AI disclosure, fixed words, every change on record.
The person on the line is told they are speaking with AI. Regulated processes run on deterministic Brilliance™ StateGuard variants, and nothing reaches your customers that is not in a recorded, versioned catalogue.
Policy → permitted action → reviewable evidence.
Human accountabilities and permitted automated actions remain defined throughout the process.
- Human authorityBusiness purpose and sensitive decisions remain allocated
- System authorityOnly configured, reviewed actions within scope
What your compliance review starts from
ISO 27001
Vee Center Sp. z o.o.
ISO/IEC 27001:2022, certificate PL21/00000044 (SGS), valid to 3 November 2027.
EU/EEA
Processing location
Service data is processed in the EU/EEA as a rule; any transfer outside the EEA only under Chapter V GDPR and the agreement.
Art. 50
AI disclosure in the call
Customers are told they are speaking with an AI system, as the EU AI Act requires.
Six controls inside the conversation itself.
Governance that lives in the call, not in a policy binder. Each control is part of the process design and leaves evidence.
AI disclosure
Article 50 of the EU AI Act applies from 2 August 2026. In a Vee call, the disclosure is a step in the dialogue and does not wait for the person to ask.
Deterministic variants
Brilliance™ StateGuard runs variants up to fully deterministic. In regulated processes there is no autonomous content generation.
Closed catalogue
What the system can say is a closed, versioned catalogue. Studio-recorded voice (Vee Lector/UX) makes it a finite set a reviewer can hear in full.
Human boundary
Disputes, complaints, vulnerability signals and anything outside the rules stop the automated path and reach a person with the context.
Market rules
Calling windows, attempt limits, intervals and recording consent are configured per market and enforced by the platform.
Full reconstruction
Recording, logs, statuses and the scenario version show what was said in each call, and why.
No change reaches a live call without a record.
Wording, contact rules and integration behaviour change only through one release path. The interaction layer does not learn from production traffic.
- 01ProposeThe need and the intended effect, in writing.
- 02AssessImpact on compliance, privacy, customer treatment and the result.
- 03TestNormal paths and exceptions, before any customer hears it.
- 04Sign offA named owner on your side accepts the release.
- 05ReleaseA new version, monitored from the first call.
- 06ReviewOutcomes compared; the previous version restored if needed.
What compliance teams ask first.
Are you compliant with the AI Act?
We do not issue blanket compliance statements. We show the architecture: deterministic dialogue where the process needs it, disclosure in every call and a full evidence trail, and we assess your configuration with you in the Diagnostic.
Does a language model write the answers to our customers?
Not in regulated processes. Answers come from the closed, versioned catalogue. Statistical models recognise speech and intent; they do not write what is said.
Is our process high-risk under the AI Act?
Classification follows the intended purpose, not the technology. We assess it with you per use case and record the result before the operation starts.
What evidence can our auditors examine?
Scenario and rule versions, the content catalogue, change records, call and connection logs, escalation and stop rules, and the mapping from each interaction to its outcome.
Classify the use case before you configure it.
For regulated processes the Diagnostic includes an AI Act review: intended purpose, disclosure, human boundary and the evidence your compliance team will need.
Performance Diagnostic
Typically 2–4 weeks from data access · fixed scope and fee
We map one process with your team: volume and segments, contact rules, systems, current results and cost, and the measure of success.
You get Baseline, compliance map, business case and a pilot design you can approve, or a clear recommendation not to proceed.
Controlled Pilot
A defined period agreed in the Diagnostic
Vee runs the process on a limited population against a comparable reference, with your rules, quality gates and stop criteria in place.
You get The metric that matters, confirmed in your ledger, calendar or CRM.
Managed Production
Ongoing
Vee operates and improves the process at scale: service levels, Vee Analytics™ reporting, a regular performance review and controlled changes.
You get One accountable operation, priced per case or account.
No commitment for the process call. The Diagnostic includes an AI Act and data-protection review for regulated processes.
Walk the AI design with your compliance team.
Bring the intended use and the points where a decision affects a person. We will show where automation stops and what evidence each step leaves.