Home · Platform · Proof vs accountability
Why AskLedger

A signed receipt is necessary. It is not sufficient.

Cryptographic proof is real, and it matters. A signed record tells you a known signer wrote specific fields and nothing changed after. But proof alone can't tell you who across your organization is using AI, what it costs, what risk it carries, or whether it worked. To answer those, you need the proof layer plus the cost and control that sit on top of it. AskLedger gives you both.

What a receipt is for

Proof answers one question: did the record change?

A cryptographic receipt is a verifiable claim about a record's integrity, and AskLedger is built to produce the strongest one. Integrity is not the same as truth, coverage, or impact. It's worth being precise about exactly what a receipt does and does not prove, because overclaiming is how trust in evidence gets lost.

A receipt proves
  • A known signer recorded specific fields, and the record has not changed since.
  • The relationship among linked events and checkpoints in a chain.
  • Which policies were evaluated, and what decision was recorded.
  • That an exported evidence package matches the stored evidence.
A receipt does not prove
  • That every field supplied by a model provider was factually true.
  • That all activity was captured, since users can bypass instrumentation entirely.
  • That the policy itself was adequate, sufficient, or legally compliant.
  • That a correlation the platform recorded equals causal business impact.
The gap proof leaves

Five questions proof alone can't answer.

Answering for enterprise AI means answering, defensibly, five things about every use of AI. A stack of perfectly signed receipts still leaves most of them open. These are questions of who used the tool, what it cost, what risk it carried, and what it produced, not questions of integrity.

Who or what is using AI: people, apps, service accounts, agents?
Which tools, models, data and permissions were involved?
What did it actually cost, end to end?
What business outcome or operational risk resulted?
Can you independently verify it later, when challenged?

A receipt answers the last question well. It barely touches the first four. That is the gap between proving a record and running the AI behind it.

What sits on top of the proof

Receipts, then the things receipts can't do alone.

AskLedger starts with the strongest proof layer, then joins it to the context that makes it mean something. This is the layer proof-only tools stop short of.

Discovery

Org-wide discovery

Signed telemetry only sees instrumented systems. AskLedger joins identity, billing and security signals to surface shadow AI, the usage no receipt was ever written for, and labels each source's confidence honestly.

Cost

Normalized cost + verified savings

Every provider prices differently. AskLedger normalizes tokens, requests, tools and licenses into one cost model, attributes it, and then proves realized savings with a baseline and the math, not an estimate.

Control

Agent accountability

Pre-approved scopes and limits, attempted vs. executed actions recorded separately, and tested pause, revocation and rollback, so an autonomous agent has a named owner and a defensible trail.

Outcome

The AI Activity Graph

One normalized graph joining identity, cost, policy, outcome and evidence, turning disconnected logs into a queryable history of who acted, under what authority, at what cost, producing which result.

The position

The deepest proof, plus the cost and control it can't give you alone.

This is the same proof that verifies your AI savings, put to work on cost and control. We're a design-partner-stage company building toward it, in the open. The receipt specification and the verifier stay free and open, because proof should never be something you rent. The cost and control layer that sits on top is where the enterprise value lives. Try the proof yourself, then see what sits on top of it.