PlaneConnection integrates AI capabilities throughout the safety module to assist safety managers and reporters with classification, analysis, and report generation. These features use large language models and structured data analysis to accelerate safety workflows while keeping human decision-making at the center of all safety assessments.
AI suggestions are advisory only. Per 14 CFR Part 5, all safety assessments, investigation
findings, and risk evaluations require independent human evaluation by qualified personnel. AI
recommendations should be used as one factor among many in decision-making, not as the sole basis
for any safety, operational, or maintenance action. See Legal
Notices. AI Capabilities Overview
| Capability | Location | Purpose |
|---|
| Safety Copilot | Safety > Copilot (sidebar panel) | Context-aware AI assistant for drafting, analysis, and guidance across all safety pages. |
| AI Insights | Safety > AI Insights | Automated trend analysis and pattern detection across safety reports, investigations, and SPIs. |
| NLP Report Submission | Safety > Reports > Natural Language | Free-form narrative submission that extracts structured report fields automatically. |
| AI Classification | Safety report processing | Automatic categorization, severity suggestion, and risk scoring for incoming reports. |
| AI Risk Assessment | Safety report processing | Preliminary risk evaluation based on report content and historical data. |
| Investigation Assist | Investigation detail page | Drafting summaries, identifying contributing factors, suggesting lines of inquiry. |
Safety Copilot
The Safety Copilot is a context-aware AI assistant available throughout the safety module. It appears as a sidebar panel and has access to the current page context — including report numbers, investigation status, findings counts, and hazard data.
What the Copilot Can Do
- Draft summary language for investigations and reports.
- Identify potential contributing factors based on event descriptions.
- Suggest additional lines of inquiry during investigations.
- Explain regulatory requirements in plain language.
- Help compose hazard descriptions and risk control narratives.
- Answer questions about SMS processes and terminology.
What the Copilot Cannot Do
- Make safety decisions or approve investigations.
- Access data from other organizations or workspaces.
- Override human classifications or risk assessments.
- Guarantee regulatory compliance of any output.
AI Insights
The AI Insights page analyzes patterns across your organization’s safety data:
| Insight Type | Data Sources | Output |
|---|
| Report trends | Safety reports over time | Category distributions, frequency trends, seasonal patterns. |
| Risk concentrations | Hazard register, risk assessments | Areas with elevated risk density, common hazard categories. |
| CPA effectiveness | CPA completion and verification data | Patterns in overdue actions, effectiveness ratings, recurrence rates. |
| Investigation patterns | Investigation findings and root causes | Common root cause categories, average investigation duration, findings per category. |
NLP Report Submission
Natural language report submission allows reporters to describe a safety event in free-form text. The AI parses the narrative and extracts structured fields.
| Field | Extraction Method |
|---|
| Report type | Classified based on event description (hazard, incident, near-miss, concern, observation, audit finding). |
| Date occurred | Extracted from temporal references in the narrative. |
| Location | Identified from airport codes, facility names, or geographic references. |
| Description | Structured from the narrative with key facts highlighted. |
| Severity indicators | Assessed from the described outcome and contributing factors. |
| Category | Mapped to the standard safety report categories. |
Submission Flow
- Reporter selects Natural Language submission mode.
- Reporter writes a free-form description of the event.
- AI extracts structured fields and presents them for review.
- Reporter reviews, corrects, and confirms the extracted data.
- Report is submitted with the finalized structured fields.
The reporter always reviews and confirms AI-extracted fields before submission. No report is filed
without explicit human confirmation of the structured data.
AI Classification
When a safety report is submitted, the AI classification engine analyzes the report content and provides:
| Output | Field | Description |
|---|
| Classification | ai_classification | Suggested category and subcategory based on report content. |
| Risk Assessment | ai_risk_assessment | Preliminary risk evaluation including suggested severity and likelihood. |
Confidence Levels and Reliability
AI outputs in PlaneConnection should be interpreted with the following confidence guidance:
| Confidence Level | When It Applies | Guidance |
|---|
| High | Structured data extraction (dates, locations, aircraft registrations) from clear text. | Generally reliable. Verify against source. |
| Medium | Category classification, severity suggestion, trend identification. | Use as a starting point. Apply professional judgment before accepting. |
| Low | Root cause suggestions, contributing factor identification, predictive analysis. | Treat as exploratory input only. Always validate with domain expertise and evidence. |
AI confidence levels are not displayed in the application as numerical scores. The levels above
are guidance for interpreting AI suggestions across different use cases. All AI outputs require
independent human evaluation.
Regulatory Advisory
AI features in PlaneConnection are designed to assist, not replace, human safety decision-making. The following regulatory principles apply:
| Principle | Regulatory Basis | Application |
|---|
| Human authority | 14 CFR Part 5 | All safety assessments, investigation findings, and risk acceptability decisions require qualified human evaluation and approval. |
| Accountability | 14 CFR 5.23 | The accountable executive and safety manager remain responsible for all safety decisions, regardless of AI input. |
| Documentation | 14 CFR 5.97 | AI-assisted outputs that contribute to safety decisions should be documented alongside the human rationale. |
| Non-punitive use | 14 CFR 5.21(a)(4) | AI classification outputs are not used for punitive purposes. |
Data Sources and Limitations
Data Sources
AI features in the safety module draw from the following data within your workspace:
- Safety reports (all types and statuses).
- Investigation records, findings, and root cause analyses.
- Hazard register entries and risk assessments.
- CPA records and verification outcomes.
- SPI data and trend history.
- Safety policy and training records.
Limitations
| Limitation | Description |
|---|
| Workspace scope | AI analysis is limited to data within your workspace. It does not access industry-wide data, other organizations’ records, or external databases. |
| Historical depth | Trend analysis quality improves with data volume. New organizations with limited history will receive less detailed insights. |
| Regulatory currency | AI models are trained on regulatory content available at the time of model training. Always verify regulatory citations against current published regulations. |
| Context gaps | AI cannot observe physical conditions, interview witnesses, or assess non-documented factors. It works only with the data entered into the system. |
| No real-time data | AI does not integrate real-time operational data (weather, ATC, flight tracking). It analyzes recorded safety data. |
Use Safety AI
How to use the Safety Copilot and AI Insights features.
Submit a Natural Language Report
How to submit safety reports using free-form text.
Report Types
Report types that AI classification targets.
Investigation Workflow
How AI assists during investigations.
