Due diligence
Most ESG diligence dies at signing.This one becomes year one.
The usual pre-investment ESG review is a deck, authored in Office, read once, and filed. It ends at the signature — the budget spent, the data stranded. This one is a platform record: it grades the target against a sector framework from answers the company already gave, and the day you close, that grid becomes the company's entry baseline and the action plan becomes its ESG roadmap.
What it does
A framework per sector
A 32-topic grid for the kind of business you're buying — how exposed the sector is on each, what documents to expect, which questions to read. Matched to the target by its NACE code, authored once, reused on every deal.
Reads what you already have
The target's own questionnaire and its documents — the data room, files pinned to answers — are read automatically. No fresh data pull, no new collection exercise imposed on a company you don't own yet.
Grades, it doesn't guess
A topic with no evidence and no answer reads as "information not available." That's written in code, not asked of a model — a thin data room shows as thin rather than dressed up as a completed assessment.
The rating is arithmetic
Low, medium or high is computed from the share of material topics that fall short — not an opinion a model can be talked into. Bigger frameworks aren't penalised for being thorough, and an unresolved red flag can't sit under a "low."
Trajectory, not a snapshot
Every reporting year the target has is read, not just the latest, so the memo argues movement — "the accident rate fell from 26 to 12 over three years" reads differently from a single number.
It carries into the hold
On closing, the graded grid becomes the company's entry baseline and the action plan its ESG roadmap. The diligence stops being a document and becomes the first year of monitoring — the same platform, no re-keying.
What it reads, and what each part is for
| Source | What it contributes |
|---|---|
| The linked questionnaire | The graded position — this year's answers, judged against the framework |
| Earlier campaigns | The trajectory — every prior year of the same company, read as trend |
| The data room | Deal-room evidence a questionnaire never holds — a litigation register, a supplier analysis |
| Files pinned to answers | Pre-classified evidence — the question they hang off already names the topic |
Only answers from the dataset the framework was written for are used. The same question code means different things in different question sets, so rather than mis-read one, it reads none and says why.
How a review runs
- 01 · Setup
Point at the target
Create the company, link one of its questionnaires. The framework auto-matches from the NACE code; the earlier years come in on their own.
- 02 · Read
Gather the evidence
Documents are found and read automatically — from the data room, the questionnaire, and files attached to individual answers. Nothing to upload.
- 03 · Grade
Judge the grid
Each of the topics is graded from the answers, the earlier years and the documents — and left as "information not available" wherever the evidence isn't there.
- 04 · Output
Memo, report, handover
A short IC memo for the committee and the full report behind it. Proceed, and the grid becomes the company's entry baseline for the hold.
What one topic actually rests on
| Topic | Maturity | Rests on |
|---|---|---|
| Health & safety | Managed | Frequency rate, three years, plus the certificate |
| Climate targets | In progress | Carbon footprint on file, no SBTi validation |
| Board diversity | Not addressed | Zero women on a five-person board |
| Biodiversity | Information not available | No answer, no document — stated, not inferred |
| Supply-chain risk | In progress | Supplier concentration read from the data room |
The row that matters most is the fourth. Where the evidence isn't there, the grid says so plainly rather than grading confidently on nothing — the single failure mode that makes an automated review untrustworthy.
The regulation behind this
Written for someone who has to comply with it, not to rank for its name.
Where this differs
Most diligence is a one-off deliverable that ends at signing.
The graded grid becomes the company's entry baseline at close, so the review is year one of monitoring rather than a document filed and forgotten.
Most tools ask the target for a fresh data pull.
It reads the answers the company already gave and every prior year alongside them — the review is assembled from data that already exists, not a new burden on a company you don't own yet.
Most AI tools grade confidently on thin data.
No evidence reads as "information not available," and the rating is arithmetic rather than a model's opinion — the two places an automated review usually earns its distrust are closed in code.
“We already run ESG diligence — a consultant sends us a deck.”
And the deck is read once and filed, its data stranded in a slide the day after signing. The work is the same; where it lives is not. Here the grid is a platform record, so the thing you paid a consultant to assemble becomes the first year of the monitoring you'd have started from scratch anyway.
Bring us a live target.
Thirty minutes on a deal you're looking at. We'll grade it against its sector framework from the data it already reports and show you the memo the committee would read — including the topics where the honest answer is "we don't know yet."