RGBY.ai

RGBY is an operational-intelligence platform for regulated sectors — it catches risk and drift early, and turns every call into an auditable, defensible decision, not a black-box guess.

Deterministic AI across housing, maritime, construction, financial services, energy, defence, and AI governance. Same inputs, same answer, every time — with a full evidence trail.

Request Briefing See how it works ↓

This is the engine, live. The same input vector always yields the same coherence score — so every decision is reproducible and auditable, not a black-box guess. Hover to inject noise and watch it re-resolve, then recover.

What you're looking at

The engine, in plain English

The panel above scores a case across four channels — Demand, Packing, Structure, Noise. Below is what each means for a housing case, and the one question they answer: are we sure enough to act — and could we defend that decision?

EvidenceDemand
how strongly the data points to a problem
AgreementPacking
whether the different signals tell the same story
GroundingStructure
how complete and reliable the records are
Noise
how much is missing, broken, or contradictory

Flat 14 · possible damp

  • Evidence — stronghumidity climbing for 3 weeks, 2 repair requests, a condensation flag at the last inspection
  • Agreement — strongthe sensor, the tenant, and the inspector all point the same way
  • Grounding — strongfull property and ventilation record on file
  • Noise — lowclean, consistent data
ACCEPT — act nowRaise a proactive damp case before a complaint lands, with the evidence pack already assembled — Awaab's Law compliance, by design.
coherence 0.94

Same home — the sensor's been faulty a week and two inspection notes now contradict each other

  • Evidence — still concerningthe picture still looks like damp
  • Agreement — weakthe two inspection notes disagree
  • Grounding — shakya gap where the record should be
  • Noise — highthe faulty sensor is feeding unreliable readings
HOLDToo much noise to act on defensibly. Replace the sensor and re-inspect — then it re-runs and commits.
coherence 0.60

Same facts in, same answer out — every time, and all of it logged. When Awaab's Law or the Ombudsman asks why you did (or didn't) act on this home, the answer is reproducible and defensible — not a black-box guess.

How it works

From raw signals to decisions you can defend

One engine behind every sector. RGBY runs on the R-KID protocol — a structured reasoning layer that sits between raw AI and the decisions your regulator will scrutinise.

01

Detect the signals that matter

RGBY reads the data your operation already produces — records, reports, sensor readings, model outputs — and surfaces emerging risk, inconsistency, and drift early.

e.g. across a social-housing stock, spotting the conditions for damp and mould before a complaint — not after.

02

Reason under hard constraints

The R-KID protocol applies your rules, standards, and domain knowledge — Constraints · Knowledge · Inference · Detection — to interpret what's happening. Deterministically: same inputs, same conclusion, every time.

No black-box variability — a result you can reproduce and explain.

03

Evidence every decision

Every output ships with its reasoning, the constraints applied, and a full audit trail — evidence-ready for regulators, boards, and inquiries. Deploy in your cloud or on-prem.

Defensible by design, not after the fact.

Not a black box. Where a general AI gives you a plausible answer you can't justify, RGBY gives you a consistent, traceable one you can defend.

Platform

The platform

A deterministic reasoning and monitoring layer for complex operational environments.

RGBY brings together signal detection, structured reasoning, drift monitoring, and auditable outputs to help teams interpret change, prioritise action, and maintain control in regulated settings.

Learn more about the platform

Scope

Signals of scope

4.4M

Social homes in scope

29

HHSRS hazard categories

12K+

High-rise buildings in scope

Deterministic

Auditability layer

Protocol

The R-KID Protocol

A governance layer for AI systems.

R-KID adds structured controls, consistency monitoring, and auditable decision traces to AI-enabled workflows operating in regulated or high-consequence settings.

Constraints

What the system must not do.

Knowledge

The rules, procedures, and structures it must follow.

Inference

How it interprets patterns, signals, and context.

Detection

How it monitors drift, inconsistency, and control failure over time.

AI Model
R-KID Protocol
Structured Output
Explore the R-KID Protocol

Product

DewPoint Guardian

A damp and mould intelligence system for risk, causation, mitigation, and compliance.

Not just detection.
Not just reporting.
A new layer for understanding what is happening, why it is happening, what needs to change, and how the response is evidenced.

Coming soon
RiskCausationMitigationCompliance
Learn more

People

The team

Gerald Manton

Gerald Manton

Strategic development, product direction, and commercial positioning.

John O'Leary

John O'Leary

20 years across social housing and major construction delivery. RGBY was developed from live failures in compliance, safety, and operational control.

Hadley Christoffels

Hadley Christoffels

Board member. Founder of dataDecisions.ai and author; data architecture and applied AI.

David Marriage

David Marriage

Board member. Enterprise transformation and AI in regulated financial services; former PwC partner.

Jon King

Jon King

Co-founder. AI safety, governance, and cognitive-risk systems; building provable assurance into the platform.

About RGBY

Connect

Talk to us

For pilots, partnerships, technical discussions, and sector use cases.

Request Briefing

Available for technical reviews, sector conversations, and early-stage deployment discussions.