01 · REDACTION
Names, account numbers and contact details are removed on the way in, before the analysis runs, and the redaction is part of the record. It is rule A4, checked when a result is written.
Governance
REVEAL · GOVERNANCEKeep findings open to review. Review the evidence behind results, identify those flagged by the confidence gate and apply the controls agreed for your operation.
[01 / THE REVIEW]
From a finding to a reviewed decision.
Every result carries the passage and the rule it was judged against, and a result below the confidence threshold is flagged for a person.
- 01The evidenceEach finding carries the passage it was based on and the rule, quoted in full, it was judged against.
- 02The confidence gateBelow the confidence threshold a result says so and asks for review rather than presenting itself as settled.
- 03Human reviewReviewers score a slice against the same items, without seeing the machine result first; variance is measured per item, and a rule everyone reads differently is rewritten once.
- 04The responsibilitiesWho reviews, who approves and what reaches a worker are agreed for your operation, and nothing that ranks a person reaches anyone without a human in the path.
[02 / INFORMATION]
How information is handled.
Redaction is applied before the analysis runs, access follows your rules, and the record can be exported and checked without us.
02 · ACCESS
Operator, reviewer and executive views under your identity provider and your access rules.
03 · RETENTION
Results are kept on the retention agreed for your deployment and deleted on that schedule.
04 · THE RECORD
Input, rule, model and result under one hash, exported as a bundle you keep and can verify yourself.
[03 / THE FRAMEWORKS]
What this is held to.
Two regulations, one operational standard and our own rules. What each one asks, and the measure in the platform that answers it.
01 · EU AI ACT
A system used to evaluate work puts obligations on the operation that deploys it: a person in the path, results a team can explain, and disclosure that a machine produced them. The controls above are built for exactly that.
02 · GDPR
Personal data is removed on the way in, so the analysis reads a conversation without the identifiers in it. Results are deleted on the retention you agree, and handling is defined for your deployment. Read the policy.
03 · COPC CX STANDARD
A framework aligned with COPC releases 7.1 and 8.0: customer, business and compliance critical errors are scored apart from weighted non-critical ones, so one critical error fails the transaction instead of being averaged away.
04 · OUR OWN RULES
Eight rules every FQ agent follows, checked when a result is written; an output that breaks one is refused. Four of them are the controls on this page. Read the eight.
Which obligations apply to your operation depends on how you deploy the platform, and the scope is agreed in your proposal. Rights requests are answered within thirty days and the supervisory authority is the Hellenic Data Protection Authority. A machine result never overrides a reviewer on a dispute, and alignment describes a framework, not a certification. We hold no ISO or SOC certificates yet, we protect data with measures suited to our size, and nothing here is legal advice.
[04 / IN DEVELOPMENT]
Additional checking and risk controls, in development.
A further AI checking stage and a set of risk controls are being designed. They are not available today and are not part of the controls above.
THE CHECKING STAGE
An assessment of results inside each analysis run, surfaced to a person only when needed. Planned.
RISK CONTROLS
Classes of risk on an interaction, such as personal data, vulnerability or regulatory exposure, with the handling each requires. Planned.
Availability and scope of development-stage controls are confirmed in your proposal, not on this page.
[05 / CONSOLE]
Discuss the review controls you need.We will walk through the controls, their scope and the responsibilities of your team.
Let’s makequality visible.