Analysis results

Where: workflow detail → profile detail → Analysis Results tab Results collect on this tab every time you run an analysis. With no run history, the screen shows that there are no analysis results yet.

Element	Contents
Analysis ID	`analysis_` prefix
Status badge	Pending · Processing · Completed · Failed
Analysis selector	Switches to an earlier analysis of the same profile. The 5 most recent are kept per profile
Download PDF	Enabled for completed analyses only. The report is generated automatically when the analysis finishes
Delete analysis	Requires an administrator email to confirm

When an analysis fails, the error code and message appear in an analysis error area.

Analysis Score

Part	Weight	Meaning
Step Execution	40%	Share of the steps defined in the playbook that actually ran
Step Quality	20%	Pass status of the steps that ran, and whether engines were called
Output Completeness	40%	Share of output-schema leaf fields that received a value

Score Level is HIGH (80–100) · MEDIUM (65–79) · LOW (0–64).

The Analysis Score measures how completely the output was filled in, not the verdict. A run where every step passed can still score low if the output schema came back empty. Read it alongside the verification verdict when you decide on approval — reading analysis results.

The verification verdict is shown separately — Verified · Pending review · Rejected · Unverified.

The five sub-tabs

Summary

Summarizes what the agent looked at and how it decided.

Decision variable detail panel — Selecting a decision variable opens a detail panel on the right.

Area	Contents
Decision variables	The fields defined as verdict items in the output schema, with their final values. Select one to open its detail panel
What the agent checked	What each step did, shown as a pair — Query (the instruction the agent received) and Reasoning (the grounds for the result)
Recommendations	Suggested follow-up actions based on the result

When there are no decision variables or no steps, the screen says so in place of the list.

Report

Shows every field defined in the output schema, in the schema’s own structure. Fields with no value are marked as not extracted.

Element	Contents
Completeness	Share of fields that were filled in
Reasoning	Why each field ended up with the value it has. When none was recorded, the screen says no reasoning was provided
Not extracted	Marks fields whose value is empty. An empty list shows “None”
Verification results	Category and detail per action
Copy	Copies the result to the clipboard
Download PDF	Saves the report file

“No reasoning provided” does not mean there were no grounds — it means the reason could not be supplied. It appears for analyses that finished before the reasoning feature existed, and when output-schema assembly failed.

PDF download errors

Condition	Cause
The report can be downloaded after the analysis completes	The analysis is not `completed`
The report is still being prepared	Report generation has not finished
The download period has expired, and the report must be regenerated	The link expired
The report file cannot be found, and the report must be regenerated	The file is missing
You do not have permission to download this report	Insufficient permission

The downloaded PDF lays out the score, the summary, and the decision variables as a document, along with the analysis ID, completion time, and profile ID, so you can file or submit it as-is.

PDF analysis report — The downloaded PDF report.

To get the same file through the API, use GET /analyses/:analysisId/report.

Audit

The record of the steps the agent ran — status, tokens used, execution time, and the tools it called.

At the top are totals for steps, passed, needs review, tool calls, RAG citations, and tokens. Each step carries its rule, query, and reasoning, plus the list of tools it called. The raw audit trail (JSON) gives you the unprocessed log.

Output Schema

The raw JSON of the data the agent extracted, shown exactly as it is with no display processing.

When nothing was extracted, the screen says there is no extracted data.

Info

Area	Contents
Processing info	Processing time, playbook ID, and the requested, completed, and created times
Items used	The documents this analysis verified — item ID, name, and type
Engine	Details when the run pinned a specific engine
System metadata	Iteration count, token usage, and so on
Options · client metadata	Values passed with the request. When there are none, the screen says no options were set
Generated Files	Evidence PDFs the AML engines retrieved for each match. See below

Generated Files

In one screening call, the AML engines (AML Search - Person · AML - Business) retrieve the search results, the details of each match, and an evidence PDF per match. The PDFs appear one per row in the Generated Files (N) area under Metadata on the Info tab.

Element	Contents
File row	PDF badge · file name · size · creation time · Download
One-line description	Per file: `subject · threshold · date-of-birth basis · screening result · match count`
Download	Changes to Downloading while it runs, then saves the original PDF

Examples of the one-line description:

Jordan Avery · Threshold 85 · Date of birth within 2 years · Needs review · 4 matches (31 excluded, no date of birth)
Jordan Avery · Threshold 85 · Date of birth not used · Hit · 53 matches
ACME TRADING LLC · Threshold 90 · Hit · 3 matches

Fragment	Meaning
Subject	Name that was screened
Threshold	Match threshold actually used
Date-of-birth basis	Date-of-birth matching actually used. Business screenings and older analyses run without options have no such fragment
Screening result	Hit or Needs review for the party in this file. It can differ from party to party within the same screening
Match count	Count after removing matches with no date of birth. When 1 or more were removed, the number appears in parentheses

Screening result	Person (AML Search - Person)	Business (AML - Business)
Hit	Current sanctions · current PEP	Current sanctions · former sanctions
Needs review	Any other match	Any other match (state-owned enterprise, reputational risk, etc.)

The screening result belongs to one PDF. It is separate from the verification decision for the whole analysis (Verified · Pending review · Rejected). The threshold and the date-of-birth basis can be fixed on the workflow step — Fixing AML engine options.

Banners — when evidence is missing, a banner appears above the file list.

Banner	Text	When
Cap	`(subject) · Only 50 of N matches were in scope (M excluded by the cap)`	More than 50 matches. PDFs are retrieved for the top 50 matches by score only
PDFs not retrieved	`(subject) · Could not retrieve N PDF(s). Please run the analysis again later.`	A temporary error persisted after the server retried up to 3 times. Shown below the cap banner

If a banner is present, the area shows even when no PDF was retrieved. In that case the title carries no count.
With no files and no banners (for example, zero matches), the area is hidden.
Retrieving evidence PDFs does not change credits: it is still one screening call.

Download error	Cause
File not found	The file does not exist
Could not download the file. Please try again.	An error during the download

Analyses run before 2026-10-22 18:12 (KST) carry a per-screening result, so every file from one screening shows the same value. To get the same file list through the API, use artifacts[] in GET /analyses/:analysisId.

Deleting an analysis result

Selecting Delete analysis opens a confirmation dialog, and deletion requires an administrator email.

The report generated with the analysis is permanently deleted too. This cannot be undone. Only the 5 most recent analyses are kept per profile, so download the PDF for any result you need to keep long term.

Reading analysis results — the order to read the response in, and the interpretation rules
GET /analyses/:analysisId — API field definitions for the values on this screen
Profiles and items — the screen you run an analysis from

Overview

Project

Workflow Setup

Profiles & Analysis

Header

Analysis Score