AI ASSURANCE — UK & EU
Probari is the assurance-led design platform for AI. It guides every use case from impact assessment through to live monitoring — so your teams ship faster and your regulators see the evidence at every step.
FOR HEADS OF AI · CHIEF RISK OFFICERS · BOARDS
THE PROBLEM
Typical AI governance tools assess models that already exist, after teams have built them. By that point, redesign is expensive and assurance feels adversarial.
Probari inverts this: assurance is woven into the design phase, so the evidence is generated as the use case is built, not retrofitted.
THE PROBARI METHOD
01
Decide what to build before you build it. Triage every candidate against the rules in GB, Ireland and Northern Ireland at the same time, then capture the use case with assurance built in from the start.
02
Run a formal AI Impact Assessment (AIIA) — the document that explains what an AI agent does, why it can be trusted, and where it is constrained — then auto-generate a Product Requirements Document (PRD), the build brief engineering works to, from the same evidence chain.
03
Translate the assessment into operational controls. Map to UK GDPR, EU AI Act, ISO 42001 — your frameworks, surfaced in one place, gaps named.
04
Monitor live AI behaviour against the assurance baseline. Drift triggers re-assessment, not surprise audits.
WHY PROBARI
FOR HEADS OF AI
Ship faster, prove it.
Probari builds the assurance evidence as you build the system. Governance review stops blocking launches.
FOR CHIEF RISK OFFICERS
Evidence on demand.
Every use case carries a live AI Impact Assessment, a control map and a signed sign-off chain. Auditable in minutes, not months.
FOR BOARDS
One register, all of AI.
A single source of truth for every AI use case across the organisation, with risk tier, owner, and assurance status visible at a glance.
WHERE PROBARI IS DEPLOYED
Probari covers the energy retailers that build and operate AI systems, the Third Party Intermediaries (TPIs) — energy brokers and consultants — who advise on them, and the debt collection agencies (DCAs) that act on the supplier's behalf. The platform is configured for GB and all-island Ireland coverage in the same product.
VERTICAL 01
Heads of AI, Chief Risk Officers
For energy retailers operating AI systems under Ofgem, the Utility Regulator, the CRU and the SEM Committee. Probari maps every use case to UK GDPR, sectoral consumer standards and the EU AI Act, with assurance evidence built in alongside the system.
VERTICAL 02
Compliance Officers, Commercial Directors
Ofgem will become the TPI regulator under a hybrid authorisation regime confirmed in October 2025, with a phase-in window of 12–18 months once legislation lands. Probari builds the registration evidence, conduct logs and complaints record before authorisation becomes mandatory.
VERTICAL 03
Compliance Directors, Operations Directors
Ofgem's 2024 Debt Strategy and 2025 update put AI in arrears — ability-to-pay scoring, vulnerability inference, PPM-pathway prioritisation, customer comms — firmly inside the assurance perimeter. SLC 27.8A (the licence condition that flows a supplier's obligations down to a debt collection agency (DCA)) applies directly. Probari Collections is sold to DCAs: one tenant covers every supplier on your book, with an Evidence Bundle export for any supplier on request.
ADD-ON · HORIZON
Heads of Compliance, Regulatory Affairs, Policy Leads
Probari Horizon turns a regulatory consultation into a structured Impact Memo in a single run. Cross-regulation map, named business areas affected, named customer groups affected, the questions your analyst must answer to land a position, and the opportunities your competitors will miss. £6,000 standalone or £4,000 as an add-on; included with Assured Operate.
REFERENCE DEPLOYMENT · NORTHBRAE ENERGY
Northbrae Energy is a GB Energy retailer running the full Probari assured chain on a customer-churn use case. The deployment is the reference configuration shown to evaluators — a complete, regulator-grade trace from triage through live signals. Six stages, each tamper-evident and linked to the next so the audit trail cannot be quietly altered. The methodology is reproducible; the outputs below describe what the platform does, not a marketing claim.
STEP 01
Candidate scored against the Probari prioritisation matrix — five weighted dimensions, jurisdiction-specific regulatory rubric. The agent answers "should we build it?" with a quadrant placement and an audit-anchored rationale.
OUTPUT · triage finding, quadrant, audit anchor
STEP 02
Use case captured through the Probari intake wizard — purpose, data classes, lawful basis, decision affected, vendor model. Risk tier automatically graded against the EU AI Act schedule, seeded by the triage outcome.
OUTPUT · signed intake record, risk tier, owner
STEP 03
Every decision follows a five-step pattern — Discover, Identify, Assess, Mitigate, Confirm (DIAMC). It means nothing reaches a regulator without being checked the same way every time. The completed decision map produces the structured evidence the AI Impact Assessment consumes. Every stage entry is regulator-cited and reviewable.
OUTPUT · approved decision map, evidence packets
STEP 04
Forty-seven AI Impact Assessment (AIIA) questions across nine sections, drafted with Perplexity Sonar assistance — the only AI provider Probari uses, chosen because every answer comes with the source — and linked to the upstream decision map. Every answer carries a citation, an author and a timestamp.
OUTPUT · completed AIIA, citation chain, sign-off log
STEP 05
The approved AIIA auto-generates a Product Requirements Document (PRD) — the build brief engineering works to — using Sonar. A second AI agent (the patrol agent) then runs six structured checks — source anchoring, regulator citation, risk-tier consistency, jurisdiction, approval chain and PM-craft — before any human reviews the work.
OUTPUT · patrolled PRD, six-check report, tamper-evident audit chain
STEP 06
Operational signals tracked against the assurance baseline. Drift in input distribution or fairness metrics opens a reassessment ticket rather than a surprise audit. The full chain — triage to signals — replays as an evidence pack.
OUTPUT · signals feed, drift triggers, reassessment queue
Reference configuration in the Probari demo environment. The Northbrae deployment is illustrative — every output above is generated by the live product against real framework text. Six agents in the chain: triage, wizard, decision map, AIIA, PRD generator and PRD patrol.
Probari is founded and designed by Ian Curran. If you are a Head of AI, Chief Risk Officer, or board director with an AI assurance problem, get in touch — we will book a 30-minute briefing within five working days.
Request a briefingOr write directly to info@probari.uk