Third Party AI Summary

Follow

About The AI Summary

The AI Summary panel gives you a factual summary of a third-party record, so you can see what the record holds without opening each assessment and screening in turn. It covers the entity, its risk rating and due diligence status, the key findings from its assessments, questionnaires and screenings, and the mitigations and controls in place. Each finding links to the record item it comes from, so you can check it at the source.

The summary reports what the record holds. It does not set a risk rating or recommend a decision. The risk decision stays with the reviewer.

 

Before You Start

You see the AI Summary panel on a third-party record when your organization has AI enabled. Your administrator turns on TPRM AI Summary in System Administration > AI Features (see Enabling AI Features). If you can view a third-party record but don't see the panel, contact your administrator.

 

Generate A Summary

  1. Open a third-party record in Third Party Risk Management.

  2. Open the AI Summary panel on the right of the record.

  3. Select Generate summary.

18535ff7-6a89-4f25-9187-dcf6c865881b.png

 

The summary is generated each time you select Generate summary. It is not saved to the record and does not update on its own, so it reflects the record at the moment you generated it. If the record changes, generate the summary again.

 

210ea555-b896-452f-bf81-f81ddbc74370.png

 

What The Summary Contains

A generated summary has three parts: a Summary for a quick view of the whole record, Key Findings about the record including screening results, and Mitigation & Controls currently associated with the record.

 

Risk Rating

The Risk Assessment line reports a rating only from the Final Evaluation or the Assessment. It never works one out from CPI, screening results or other values on the record.

If the record has a Final Evaluation risk rating, the summary reports it exactly. If there is no Final Evaluation but the Assessment holds a risk rating, the summary reports that rating and says where it came from, for example "Medium (Assessment risk rating; no Final Evaluation completed)". If the record holds neither, the summary gives no overall rating. A blank rating means the record doesn't have one.

 

Key Findings

Key findings are grouped by theme in a fixed order, for example sanctions, export controls, forced labor, adverse media, political exposure, regulatory action, environmental risk and shell company risk. Record-specific items follow, such as contract value, dates, questionnaire findings and screening status.

The summary also reports other items linked to the third party, such as further assessments, Evaluations, Enhanced Due Diligence requests, Conflicts of Interest declarations and Gifts & Hospitality requests. A completed Internal Questionnaire or External Questionnaire is included with its material answers, even when those answers are clean.

A recorded Highest Severity value is reported exactly. The CPI score is shown with its country, for example CPI 89 (Denmark). If the Assessment records a different CPI, both values are given. If a recorded Location and the address country disagree, the summary points this out. For a third party that is a person, the summary gives the recorded date of birth.

Wording in a finding, including risk language, comes from the provider's report. The summary doesn't add its own.

 

Screening Results

Each screening outcome is stated as it is recorded, for example Pending Screening, No Match Found, Match Found with matches awaiting review, Matches Applied or Matches Discarded. A screening that is pending, was discarded or was never run is not presented as a clean result. Repeated screenings with the same result and provider are combined into one line with a count.

For every screening, the summary also gives the provider, screened entity, location and date requested where the screening record holds them. If no provider is recorded, the summary says so.

A matched entity is named only when a reviewer has applied the match. Matches awaiting review and discarded matches are reported at the screening level and are not named. If a screening was run against a different name from the third party, for example an associated person, the summary names that entity and does not present its findings as the third party's.

How much detail the summary gives beyond the status depends on the provider. See Screening Provider Support below.

 

Mitigation & Controls

This section lists only the mitigations and controls held on the record. Ongoing monitoring is listed as a control when the screening record shows monitoring. Assessments, approvals, Final Evaluations, reports, Enhanced Due Diligence requests, questionnaires and workflow statuses are not listed here. If the record has no mitigations and no controls, the section says none are recorded.

 

Screening Provider Support

Some providers store their findings as text on the screening, and the summary reports those findings. Others store only a status, with the detail held in a report or on the match screens. For those, the summary tells you only that the screening ran and what its status is.

Provider

What the summary includes

Sayari

Risk factors grouped by theme, and the highest severity, once matches have been applied. Not the ownership network or corporate hierarchy.

Integrity Essential

The narrative text for each category, for example Adverse Media, Bribery & Corruption, Political Exposure and Sanctions/PEPs/Watchlist Match.

Integrity Enrich

Everything for Integrity Essential, plus each red flag and its figures from the Deep Search Report.

Vantage

The matches you have applied: name, location, match score, reason, source, offence and notes. Not the Vantage report PDF.

Dow Jones

The Dow Jones risk categories, where the screening has them, with no further detail.

RDC

Status only.

Orbis (Moody's)

Status only.

Note - Integrity Essential and Integrity Enrich screenings show the provider as Integrity Due Diligence, on the screening and in the summary.

Applied matches from Sayari, RDC, Dow Jones and Orbis are not listed individually. For Sayari, the risk factors from all applied matches appear as one list, so the summary can't say which matched entity a risk factor belongs to. For RDC, Dow Jones and Orbis, a Matches Applied status tells you that matches were applied, but not what they are.

An applied Vantage match appears without a link to its screening and without the screened entity it was matched against. Associations, source links and additional match reasons are not included. Check the match screen before you rely on a match in the summary.

For Integrity Essential and Integrity Enrich, the articles behind the narrative text, and their links, are not included.

 

Check The Summary Against The Record

AI-generated content may contain inaccuracies. Before you make a compliance decision, check any finding you rely on against the source data on the record.

Every citation link opens the assessment or screening it refers to in a new browser tab, so the summary stays open while you check the source. Applied screening matches have no citation link. Open the screening to check them.

 

What The Summary Does Not Cover

The summary works from the data held on the record. It does not open attached files, so a report added to the record, or a provider report delivered as a PDF, is noted but not summarized.

The summary doesn't give a count of matches awaiting review. It says that matches are awaiting review, and you open the screening to see them.

The screening status in the summary can lag the screening itself. Open the screening to confirm its current state.

 

Your Data

The summary is generated by Claude Sonnet 5.5, a model from Anthropic, which GAN Integrity accesses through Amazon Bedrock. Your record data is not used to train the model and is not shared with Anthropic.

Your data is processed in the same region as your GAN Integrity system. If your system is hosted in the EU, the model processes your data in AWS regions in the EU. If your system is hosted in North America, the model processes your data in AWS regions in the United States.

Was this article helpful?
0 out of 0 found this helpful