Articles · Money & Agents
Show me exactly what happened
When an AI gets it wrong, the first question is always the same — and most businesses cannot answer it. The AI Receipt Book writes a signed receipt for every action, at the moment it happens.
By Lucian · Lux Automations
The problem
A company has already been held liable for its own chatbot’s bad advice — the business argued the chatbot was a separate entity, and the tribunal disagreed. The lesson travels well beyond one case: when an agent or chatbot does something wrong, the business answers for it, but most businesses cannot show exactly what their AI did, said or was told.
Logs are scattered across tools, editable by anyone with admin access, or simply missing. An editable log is weak evidence, because the party it describes could have rewritten it. So the first question from a regulator, an insurer or a court — show me exactly what happened — gets an answer built from fragments, memory and screenshots.
The technology
The AI Receipt Book gives every action a signed, tamper-evident receipt, written at the moment it happens. Each receipt names the agent that acted and the mandate it acted under, because an action without an accountable identity and a permission behind it is only half a record.
The receipts are written into a hash-chained, append-only book. A page quietly changed afterwards breaks the chain — in the open — so the book’s honesty does not depend on trusting whoever kept it. When someone asks what happened, the whole story can be exported as one readable report, in a form a non-technical reader can follow. The direction of law points the same way: the EU AI Act already writes logging and traceability duties into law for covered systems, so a business that keeps receipts is preparing for questions that are coming anyway.
What it changes
What changes is who carries the burden of a dispute. Instead of scrambling to reconstruct an incident after the fact, the business hands over the book: what the agent did, when, under whose instruction and inside what mandate, in a chain that shows if it has been touched.
That record also helps on ordinary days, not just bad ones. Patterns become visible — which agent does what, how often, and where its work bunches up — because the receipts were written as the work happened, not assembled later from whatever the tools happened to keep.
The book also settles smaller arguments, the kind that never reach a tribunal. A customer says the agent promised something; the team remembers it differently; the receipt shows what was actually said and done, and under what mandate. Disagreements that used to run on memory and confidence start running on a record both sides can read — which is, in the end, what a receipt has always been for, now written for a worker that is software instead of a person.
See it
The AI Receipt Book turns what your agents did into a book you can hand to anyone who asks. AI Receipt Book — the product page · All the house articles · Lux Automations — ailux.ca