Research on the constraints that decide market access.

Published analysis on the rules, trade flows and physical conditions that determine whether a product can enter a market, what it costs, and what changes next. Articles, reports, and data, all freely available.

Cover of Regulatory Intelligence Beyond Retrieval, a product field guide
E-book · PDF edition · September 2026 · 55 pages

Regulatory Intelligence Beyond Retrieval

A product field guide

A field guide for product leaders and decision-makers evaluating whether an AI system's output can be relied on, whether to buy or build, and what to fund first in either case.

Buy the ebook · €15.99 (excl. VAT)

You receive the full PDF by email immediately after purchase. Payment and VAT are handled by Gumroad. No account needed.

Product leaders deciding whether to build a decision system or buy one.

Regulatory, compliance and risk teams asked to rely on AI output.

Consultancies and advisory firms choosing the layer they will put their name on.

01

A fourteen-test evaluation you can run on any vendor or internal build.

Each test is phrased so that you choose the case, the date and the ground truth, so a prepared demo cannot pass it.

02

The difference between a computed verdict and a composed one, and why it matters the day the law changes.

Same evidence, same rules, same answer, with the path from source to conclusion inspectable at every step.

03

A clear view of which half of the system is cheap and which half is the business.

Ingestion and drafting get cheaper every quarter; classification, review and maintenance do not. Fund the second before the first.

Intro.
The argument of this book
The case for computing verdicts rather than composing them, and who this book is written for
01
Verdicts are computed, not composed
Why retrieval plus generation cannot give you the same answer twice, and what does
02
An ontology that holds
Classifying constraints by what they do, not where they came from
03
A graph that is maintained
Names, time, relationships and propagation: the four costs of keeping a decision current
04
A harness you own
Capability is rented; the controls around it are the asset
05
Provenance and stated limits
Every claim with its source, its check date, and what was not checked
06
Structure that is paid for
The labelling operation, the review desk, and what this category actually costs
07
Observed outcomes, not just documents
What happened when the analysis met the world
08
Build for regulatory motion
The EU, India and the US in one year of change, and what a system must do about it
Concl.
How the 8 components generate value
Which parts of the system appreciate as capability gets cheaper, and which do not
Checklist
The Evaluator Checklist
Fourteen tests to run on your case, your date and your ground truth

Chapter 5, Provenance and stated limits, is sent free by email as a PDF. It is the chapter to send to anyone who has to sign off on an AI-generated document.

Anirudh Valluri

Anirudh Valluri is the founder of Jungle Systems, a risk modelling firm for market entry. He built and operates the system the book describes: a maintained graph of regulatory and market constraints, deterministic applicability rules, and a signed review desk, delivered today on the EU and India corridors. His Systems Impact Framework, the classification method in chapter 2, is published as a working paper on SSRN. The book is written from operating experience, not borrowed theory, and it says so where the experience runs out.

Regulatory Intelligence Beyond Retrieval

PDF edition · 55 pages · legal facts checked as at 13 September 2026, errata published on this page

Buy the ebook · €15.99 (excl. VAT)

Delivered by Gumroad. VAT included where applicable.

Every legal claim in the book carries a check date. Corrections since publication are listed here.

Errata for Regulatory Intelligence Beyond Retrieval
DateCorrection
No corrections have been issued since publication.