Overview
Suggested Categorisation helps Safety teams classify reports faster and more consistently. It suggests best-fit categories for each report, learned from how your organisation classifies its own reports - the reviewer can accept each suggestion as-is, adjust it, or remove it entirely. It speeds up the category selection of the Classify Risk phase, while keeping the final decision with the person reviewing the case.
Before you start:
This feature is optional and disabled by default. To use it, an organisation admin must:
- Opt in to the feature for your organisation
- Agree to the AI data processing terms, as part of enabling AI-powered functionality
Once enabled, the system begins analysing your reports to generate categorisation suggestions. There's no per-user setup - it applies organisation-wide once switched on.
If you're interested in activating this functionality, please contact TrustFlight Support.
Access Rights
To view and action suggested categorisation, at least one of the following is required:
- The 'Manage Cases' permission for the role or individual.
- Delegation of a specific case, even if the user doesn't hold 'Manage Cases' more broadly
How It Works
Step 1: Learning From Your Data (background)
Once enabled, the system analyses how your organisation has historically categorised its reports.
Categories and their selectable options are configurable per organisation - each customer requests their own set of categories and the options available within them, so the fields and values listed below are examples only, not a fixed list. Commonly seen categories include:
- Operating Base
- Risk Category
- Department
- Affected Fleet
- Link to AOC Risk Register
This lets the model learn your organisation's own categories and its own historical classification patterns, rather than applying a generic, one-size-fits-all model. No user action is required - this runs in the background.
Step 2: Suggestion Generation (background)
When a report reaches the Classify Risk phase, the system generates recommended values for your organisation's configured categorisation fields, based on the report's content and how similar reports have been classified previously within your organisation's own categories.
- Suggestions only appear for fields where the model has sufficient confidence in a value
- Fields with low confidence are left blank for the reviewer to complete manually, rather than guessing
- No suggestion is confirmed until a reviewer actively saves
Step 3: Review & Apply (reviewer-facing)
On the case's Classify Risk screen, a Smart Suggestions banner appears, showing how many suggestions are ready - e.g. "You have 5 suggestions ready to review and apply" - with a Review suggestions button.
1. Review Suggestions opens a panel from the right-hand side of the screen, titled "Recommended categorisations." This lists each of your organisation's configured categories and its suggested value - for example (illustrative only, using the common example categories):
- Operating Base - Prestwick
- Risk Category - Safety > Equipment > Vehicles
- Department - Third Party > Supplier
- Affected Fleet - B737-800
- Link to AOC Risk Register - Runway Incursion (CAA S7) (07)
An "About AI-powered suggestions" note sets expectations for the reviewer:
- You stay in control - nothing is recorded until you save the form
- Suggestions only appear for fields with sufficient confidence; low-confidence fields are left blank for you to complete
- Your edits train the model - whether you accept, change, or remove a suggestion is fed back to help improve future suggestions
2. Back on the case form itself, each suggested field is marked with a "Suggested" tag next to it, pre-filled with the recommended value. The reviewer can, per field:
- Accept the suggested value as-is
- Adjust it to a different value from the dropdown (or add a new one via "+ Add new...", e.g. "+ Add new department")
Leave it blank if it doesn't apply
3. The reviewer clicks Save to confirm the categorisation (or Cancel to discard).
Things to Know
- Suggestions, not decisions. Nothing is saved until the reviewer clicks Save - every suggestion can be accepted, changed, or ignored.
- Suggestions are confidence scored. A field with low model confidence is simply left blank, rather than filled with a low-quality guess.
- The model learns from your organisation specifically. Suggestions are based on how your organisation has classified reports historically, not a generic external standard.
- This applies to any case reaching classification
Key Benefits
- Speeds up the categorisation within the Classify Risk phase of a case
- Learns from your organisation's own historical classification patterns, not a generic model
- Reduces inconsistency between how different reviewers classify similar reports
- Keeps the final decision with the reviewer - every suggestion can be accepted, adjusted, or discarded
- Confidence-based: only surfaces a suggestion when the model has enough basis to make one
Summary
Once enabled, AI Smart Suggestions – Suggested Categorisation:
- Suggests best-fit categories for a report, learned from how your organisation classifies its own reports
- Only surfaces a suggestion where the model has sufficient confidence, leaving other fields blank
- Lets the reviewer accept, adjust, or discard each suggested field before saving
- Applies to any case reaching classification
Helping Safety teams:
- Classify reports faster and more consistently
- Reduce reviewer-to-reviewer variation in categorisation
- Keep full control over the final classification decision