A different architecture for verifying AI

We are delighted to share that LogicalMetricsAI has been named as one of the Tech Nation Breakout 50 for 2026, an annual showcase of the UK's most promising early-stage technology companies, selected from the shortlists of Tech Nation's NextWave and OneToWin pitch competitions.

The Breakout 50 cohort represents over £27.5 million raised between 50 companies, with more than 300 employees across every region of the UK. We are proud to be featured in the AI and Deep Tech section, alongside companies building some of the most interesting technology in the UK right now.

What we are building

LogicalMetricsAI is developing AARF, the Auditable AI Reasoning Framework. As AI is deployed at scale into healthcare, finance, legal, and public services, organisations face a critical and largely unsolved problem: they cannot verify whether an AI system's reasoning is correct, evidence-grounded, or safe.

AARF addresses this directly. It breaks any AI output into individual claims, checks each one against a library of authoritative, version-controlled sources, applies deterministic safety rules that cannot be overridden by probabilistic model outputs, and produces a full audit trail before a decision reaches a human.

Where we come from

LogicalMetricsAI is headquartered in Durham. We are one of a small number of North East companies in this year's Breakout 50 cohort, and we think that matters. Genuinely differentiated technology is being built outside London, and the North East ecosystem, from Business Durham to the regional investment community, is an increasingly important part of the UK's innovation landscape.

The founding insight behind AARF came from original university research into misinformation detection. The core realisation was that AI systems hallucinate because they cannot cleanly separate true from false during inference.

The solution was counterintuitive. By training reasoning models exclusively on false content and maintaining truth as a separate verified library, hallucinations become structurally contained. A hallucination of false reasoning is simply more false and cannot corrupt the truth library. This architectural approach underpins everything we build.

What comes next

We are currently in active conversations with investors and potential PoC partners across healthcare and financial services. SEIS Advance Assurance is in progress. The system has been built and validated in controlled synthetic environments and is now moving toward real-world pilot deployments.

If you are working in NHS digital health, AI governance, regulated financial services, or early-stage investment, we would be pleased to speak.

Contact: andrew@logicalmetricsai.com

Read the full Tech Nation Breakout 50 report here