Disclosurely
Disclosurely →

AI Case Analysis

Use the AI Assistant to help an authorised case handler review a whistleblowing report. AI output is a draft for human review, not a decision.

Overview

When a case handler chooses to start it, the AI Assistant can help them:

  • Summarise the key points of a report
  • Suggest a risk level and the concerns behind it
  • Suggest questions to ask and next steps to consider
  • Draft wording for a message, which the handler edits before sending

AI never runs automatically when a report is submitted. It runs only when an authorised case handler starts the AI Assistant or asks for a report or message to be translated.

How It Works

1. A case handler starts it

Nothing is sent for AI processing until an authorised case handler opens the AI Assistant on a case, or asks for a report or message to be translated.

2. Processing on Amazon Bedrock in London

The request is processed by Amazon Web Services (Amazon Bedrock) in the London (UK) region.

3. The handler reviews the output

The output is a draft. The case handler checks it against the original report and decides what, if anything, to use.

Privacy & Security

What is sent

  • The report and message content the handler asks the AI to work with
  • Any policy documents the handler chooses to add

Personal information in report text is not removed or redacted before processing. If a reporter has included names or other identifying details, those details are part of what is processed.

Safeguards

  • Handler-initiated only - No AI processing on submission
  • Access control - Only users who can already see a case can use the AI Assistant on it
  • Encrypted transit - Data is encrypted in transit to Amazon Bedrock
  • Human review - AI output is a draft, never a final decision

Usage

Analysing a Report

  1. Open a report in your dashboard
  2. Open the AI Assistant
  3. Optionally add your own policy documents
  4. Start the analysis
  5. Review the suggestions before acting on them

Asking Follow-Up Questions

After the first analysis, you can ask follow-up questions, for example:

  • "What are the main risks here?"
  • "What should we ask the reporter next?"
  • "Draft a reply to the reporter"

Best Practices

When to Use AI

✅ Good use cases:

  • Getting a first summary of a long report
  • Identifying possible investigation steps
  • Drafting communications for a handler to edit

❌ Avoid using AI for:

  • Final legal decisions (consult counsel)
  • Disciplinary actions (requires human judgment)
  • Cases requiring nuanced cultural understanding

Reviewing AI Output

Always review AI suggestions critically:

  • Verify facts against the original report
  • Consider organisational context
  • Consult legal/HR for serious cases
  • Use it as a starting point, not a final answer

FAQ

Who processes the data?
Amazon Web Services, using Amazon Bedrock in the London (UK) region.

Are names and other personal details removed first?
No. Personal information in report text is not removed or redacted. Only start the AI Assistant on a case when that processing is appropriate.

Does AI run on every new report?
No. AI runs only when an authorised case handler starts it.

How accurate is the AI?
AI output can be wrong or incomplete. Treat it as a draft for human review.