Touchledger
A per-transaction evidence ledger for AI bookkeeping vendors that records source, proposed treatment, reason code, model and policy versions, human review, destination readback, correction, and signed integrity receipt.
AI bookkeeping systems can classify, code, match, reconcile, and post transactions, but a generic application log rarely answers what source supported a treatment, which model and policy proposed it, whether a human changed it, what reached the ledger, and how it was corrected. Research confirms a 2026 regulatory date and no direct per-transaction competitor. The product should store bounded reason summaries rather than hidden chain-of-thought, and a signed receipt must prove integrity only -- not accounting correctness, audit sufficiency, legal compliance, or survival of an inspection.
The product, controls, accounting, risk, or audit lead at an AI bookkeeping vendor or CPA firm consuming its evidence. The input does not establish scale, budget, or current logging cost.
A confirmed August 2, 2026 date creates a preparation window for governed systems.
Event capture, schemas, signing, tags, exports, and verification scale through software.
Two cross-references and one inbound connection support a modest cluster.
A confirmed August 2026 date, accounting-agent convergence, free signing infrastructure, and no direct competitor make the evidence layer timely and scalable.
Two API references were unverified, the exact buyer is incomplete, legal and audit acceptance cannot be promised, and incumbents can add transaction-level logs.
Discussion
No comments yet — be the first to weigh in.