Skip to main content
Book an AI strategy call
GOVERNANCE

Where your data goes, and who can see it.

Ten controls, all decided before a build starts rather than defended afterwards. If any of them conflicts with your policy, the design changes, not the policy.

Named providersNo training on your dataAudit record per actionNo lock-in
THE CONTROLS

Ten controls, and when each one is decided.

Most of these are design decisions, which is why they are cheap to honour and expensive to retrofit. One of them, the human approval points, cannot be removed later, by us or by you.

See the six stages
01

Data minimisation

Fields a step does not need are removed before anything leaves your systems. An answer about delivery gets an order status, not a customer history.

Design stage
02

Named providers

Every model, hosting, telephony and verification provider is named in writing before build, with its processing region and retention policy.

Before build
03

No training on your data

Providers are configured, where the provider supports it, so submitted data is not used to train their models. Where a provider cannot guarantee it, it is not used.

Contractual
04

Jurisdiction constraints

If your policy prohibits data leaving a region, that is a design input that changes which providers are eligible, not a launch-week discovery.

Design stage
05

Human approval points

Customer-facing output held by default, low-confidence work escalated to a named person, destructive and financial actions always behind a human step.

Not removable
06

Audit record

Every system action, escalation, approval and override is written with who, when and what changed.

Per action
07

Access control

Role-based access in the operator interface, credentials held in your accounts, and a transfer of admin rights at handover.

Your accounts
08

Retention schedule

Stated in one sentence per system, verified against the provider settings rather than assumed.

Written down
09

Incident route

A named contact, a defined response window and a written record. Nexvix reports rather than waits to be asked.

One business day
10

Exit

The system runs without a Nexvix retainer. Support is optional and priced separately.

No lock-in
IN WRITING

Four commitments, and what each one costs us.

A commitment with no cost attached is marketing. These four are in the engagement terms.

A

Your accounts, not ours

Systems are deployed into infrastructure and model accounts billed to you. It costs us the easiest form of client lock-in there is.

Ownership
B

No retainer required

A delivered system runs without paying us anything. It costs us predictable recurring revenue.

Commercial
C

Providers named before build

Including the ones a client might object to, early enough to change the design. It costs us the option of quietly picking the cheapest.

Transparency
D

We report our own incidents

A named contact, within one business day, in writing. It costs us the comfort of silence.

Accountability
QUESTIONS

What security reviewers ask.

Is our data used to train an AI model?
No. Providers are configured so submitted content is not used for training, and any provider that cannot offer that setting is not used for client data.
Which providers see our data?
The ones named in your design document, with region and retention. That list is produced before the build, not on request afterwards.
Can everything stay in one region?
Often, yes, and where it is a hard requirement it constrains which providers are eligible. That constraint belongs in the design, and it changes the build rather than the launch.
What happens if the system gets something wrong?
Approval points mean most errors are caught before a customer sees them. Anything that reaches a customer is recorded, reported to you and fixed as part of the engagement, not as a change request.
Do you sign a data processing agreement?
Yes. Nexvix acts as a processor on client instructions for anything in a delivered system, and the DPA sits alongside the scope document.
Who can access the production system?
Named people, in your accounts, with role-based access. Nexvix access is time-limited and removed at handover unless you ask for a support arrangement.
WORKFLOW AUDIT

Send us your security questionnaire before the call.

One call, thirty minutes. Nexvix maps the process, names the measurable outcome and tells you whether a system is worth building. If it is not, you will hear that instead of a proposal.

A written workflow map you keep either wayA named measurable outcome, not a feature listA direct answer if the system is not worth building