Accounting & Audit by Design (A&AD) Framework (Technical)

This blog post is from a very good email sent by Richard Gasca, the creator of Accounting & Audit by Design (A&AD) Framework, to a private email list of accountants interested in A&AD. I thought it would be a good thing to share more broadly.

Auditchain's demonstration of Pacioli Ledger Alpha and the Integrity Audit Log, reinforced by yesterday's announcement of Auditchain's partnership with PwC, represents a formidable and welcome technical milestone.

Anchoring event sequences via the Haber-Stornetta cryptographic lineage (1991, 1993, 1997), utilizing Ethereum EIP-712 structured signatures, and batch-stamping Merkle roots onto public blockchains definitively solves post-hoc ledger tampering. Auditchain is entirely right on a fundamental premise: the substrate must come first, and establishing physical, append-only non-repudiation is the foundational prerequisite for any modern financial platform.

To bridge the chasm between raw cryptographic bits and a legally sufficient Audit Evidence Graph, we must apply the foundational test established in ISO/TC 307 WG 10 (Forensics and Evidence), drawing upon the work of Eric Cohen: (here is the presentation)

"Evidence of what, for whom, and for what purpose?"

A ledger entry may be cryptographically tamper-proof, yet legally and economically mute: it does not verify who held the key, statutory authority, the underlying business event, or compliant accounting recognition.

Immutability guarantees non-tampering post-facto; only semantic governance guarantees audit sufficiency at inception.

1. The Fallacy of "Orthogonality": Why Writing Cannot Be Severed from Semantics

In computer security circles, it has often been posited that authentication (proving which private key signed a payload) and authorization (determining whether that actor possessed the statutory capacity to bind the enterprise) are "orthogonal"; problems that systems can handle in completely decoupled phases.

In cybersecurity, perhaps. In accounting, auditing, and corporate governance, never.

Auditchain noted in his latest intervention that at the "writing stage," the ledger is concerned with recording bits rather than formats or XBRL, and that "the semantics are a separate layer." But herein lies the architectural fork in the road:

When an auditor examines an entry, mathematical integrity alone cannot establish audit sufficiency. If a cryptographically signed transaction was executed by an actor exceeding their board-delegated authority, if it lacked real-world economic duality, or if it violated IFRS/US-GAAP recognition criteria, the blockchain faithfully preserves the record; yet the audit fails.

Without semantic governance at inception, distributed ledgers merely guarantee:

"Garbage In, Impeccably Preserved and Immutable Garbage Out."

Cryptographic immutability proves that nothing was altered after the fact. It does not prove that what was recorded was true, authorized, or economically valid at the moment of inception.

2. Reframing the Metaphor: The Auditor as Public Health Inspector

Auditchain shared an evocative metaphor regarding large advisory firms: that busy accountants with billable quotas care less about "the viscosity of the molecules inside the udder of the cow" and more about "the milk itself" and the certificate attesting that the dairy complied with standards.

This observation is spot on for the consumer or preparer of a report. But it reveals why contemporary audit is in crisis:

The financial controller drinks the milk. They need the trial balance to close the period on time.

The Auditor is the Public Health Inspector. Under International Standards on Auditing (ISA 500) and PCAOB AS 1105, an auditor cannot take the milk on faith merely because a certificate is attached. The statutory obligation of the auditor is precisely to inspect the health of the herd, the hygiene of the milking process, the pasteurization temperature, and the integrity of the cold chain.

The 40% audit failure rate that occurs precisely because auditors have spent decades inspecting only the glass of milk at year-end, completely blind to how the cow was managed throughout the year. Continuous assurance cannot skip the operational reality to only inspect the final output.

3. The Accounting & Audit by Design (A&AD) Solution: The 3-Layer Assurance Stack

Within the Accounting & Audit by Design (A&AD) Framework, we address this dilemma not by replacing Auditchain's cryptographic substrate, but by completing it through a 3-Layer Assurance Stack powered by W3C semantic graph standards:

THE 3-LAYER ASSURANCE STACK

LAYER 3: DETERMINISTIC GOVERNANCE & AUDIT ASSERTIONS (Gatekeeper at Moment 0)

W3C SHACL 1.2 Validation Shapes (Fails closed before polluting the immutable ledger

Statutory Authority & Delegation Limits bound to Cryptographic Signatures

ACTUS Algorithmic Financial Contracts (Deterministic Cash Flows - Willi Brammertz)

OCEG GRC Triad: Risk -> Control -> Control-Activity

LAYER 2: MULTIDIMENSIONAL SEMANTIC EVIDENCE GRAPH (Causal Lineage & Reality)

W3C PROV-O (Bitemporal Causal Lineage: Entity -> Activity -> Agent)

Economic Duality: ISO 15944-4 (Accounting and Economic Ontology) REA Ontology - Bill McCarthy & Cheryl Dunn)

Canonical Ingestion: ISO 21378 (ADCS ERP Tables) & XBRL GL (1:1 XSD Parity) & UBL e-Invoices

Graph Engine: DFRNT / TerminusDB (Bitemporal Valid Time vs. Transaction Time)

LAYER 1: CRYPTOGRAPHIC SUBSTRATE & TAMPER EVIDENCE (Physical Bedrock)

Haber-Stornetta Hash Chains (1991, 1993, 1997) & Merkle Trees (Pacioli Ledger - Jason Meyers)

Ethereum EIP-712 Typed Structured Signing

Public Blockchain Batch Anchoring & Non-Repudiation

Pillar A: W3C PROV-O — Deep Provenance Beyond the Hash Chain (Layer 2)

A cryptographic hash tells you that an entry has not changed; it cannot tell you why it happened or how it connects to supporting evidence. By adopting W3C PROV-O within a bitemporal Knowledge Graph (DFRNT):

Every journal entry (`prov:Entity`) is explicitly bound to the economic activity (`prov:Activity`), the originating statutory source document (UBL invoice, digital contract), and the executing legal actor (`prov:Agent`).

We maintain strict Bitemporality: distinguishing when the real-world economic event occurred (Valid Time) from when the system recorded and asserted it (Transaction Time).

We standardize ERP and transactional ingestion via ISO 21378:2019 (Audit Data Collection Standard - ADCS) and XBRL GL (with 1:1 XSD parity), mapping core cycles (GL, P2P, O2C, Inventory) directly into graph nodes.

We implement Bill McCarthy's REA (ISO 15944-4) ontology: every increment is semantically bound to its reciprocal decrement (economic duality), transforming opaque token transfers into auditable economic exchanges.

Pillar B: W3C SHACL — Deterministic Invariant Enforcement at Moment 0 (Layer 3)

Rather than waiting for an auditor to discover discrepancies during a retroactive "4K Replay" months later, A&AD utilizes W3C SHACL (Shapes Constraint Language) as an active gatekeeper at Moment 0:

a. Unifying Authentication with Authorization:

The signing key is mapped to an ontological role in the graph. If an EIP-712 transaction exceeds the signer’s statutory spending limit or violates segregation-of-duties rules, the SHACL shape fails closed, rejecting the transaction before it can pollute the immutable chain.

b.. Contractual Computability with ACTUS:

As Willi Brammertz has established, financial instruments are deterministic state machines. A SHACL shape verifies whether an interest accrual or repayment conforms mathematically to the ACTUS contract terms prior to posting.

c. Deterministic Quality Management & ISO 21378 Compliance:

Deterministic Quality Management is an approach to quality control and assurance that relies on fixed rules, repeatable logic, and exact measurements where identical inputs consistently produce identical outputs. Unlike probabilistic methods; which account for statistical likelihoods, estimations, or human visual judgment; deterministic quality management uses invariant controls to guarantee compliance, safety, and traceability

SHACL enforces accounting invariants (debit/credit equality, XBRL GL canonical conformance, OCEG control activities) and validates ISO 21378 Level 1 requirements (mandatory source document links, exact decimal precision, and segregation of duties such as `createdUserID != approvedUserID`) deterministically before commit.

4. Beyond "Homogenizing Semantics": The Power of Knowledge Graphs

Auditchain rightly warned that attempts by legacy vendors to artificially "homogenize" semantics have caused endless frustration for CFOs and controllers. We agree completely: forcing disparate business realities into rigid, flat relational schemas is a recipe for failure.

This is precisely why Knowledge Graphs (DFRNT / RDF / JSON-LD) represent the paradigm shift:

Graphs do not flatten or homogenize data destructively. They allow local, context-rich business models to retain their native fidelity while mapping to universal taxonomies / ontologies (XBRL GL, REA, ACTUS) via flexible semantic relationships.

5. The Path Forward: A Symbiotic Architecture

Auditchain's Pacioli Ledger and the A&AD framework are not competing narratives; they are complementary layers of the exact same revolution:

[ Pacioli (Layer 1: Cryptography & Immutability Bedrock) ]

(Haber-Stornetta / Merkle Trees / EIP-712 / Blockchain)

[ A&AD (Layers 2 & 3: Semantics, Causal Lineage & Normative Rules) ]

(XBRL GL / W3C PROV-O / W3C SHACL / REA / GRC)

AUDIT EVIDENCE GRAPH

Pacioli provides the irreproachable cryptographic bedrock: preventing backdating, data deletion, and ledger tampering.

A&AD provides the semantic and normative intelligence: guaranteeing that every recorded event is authorized, compliant, economically grounded, and provable against auditing standards. (For more information, see the paper, The End of Reactive Control: Algorithmic Certainty through Accounting and Audit by Design)

This is how we shift the Kuhn Cycle. We replace 800 disconnected spreadsheets not with bare cryptography alone, nor with abstract ontologies alone, but with a fully computable, semantically bound, and cryptographically anchored Audit Evidence Graph.

This is the exact intersection where algorithmic contracts, cryptographic immutability, and semantic graph auditing converge into an industrial-strength reality. (Infographic)

Framework Workflow Steps:

Documentation: Documentation for a business event is received. Documentation could be in the form of paper, electronic paper such as PDF structured as a document but unstructured in terms of meaning, a structured machine interpretable format such as XML, or a standard structured document using the Universal Business Language (UBL). Humans assisted by artificial intelligence convert documentation into business event journal entries, rich is metadata.
Business Events Journal: Information from the documentation and data and metadata which describes the information is put into a business events journal. Metadata from Data Centric Accounting (DCA) and a reporting framework are used to encode the entry into the business events journal. (prototype)
Financial Journal: From the business events journal information projections are generated to create financial journals such as subsidiary journals and general journal.
Enterprise Information Platform: From the rich set of journals, financial ledgers are projected; think accounting manifolds. This are rich traceable/trackable sets of information with connections all the way back to the business events journal and original documentation which spawned the business event. This is similar to but better than a financial data platform. A better explanation for the enterprise information platform is the Semantic Knowledge Management System or the Pragmatic Knowledge Management System, see the Pragmatic Web.

Proof of Concepts and Working Prototypes:

Business Events: The business events journal is a graph. The business events ledger is an accounting manifold.
General Journal: The general journal is a projection. Prototype2: JSONLD / Schema
Trial Balance: The trial balance is a projection.
Lead Schedules: The lead schedules are a projection.
Closing Book: The closing book is a projection.
Financial Statement: The financial statement is a projection. (Note the verification; full set of information)
Analysis Models (FP&A): Analysis models are a projection; entity variance analysis, peer benchmarking, budget/actual comparisons, etc.
Audit: Risks, Controls, Control Activities, Audit Evidence; all graphs. See Essence of Audit and Assurance (prototypes coming).

Additional Information:

Search This Blog

Seattle Method