Skip to content
FRSign inBook a walkthrough

eAvy | The AI assistant inside eAviora

eAvy, the AI assistant inside eAviora.

eAvy reads a record, traces what it links to, finds look-alike occurrences and drafts the next step. Every suggestion waits for a person to accept, change or dismiss it.

Case file | SMS-OCC-2026-0412Stage 3 of 11, Classification
  • Overview
  • Workflow
  • Linked
  • Discussion
  • History
Hard landing, runway 24R2.1 g at touchdown | Flight 412, Toronto to Montréal
ClassificationNot set
92%
eAvy proposes | 92% confidence | 2 creditsRunway Safety, Abnormal Runway Contact

From the narrative: a hard touchdown on runway 24R, 2.1 g at touchdown.

Nothing is written until a person decides.
Sample data | Northwind Air, a fictional airline
What it does

Reads the record, drafts the next step.

Four jobs on one record, each ending in a suggestion for a person.

  1. 01Traces

    Traces the links

    Follows a record to its hazard, its investigation, its actions and its evidence.

  2. 02Finds

    Finds the look-alikes

    Have we had hard landings on 24R before?A safety manager
    Yes, one in the last 12 months, on the same aircraft six days before SMS-OCC-2026-0412.eAvy | 2 credits
    SMS-OCC-2026-0405Look-alike | Firm landing on runway 24R | C-FNWK | 2026-10-31Closed
    From 214 occurrences in your register | Sample data | Northwind Air, a fictional airline

    Surfaces earlier occurrences that resemble this one, for a person to link or dismiss.

  3. 03Drafts

    Drafts the next step

    86%
    eAvy proposes | Record link | 86% confidenceLink SMS-OCC-2026-0405 to SMS-OCC-2026-0412 as a related occurrence

    Same aircraft, same runway, six days apart.

    Nothing is written until a person decides.

    A classification, a risk, a summary or an action, each as a suggestion with its confidence.

  4. 04Briefs

    Briefs you

    What changed in 24 hourseAvy brief
    The highest-risk open occurrenceseAvy brief
    This week’s safety briefeAvy brief

    What changed in 24 hours, the highest-risk open occurrences, this week’s safety brief.

A person decides

Nothing is written until someone accepts it.

Each suggestion is a card to accept, change or dismiss. The decision is kept on the record with the reviewer’s name.

ClassificationSMS-OCC-2026-0412 | stage 3 of 11Not set
92%
eAvy proposes | 92% confidence | 2 creditsRunway Safety, Abnormal Runway Contact

From the narrative: a hard touchdown on runway Runway 24R, 2.1 g at touchdown, flight 412 Toronto to Montréal.

Waiting in Classification. Nothing is written until someone decides.
Illustration | built from the sample record

Try the card. It replays a suggestion from the sample record’s History tab.

What it never does

Four rules it does not break.

  • Nothing is written without a personA suggestion stays a suggestion until someone accepts it, and the decision keeps their name.Holds
  • It does not move the workeAvy cannot move a record through its workflow or sign anything off.Holds
  • It picks, it does not inventIt picks from your taxonomy and your records. It does not invent a category.Holds
  • It does not learn from your recordseAviora does not use your records to train or fine-tune any model.Holds
Behind eAvy

A fleet of specialist AI agents, in five categories.

Each agent works at a stage of a workflow. A suggestion with low confidence is held back rather than shown.

  • ClassificationReadsProposes the event type and the risk family.
  • AssessmentRatesProposes severity, probability and the risk level.
  • GenerationDraftsDrafts causes from the evidence, or a plain narrative.
  • ValidationChecksLooks for gaps, such as a requirement without evidence.
  • ExtractionPullsPulls contributing factors from a narrative, and similar past records.
An occurrence, stage by stage | where an agent works9 of the occurrence agents shown | names from Settings, AI agents
  1. Stage 1, Report: Narrative Reformulator (Generation)
  2. Stage 2, Triage: Occurrence Triage Assist (Assessment)
  3. Stage 3, Classification: Risk-Family Classifier (Classification), ICAO Event Classifier (Classification), Contributing Factors Analyzer (Extraction)
  4. Stage 4, Initial risk: Risk Assessor (Assessment)
  5. Stage 5, Investigation: Bow-Tie Walk Proposer (Assessment), Root Cause Analyst (Generation)
  6. Stage 6, Action plan: no agent
  7. Stage 7, Governance review: no agent
  8. Stage 8, Action execution: no agent
  9. Stage 9, Residual risk: no agent
  10. Stage 10, Intelligence impact: no agent
  11. Stage 11, Closure: Link Suggester (Extraction)
Settings | AI agents, 30 days
4,618runs
Accepted, of decided71%
Suggestions waiting on a person22

Each agent is on or off, per organization. An administrator sees how many of its suggestions were accepted.

Five occurrence agents | runs, 30 days
Narrative Reformulator40288%
Risk-Family Classifier38874%
Risk Assessor36669%
Link Suggester31081%
Root Cause Analyst9658%

Sample data. The share accepted is counted over the suggestions a person decided.

Credits, not invoices

Measured in credits, with a ceiling you set.

400AI credits a month on Community
Select and Enterprise carry a shared AI allowance for your team.
  • One shared monthly pool for the whole organization.
  • Extra capacity in prepaid packs, with hard limits: reach the ceiling and AI use pauses.
  • Credits are the only unit you see, per session and over the month.
Confidentiality holds

It cannot show what you cannot see.

eAvy works only with what the person asking is allowed to see. A confidential report stays invisible to anyone not cleared for it, even when they ask eAvy about it.In place
Questions buyers ask

About eAvy.

No. eAviora does not use your records to train or fine-tune any model. The processor that handles AI requests is named, with its region, on the sub-processors page.

See eAvy on your own records.

A walkthrough with the people who build eAviora.