Our approach

FQ CENTRAL · OUR APPROACH TO AIAI should support human judgement. Our approach starts with the people affected by the systems we build. It shapes how we develop, use and review AI, and the responsibilities we retain for its use.

[01 / THE VIEW]

AI should support human judgement.

The manifesto asks for one thing: give people back the time to lead. The Constitution makes that binding for the AI we build and use, and we apply it to our own agents first.

We treat the AI in our products and in our own work as a coworker in one sense only: it is held to written rules, its work is reviewed, and the people who use it keep the decisions. Accountability stays with people.

[02 / THE CONSTITUTION]

Seven principles for the way we design and use AI.

The FQ AI Constitution, written with our Responsible AI lead. Principles here; what is implemented in REVEAL today is on the Governance page.

01

Human dignity above efficiency

No person is reduced to a score without context, explanation or recourse.

02

Transparency by default

Every AI-produced line is identifiable as such, with a readable explanation of how it was reached.

03

Fairness is measurable

Claims about fairness are backed by specific tests on specific data, with results published.

04

Human authority over AI judgement

No FQ system makes a final decision about a person’s employment, pay or standing without human review.

05

Data minimalism and sovereignty

Only the data the task requires, no secondary use without consent, and the data belongs to the client.

06

Worker protection, not surveillance

REVEAL exists to coach and improve. We will not sell, configure or support deployments whose purpose is punitive action on AI scores.

07

Continuous accountability

We subject our own AI systems to ongoing monitoring and testing, and an ethics agent reviews our agents’ output against these rules.

[03 / THE COMMITMENTS]

Human dignity above efficiency. Worker protection, not worker surveillance.

Two of the Constitution’s governing principles, in the order they take priority. REVEAL exists to coach and improve, not to surveil and punish. FQ Central will refuse to sell, configure or support deployments where the primary purpose is punitive action against workers based on AI scores.

WRITTEN INTO EVERY CONTRACT

  • Onboarding includes a Use Case Declaration · the client states how REVEAL outputs will be used.
  • FQ Central reserves the right to terminate a licence if the platform is used primarily for punitive workforce management.
  • Automated actions · coaching triggers, performance alerts · require human approval before reaching the worker.

WRITTEN INTO THE COMPANY

  • No amendment may weaken human dignity, human authority or worker protection without unanimous co-founder approval and a published explanation of why the change is necessary.
  • The Constitution is designed to exceed current EU regulatory requirements, not merely meet them. Each principle maps to its counterpart in the OECD AI Principles and the GDPR.
  • Every AI output that evaluates a person’s performance must preserve the dignity of the person being evaluated · never a score without context, explanation or recourse.

[04 / THE RULES]

Eight rules every FQ agent follows.

The agent rules of the Constitution. Each is checked when a result is written; an output that breaks one is refused.

A1

No unreviewed worker evaluation

Nothing that ranks or penalises a named person reaches anyone without a human in the path.

A2

Flag uncertainty

Below the confidence threshold, the result says so and asks for review rather than presenting itself as settled.

A3

No fabrication

Numbers come from a computation, quotes come from the transcript. Where there is nothing, the answer is nothing.

A4

PII protection

Personal data is redacted by default, on the way in, and the redaction is part of the signed record.

A5

Recommendations, not verdicts

The platform reports what the data indicates. The decision stays with the person who has to defend it.

A6

No intent attribution

What was said and what happened. Never what somebody was trying to do.

A7

AI disclosure

Every AI-produced line is marked as AI-produced, in the metadata and on the surface a person reads.

A8

Refuse harmful requests

Blacklists, discriminatory criteria, routes around human review. The platform declines and says why.

[05 / IN REVEAL]

See how REVEAL implements it.

The Governance page describes the controls available today, the evidence behind results, the confidence gate, human review and the record, and marks what is still in development.

Human Oversight Let’s make
quality visible.