whole picture on one page. Here it is.
What it is
Cogniveil makes regulated workflows deployable on AI, producing the work and the proof together. The work is done by governed AI coworkers: defined roles with a job description, measures, guardrails, memory, and a named human supervisor. The coworker does the work. The human keeps the decision. Every output is built only from your approved sources, cited line by line, and recorded so it can be reconstructed later.
Architecture in brief
Your documents stay where they are; we connect to them where they already live. When someone asks, the system retrieves the passages that actually answer, by content, under your access rules. The answer is built only from those passages; anything it cannot tie back, it does not say. When it cannot ground a claim in the approved sources, it says so instead of guessing. Every step is recorded: who asked, what was retrieved, what was answered, what a human changed. How it works, in full
The four certifications
Cogniveil is certified to ISO 9001 (quality management), ISO 27001 (information security), ISO 27701 (privacy information management), and ISO 22301 (business continuity). Certificate metadata is withheld until verified from the signed certificates. Certificates and regulatory posture
On regulation: DORA compliant by design, with an ongoing program. Mapped to the EU AI Act, where use-case risk classification stays with the deployer. Built to support GDPR. Aligned with NIS2.
Deployment models
Four, selected by your constraint, not ours: cloud, private cloud, on premises (your hardware or a managed appliance), and fully air-gapped. Data residency follows the deployment: in the strictest models, your information never leaves the place you keep it. Deployment
What your security review will ask
- Where is our data stored, and can it leave?
- Who can access what, and how is that enforced?
- Is customer data used for training? It is not. We do not train on your data.
- What happens when the system cannot support an answer?
- Can you evidence continuity and recovery?
- Can we see the documentation? Yes: certificates with full scope, architecture diagrams, and the data processing agreement are in the security pack. Get the security pack
How pricing works
Per coworker, not per seat: you pay for defined units of work, not for logins. An engagement starts small by design: process mapping, a two to four week trial with agreed measures, then a one-month pilot to size consumption. No figures on the site, because they depend on your deployment; the model and parameters are public. The pricing model
The next step
The evaluation that actually settles it is not a document. Bring one regulated workflow, define what success means, run it with Cogniveil, measure the result, and keep the evidence.
Bring us a workflowNext step
Start with one workflow
Bring the work, the approved sources, and the people who must stand behind the answer.
Or contact sales@cogniveil.ai