Explainable origination intelligence for restructuring
A UK insolvency and restructuring firm
An early-stage intelligence engine that reads public company filings to surface restructuring signals, scores them on a transparent model, and drafts cited internal reports — with human approval required before any contact.
8-factor
transparent scoring — no black box
Every quote
verified against source, or it fails
Human approval
required before any external contact
The challenge
The work this system had to absorb.
Origination in restructuring means reading a lot of public filings carefully and defensibly. The firm wanted to surface genuine signals quickly without an AI making unsupported claims or reaching out to anyone on its own.
What we built
A system in production, not a slide deck.
A pipeline that scans UK companies from official public data and enriches them with officers, charges, filings and notices.
A deterministic reading of filed accounts to surface signals such as overdrawn director's-loan exposure.
A transparent eight-factor scoring model and a fully-cited internal report, routed into a workflow that requires human approval before contact.
Governance, in the code
The controls are written into the architecture.
This is the part most firms can’t show you. Governance here isn’t a policy document. It’s constraints the system enforces on itself, every time it runs.
A denied-vocabulary blocklist is injected into every prompt and re-scanned on output; a violating result is rejected, not published.
An anti-hallucination check requires every quoted figure to be a literal substring of its source, or the extraction fails.
Nothing leaves the system without a recorded human approval — who approved it, and when.
Temperature-zero calls and eight published weights that must sum to one (the code refuses to run otherwise) keep scoring explainable.
Outcomes
What it has done.
Honest and labelled: dated where the figures are point-in-time, and described as controlled tests or early-stage where that is the truth.
A working live demo over real public filings, with a fully-cited internal report and a vocabulary-locked draft letter held for human approval.
Reproducible by design: persisted prompt versions, raw responses and a full source audit trail.
How it’s engineered
The approach under the hood.
An origination engine built on official public data sources, with deterministic, temperature-zero model calls, schema-constrained extraction and an explicit, auditable scoring model.
Keep reading
More systems we’ve built and governed.
Find out where your AI exposure sits.
We'll tell you plainly what's worth doing, what isn't, and what a board or regulator will expect to see. No pitch deck.
No obligation · no pitch.