THE PRODUCT

One platform. Every use case. Every framework.

Probari is a multi-tenant SaaS platform for AI assurance. Each tenant is a single jurisdiction with a curated framework stack, a six-stage assured chain — triage, wizard, decision map, AI Impact Assessment (AIIA), PRD generator and patrol — a control library mapped to evidence, and a live signals feed against the assurance baseline. Every stage is tamper-evident and linked to the one before, so the audit trail cannot be quietly altered. Below is what the product does today, what is on the roadmap, and how it is built.

Book a 30-minute walkthrough See the surfaces

FOR HEADS OF AI · CHIEF RISK OFFICERS · DPOs · COMPLIANCE

THE FOUR PILLARS, EXPANDED

What Probari does, end to end.

01

DESIGN

Decide what to build before you build it. The triage agent scores every candidate against the Probari prioritisation matrix — five weighted dimensions, jurisdiction-specific regulatory rubric — and answers "should we build it?". The intake wizard then captures purpose, lawful basis, data classes, decision affected, vendor model and owner, with risk tier automatically graded against the EU AI Act schedule.

  • Triage agent & prioritisation matrix
  • Intake wizard
  • Automatic risk grading (aligned to EU AI Act)
  • Stakeholder map
  • Use-case register

02

ASSESS

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 — with forty-seven questions across nine sections, drafted with Perplexity Sonar assistance, citation chain back to source documents. Then auto-generate a Product Requirements Document (PRD) — the build brief engineering works to — patrolled by a second AI agent running six structured checks before any human reviews it.

  • AIIA template
  • Perplexity Sonar-assisted drafting
  • Citation chain
  • Sign-off log

03

ASSURE

Translate assessment into operational controls. Every framework in the tenant's stack is mapped to controls; every control is tied to evidence. The evidence pack exports as a single bundle for board, regulator, audit or acquirer.

  • Control library
  • Framework crosswalk
  • Evidence pack export
  • Audit-ready trail

04

OPERATE

Monitor live AI behaviour against the assurance baseline. Drift in input distribution, fairness metrics or customer-outcome indicators opens a reassessment ticket — not a surprise audit.

  • Signals feed
  • Drift detection
  • Reassessment queue
  • Outcome monitoring

THE SURFACES

Six screens. One source of truth.

Every Probari tenant gives you the same six core surfaces. The framework stack and conduct overlays differ by profile; the surfaces are the same — so the second tenant inside a group comes online faster than the first, and an evaluator can compare across deployments without retraining.

SURFACE 01

Triage & intake

Triage scores a candidate against five weighted dimensions and returns a quadrant placement with regulator-anchored rationale. The intake wizard then captures the use case, automatically grades the risk tier, attaches the right framework obligations, assigns an owner and creates the decision-map placeholder in one flow.

  • Purpose, lawful basis, data classes
  • EU AI Act risk tier
  • Vendor and model passthrough

SURFACE 02

AIIA, PRD generator & patrol

The forty-seven-question AI Impact Assessment (AIIA), linked to the upstream decision map. Once approved, the PRD generator drafts a Product Requirements Document (PRD) — the build brief engineering works to — using Perplexity Sonar. A second AI agent (the patrol agent) then runs six structured checks (source anchoring, regulator citation, risk-tier consistency, jurisdiction, approval chain, PM-craft) before any human reviews the work. Up to three auto-remediation attempts before a human is paged.

  • Nine sections, forty-seven questions
  • Citation chain
  • Sign-off log

SURFACE 03

Control library

Every control in the tenant's stack, with its source framework, its owner, its evidence and its current status. Filter by framework, by use case, by owner or by review date.

  • Source framework attribution
  • Evidence references
  • Owner and review cadence

SURFACE 04

Framework crosswalk

Side-by-side view of how a single use case maps across every framework in scope. Where one control answers two regulators, the crosswalk shows it; where coverage is thin, the gap is named.

  • Many-to-many control mapping
  • Coverage and gap view
  • Per-framework export

SURFACE 05

Signals feed

Operational signals against the assurance baseline. Input distribution, fairness, customer-outcome indicators and conduct flags. Drift opens a reassessment ticket; the audit trail captures why.

  • Input and output drift
  • Fairness and outcome indicators
  • Reassessment queue

SURFACE 06

Evidence pack export

A single bundle covering the full assurance trail for any use case or set of use cases — AIIA, controls, evidence references, sign-off log, signals history. Ready for board, regulator, audit or acquirer.

  • One-click export per use case or tenant
  • Per-jurisdiction bundles for cross-border
  • Versioned and timestamped

FRAMEWORK LIBRARY · MAY 2026

Fourteen frameworks. Six regions. Five tenant profiles.

Frameworks are curated per profile. A GB Energy supplier tenant ships with a different stack from an All-Ireland tenant or a GB Brokerage tenant. Frameworks are versioned; when Ofgem updates its Standards of Conduct or the EU finalises an AI Act delegated act, the update lands in every customer tenant on the same release cadence.

GB ENERGY

9 FRAMEWORKS

UK GDPR and DPA 2018, EU AI Act, ISO 42001, Ofgem Consumer Standards (domestic and non-domestic), NESO Grid Code, NIS2, Equality Act 2010 and the Consumer Duty crosswalk where in scope.

GB BROKERAGE

7 FRAMEWORKS

UK GDPR and DPA 2018, Ofgem Third Party Intermediary (TPI) Authorisation, Standards of Conduct (non-domestic), Broker Fee Transparency, QDSS — the Qualifying Dispute Settlement Scheme — Redress Membership, Ofgem CHS Non-Domestic, UIA Code of Practice.

RoI ENERGY

8 FRAMEWORKS

EU GDPR, EU AI Act, ISO 42001, CRU retail and conduct standards, SEM Committee wholesale frameworks, NIS2, DORA where in scope, Equality Act equivalents.

NI ENERGY

8 FRAMEWORKS

UK GDPR + DPA 2018, EU AI Act extraterritorial, ISO 42001, Utility Regulator retail and conduct standards, SEM Committee wholesale frameworks, NIS2, NI Equality, Consumer Duty crosswalk.

ALL-IRELAND

11 FRAMEWORKS

Union of the RoI and NI stacks plus the SEM and EU layers in a single tenant. Use cases are tagged by jurisdiction; controls roll up across the group; evidence pack exports per jurisdiction or all-island.

SHARED CORE

PLATFORM

Every tenant sits on the same platform: the framework graph, the AIIA engine, Perplexity Sonar-assisted drafting, the control library, the signals feed and the evidence pack export. Vertical packs are configuration, not new code.

ARCHITECTURE

Multi-tenant, multi-jurisdiction, single source of truth.

Probari is a multi-tenant SaaS platform. Each tenant maps to a single jurisdiction or a single all-island grouping. Tenants are isolated at the data layer; the framework graph, the AIIA engine and the control library are shared at the platform layer.

The platform is built on a typed schema, exposes a documented API, and is operated under the same evidence discipline it ships to customers. Perplexity Sonar is the only permitted AI provider — used for drafting and citation, never for autonomous decisioning. Every interaction with the model is logged with prompt, response, citation and reviewer.

PLATFORM

Multi-tenant SaaS

Tenant-isolated data, shared framework and assessment engines. Super-admin role for cross-tenant operators (consultancies, group functions). Per-tenant audit log.

ASSURANCE

Evidence chain

Every answer, every control, every signal carries an author, a citation, a timestamp and a version. The evidence pack is reproducible from the chain at any point in time.

AI

Sonar-only, supervised

Perplexity Sonar is the only permitted AI provider, chosen because every answer comes with the source. Used for drafting and citation only. Every response is reviewer-assigned. The platform never auto-signs evidence.

INTEGRATION

API-first

Every surface is also an API. Customers extract their own evidence pack on their own schedule. The platform is built to be the system of record, not a walled garden.

WHAT IS NEXT

Three near-term moves.

The roadmap below is concrete and near-term. The items here are the ones you will see in the next four quarters.

Q3 2026

PRINCIPLES REFRESH

Refresh the GB Brokerage pack against Ofgem's formal authorisation principles once the H1 2026 market survey closes and the principles publish. Customers receive the update without a tenant rebuild.

Q4 2026

EVIDENCE EXCHANGE

One-click evidence pack delivery to a counterparty — supplier to broker, broker to supplier, group to acquirer. The exchange is versioned and revocable. Pilot customers get it first.

2027

MULTI-DOMAIN ASSURANCE

The control library extends from AI assurance into data protection, cyber and conduct on the same platform. AI assurance becomes a filter on a broader assurance product, not a standalone tool.

Walk through the live product.

Probari is founded and designed by Ian Curran. If you would like a 30-minute walkthrough of the live MeridianTPI or Northbrae reference deployment, get in touch — we will book the call within five working days.

Request a walkthrough

Or write directly to info@probari.uk