What does a "compliant platform" mean under the EU AI Act?

Less than vendors imply. No platform makes you compliant on its own. Cogniveil is mapped to the EU AI Act: the controls a deployer must evidence, transparency, record-keeping, human oversight, data governance, are standing properties of the platform, so your program gets its evidence as a by-product of normal use. Use-case risk classification under the AI Act stays with the deployer. EU AI Act guide

Is our data used to train models?

No. We do not train on your data. Your documents are used to answer your organization's questions, under your access rules, and for nothing else.

Is our data isolated from other customers?

Yes. Each customer's environment is separated, with two tenancy models, dedicated or logical, selected per customer. Access inside your environment is enforced at the identity layer: a user can only retrieve what their role entitles them to see. Security architecture

Can it reach a legacy system that has no API?

Yes. Where a system has no interface at all, the agent signs in the way a person does, with credentials you provide and under the same governance as everything else. This is what makes legacy and air-gapped systems reachable without an integration project. Integrations

Can end users be limited to running workflows only?

Yes. End users can be restricted to executing published workflows and working with the coworkers they are entitled to, with no access to authoring or configuration. Building and changing workflows stays with the administrators you designate.

Do we need developers to configure it?

No. Guardrails, coworkers, and workflows are configured by administrators through the studio, without writing code. The coworker does the work. The human keeps the decision. Setup is part of the engagement: we build the first workflow with you rather than handing you an empty toolbox.

Per coworker, not per seat. You pay for defined units of work with a job description and measures, not for how many people log in. The parameters that drive cost are on the pricing page; the figures are a conversation, because they depend on your deployment. See the pricing model

How does an engagement start?

With one workflow. We map the process with you, agree what success means, run a two to four week trial against those measures, then a one-month pilot to size real consumption. You see the result measured on your own work before you commit to more. How pricing and engagements work

What happens when it does not know?

When it cannot ground a claim in the approved sources, it says so instead of guessing. That behavior is enforced, not encouraged: anything the system cannot tie back to an approved source, it does not say. The gap goes to a human, on the record. Verification and citations

Can it run air-gapped?

Yes. Cogniveil deploys in cloud, private cloud, on premises on your hardware or as a managed appliance, and fully air-gapped for environments with no external connectivity. The strictest deployments answer the data question structurally: nothing leaves, because nothing can. Deployment models

Do you replace our KYC or screening vendors?

No. We support the process. We are not the vendor. You keep your identity provider and your screening provider; Cogniveil runs the process around them: calls the provider, reads the result, cross-references your file and the regulation, documents it, and escalates it. KYC, KYB and onboarding

How is this different from our GRC tooling?

GRC tools manage the estate: they record policies, risks, and controls, and track whether obligations are met. Cogniveil produces the regulated work itself, the analysis, the check, the file, together with the evidence required to stand behind it. The two meet, they do not compete: the record Cogniveil produces is exactly what your GRC process wants to consume.

Resources · FAQ

Frequently asked questions

These are the questions buyers ask before a first conversation. If yours is not here, ask it on a call and we will answer it there.

Thirteen direct answers

No. A copilot helps a person with a task the person still owns, and the person still verifies everything. Cogniveil produces the regulated work and the proof together: the output arrives grounded in your approved sources, cited, and recorded, so a qualified person can check it instead of redoing it. The difference is not the answer. It is whether you can stand behind the answer.

Still deciding?

The fastest way to answer the question we cannot answer here, "will it work on our work?", is to bring one workflow and measure it.

Bring us a workflow

Next step

Start with one workflow

Bring the work, the approved sources, and the people who must stand behind the answer.

Bring us a workflow

Or contact sales@cogniveil.ai