Technology
Applied AI built for work that has to stand up to scrutiny
Energy, carbon and ESG outputs are read by auditors, lenders, regulators and boards. That constraint shapes how we build: retrieval over invention, stated assumptions, and a human accountable for the result.
Architecture
Four layers, one accountable output
A deliberately conventional architecture. The value is in the domain modelling and the review discipline, not in novel model research.
- 01Available now
Domain data layer
Building, energy, carbon and ESG data structured into consistent internal models so analysis is comparable across assets and reporting periods.
- 02Available now
Evidence retrieval
Client documents are indexed so the system answers from the material actually supplied, with references back to the source rather than from general knowledge.
- 03Available now
Reasoning and drafting
Large language models are used for interpretation and drafting, constrained by retrieved evidence and required to state assumptions and gaps.
- 04Available now
Review and output
Every output is a draft routed to a named human for challenge and sign-off. Nothing is published or certified automatically.
Engineering principles
What we will and will not automate
Evidence before inference
If the evidence is not there, the system reports a gap. It does not fill the gap with plausible language.
Assumptions are visible
Where a calculation depends on an assumption, that assumption is stated alongside the result so a reviewer can accept or replace it.
Deterministic where it matters
All numerical output is produced by deterministic calculation engines. Figures are never generated by a language model.
Continuous regulatory watch
Compliance methodology and guidance are maintained continuously against UK and EU regulatory change, so an output reflects the current position.
Human accountability by design
Every issued report is reviewed and signed off by a qualified person — an accredited energy assessor for EPCs, an ESOS Lead Assessor for ESOS. The platform speeds report production; it does not replace the professional.
Integration
How partners connect to it
Hosted products
Use NZC AI or ESG AI directly through their assistant interfaces at nzc-ai.co.uk and esg-ai.co.uk.
MCP layer licensing
The intelligence layer — in-product assistance, document ingestion, validation and calculation — licensed to a third-party platform inside its own workflow.
Embedded and white-label
Capability delivered inside a partner's product under commercial agreement.