How Scout works

What the reviewactually does.

Scout reviews one source for one decision. We will follow the pilot report from the previous page. It presented 13.4×, returned 7.2× across the whole pilot, and the next campaign's budget is still to set.

Plan an evaluation Read the White Paper

Start with the source.

Upload the files. PDF, Word, Excel, Markdown and text are supported. VALIS Readback records what the AI received from each one, and any uncertainty about it, before the review begins.

Know what your AI actually read.

A packet becomes one source. Every document keeps its own identity and its place in the order, because an amendment after an agreement means something different from the same files in reverse.

Scout freezes that source for the review. An important change starts a new review.

Name the decision.

Tell Scout what you plan to do with the document, and what happens if its reasoning fails.

The pilot report is good enough to tell the team what happened. It is not good enough to set the next campaign's budget.

Same document, two decisions, two results. The decision you name determines which findings matter and how far the review has to go.

Choose the depth.

Scout Rapid

A fast check while the work is still changing.

Scout Rapid reviews one document. It identifies claims, assumptions, numbers, the links between them and likely missing support, then tells you whether to continue, repair the document, find a better source or move to Scout Full.

Scout Full

A deeper review before a decision that is hard to reverse.

Scout Full reviews a document or an ordered set of them as one composed source. It adds an independent second reading of that frozen source, tests what the conclusion depends on, challenges the first reading and reviews the whole of it.

The review runs.

Every review starts from the same frozen source.

Readback

Record exactly what the AI received, and any limit that could affect the review.

Check

Test claims, assumptions, evidence, numbers and stated relationships.

Trace the reasoning

Identify what the conclusion depends on, and test whether the document supports it.

Scout Full adds

F01

An independent reading

A second review reads the same frozen source for the same decision, without seeing the first review's judgments.

F02

Challenge and resolution

The two readings can disagree — one treating a headline figure as the campaign's result, one not. Scout returns to the source to settle it rather than taking a vote.

F03

Whole-document review

Check the full chain of reasoning and confirm every required step finished before a result is returned.

Every finding points at the source.

Each finding shows the exact passage, what it is doing in the argument, and what it would take to settle it.

Finding

The headline return covers part of the spend

Source: The presented return, 13.4×.
The finding: That figure is measured over an early tranche of spend. Across the whole pilot the return is 7.2×.
What depends on it: The case for committing the next campaign's budget.
Why it matters: Both figures describe the pilot. Only the whole-pilot result matches the scope of the next decision.
Still open: Which basis the next plan should be set against.
What to do: State the spend each figure is measured over, and restate the case on whole-pilot performance.

Finding

The gap is carried forward without being stated

Source: The two return figures, and the spend planned for the next campaign.
The finding: The report does not carry its own headline figure forward to the spend the next campaign is planning.
What depends on it: What the next campaign is expected to return.
Why it matters: Applied to the next $2.5 million of spend, the difference between the two figures is a $15.5 million expectation gap.
Still open: The calculation is clear. Which figure to plan against is a judgment for the budget owner.
What to do: Show both bases together, and name the one the plan uses.

You inspect the author's exact words and numbers, not a summary of them. Source locations are withheld for this run. In a review they are the passage, the figure or the table itself.

The result.

A completed Scout Full review returns one result for the decision you named.

Decision Grade

The document supports the decision you named.

Conditional

It supports the decision under conditions Scout states.

Not Yet

It does not yet support the decision you named.

Set against the next campaign's budget, the pilot report does not yet support the decision. Set against telling the team what happened, it does. Same document. Different intended use. Different result.

Repair and recheck.

Use the findings to correct the document, supply the missing evidence, run the calculation again, or narrow what you intend to do with it. A recheck records what changed and whether the problem was resolved.

The result stays tied to the source, the decision, the review and the repair history, so a later reader can see why Scout reached it.

Continue with Scout

Test it on your own work.

Plan an evaluation Read the White Paper