Log in
AI & agents

What is Deterministic Ledger?

A deterministic ledger is a general ledger where the AI handles classification and matching but the actual posting logic (double-entry maths, VAT calculations, foreign exchange conversions) is exact and rule-based, never probabilistic.

Example

The AI decides a transaction is a software subscription (judgment). The ledger posts a debit of exactly £49.99 to Computer Equipment and a credit of exactly £49.99 to the bank, with VAT calculated at precisely 20%. The maths never varies.

How Deterministic Ledger Works in Practice

The concept of a deterministic ledger addresses a core concern about AI in accounting: can you trust the numbers? The answer is yes, when the system separates judgment from arithmetic.

AI is excellent at judgment calls: is this transaction an office supply or a travel expense? Does this bank payment match invoice #1042 or #1043? These decisions require context, pattern recognition, and sometimes ambiguity resolution. AI handles them well.

But the actual ledger postings must be exact. Double-entry requires that debits equal credits. VAT must be calculated at the correct rate to the penny. Foreign exchange conversions must use the documented rate. These are not judgment calls; they are arithmetic, and they must be deterministic (the same inputs always produce the same outputs).

A deterministic ledger uses AI for the judgment layer and rules-based logic for the posting layer. This gives you the benefits of AI classification without introducing uncertainty into your financial records.

Step by Step

The system has two distinct layers. The classification layer uses AI to determine what a transaction is, who it is from, which account it belongs to, and whether it matches an existing record. This layer can be probabilistic and learns from corrections.

The posting layer is pure rules-based logic. Once the classification is made, the posting follows deterministic rules: this account is debited, that account is credited, VAT is calculated at the statutory rate, and the amounts must balance exactly. There is no randomness, no approximation, and no temperature setting.

This separation is why an AI-native accounting system can be both intelligent (in its classification) and trustworthy (in its arithmetic). The audit trail connects the two layers, showing which AI classification led to which deterministic posting.

Practical Tips

  • Run a trial balance regularly to confirm debits equal credits, which verifies the deterministic posting logic is working correctly
  • Focus your review effort on the classification layer (are things in the right account?) rather than checking arithmetic
  • Use the audit trail to trace any unexpected figure back to the AI classification that produced it

Common Mistakes to Avoid

  • Assuming that if AI is involved anywhere, the numbers might be approximate: the ledger maths is exact regardless of how the classification was made
  • Not checking the classification layer: the numbers will always balance, but if the AI categorised something incorrectly, the balanced entry is in the wrong account
  • Over-trusting the deterministic label: the maths is exact, but the inputs (bank feed data, OCR extraction) should still be verified

Frequently Asked Questions

Does deterministic mean the AI cannot make mistakes?

Deterministic refers to the ledger maths, not the classification. The AI might miscategorise a transaction (judgment error), but the journal entry that results will always balance correctly (arithmetic certainty). The classification error is caught in the review process.

How can I verify the ledger is actually deterministic?

Check the audit trail. Every journal entry should show exact amounts that balance to zero. Run a trial balance: debits must equal credits. These checks confirm the posting logic is deterministic regardless of how the transactions were classified.

Related Terms

AI Ledger

An AI ledger is a general ledger that is continuously maintained by an AI agent, with human review on exceptions and sign-off on filings, rather than being updated in periodic manual batches.

AI-Native Accounting

AI-native accounting is an accounting platform built from the ground up around AI agents and a living ledger, rather than adding a chatbot to traditional accounting software.

Show the Working

Show the working is the principle that every answer an AI accountant gives should be traceable back to specific journal entries, transactions, and source documents, so you can verify any figure rather than trusting a black box.

Accounting Rules Engine

An accounting rules engine is the deterministic layer in AI accounting that enforces double-entry posting, VAT rates, filing formats, and regulatory constraints, ensuring the ledger stays correct regardless of what the AI classification layer decides.

Confused by accounting jargon?

AccountsOS explains everything in plain English. Ask any question about your books and get a clear, jargon-free answer.

Try Free for 14 Days
AI accounting glossaryView full glossary