Compliance asked for a wallet trail and got a spreadsheet

Someone in compliance asks a simple question: where did the funds in this wallet come from, and did they touch anything we should care about?

The data to answer it is public. That is the part people find surprising when the answer takes a week.

Public is not the same as readable

Open a block explorer on any address and you get exactly what the chain stores: hashes, addresses, timestamps, values in the smallest unit the network uses. It is complete, it is verifiable, and it is unreadable by anyone whose job is not reading it.

So the work gets done by whoever can. They export transactions to a spreadsheet, filter, cross-reference addresses against whatever list they have, and produce a summary. That summary is what compliance actually receives.

And the summary has a problem that has nothing to do with blockchain: nobody who reads it can check it. The reviewer sees a conclusion and a CSV. Reproducing the reasoning means redoing the work, so it does not get reproduced, and the conclusion is accepted on the strength of who wrote it.

The three questions that are actually being asked

Strip the jargon and compliance questions about an address are usually one of three.

Where did this come from. Trace inbound funds back through however many hops until you reach something identifiable — an exchange deposit, a known service, a contract.

Where did it go. The same walk forwards, which matters more than people expect, because the interesting behaviour is often the split immediately after a receipt.

Does it touch anything flagged. Compare the addresses on those paths against sanction lists, known-incident addresses, or the firm’s own list.

None of these needs a blockchain engineer. All of them currently get one, because the tools speak in hashes.

What a usable answer looks like

Three properties, and the third is the one people forget.

It answers the question that was asked, in the words it was asked in. “Show me the last five transactions for this address and tell me if any came from an exchange” should be a sentence you type, not a query you construct.

It returns the reasoning, not just the result. Which addresses were walked, in what order, how many hops, where the trail stopped and why. A conclusion without a path is an opinion.

And it is reproducible by the person reviewing it. This is what separates a tool from a report. If the reviewer can re-run the same question and get the same answer, the finding stands on the method. If they cannot, it stands on trust, and trust is not an audit control.

The honest limits

Any tool that reads a chain can tell you what happened on it and cannot tell you who did it. Attribution comes from off-chain sources — exchange records, subpoenas, disclosures — and anyone claiming otherwise is selling inference as fact.

Mixers and privacy protocols end a trail. The correct output there is that the trail ends, not a probabilistic guess presented with the same confidence as a traced hop. A tool that quietly downgrades certainty without saying so is worse than no tool, because the reviewer cannot see where the evidence stopped.

We built the AI blockchain explorer around asking in plain language and getting the path back with the answer. The demo page shows the kinds of question it handles and where it stops, and it links to a live version you can point at a real address.

If the requirement is broader than reading — custody, contracts, an internal chain of your own — that is custom blockchain development, and the questions worth settling first are about who needs to verify what.