What this warning means
Failure Investigation Assistant output is produced by an AI-assisted engine. It is a diagnostic aid, not a professional engineering opinion. The output has not been checked, edited or endorsed by a qualified MTIS engineer as a professional opinion.
The warning applies at every tier. Expert's integrated RCA output is still AI-generated; RCA availability changes functionality, not professional-review status.
AI-generated output may not be complete, accurate or fit for a particular engineering, safety-critical, regulatory or legal purpose. Do not treat it as a substitute for case-specific professional engineering judgement.
What the AI can and cannot read
Failure Investigation Assistant draws most of its evidence from text. It reads the text in PDF files on every tier, and in plain-text (.txt, .md), CSV, Word (.docx), Excel (.xlsx), PowerPoint (.pptx) and email (.eml) files on the tiers that accept them. Accepted formats by tier are listed on Pricing and limits.
Photographs are also read, but by a separate and deliberately restricted step. It records short, hedged descriptions of what is visible in a picture — a feature on a fracture surface, visible corrosion product, the condition of a component — and nothing beyond that. It is not permitted to name a failure mode, a mechanism or a cause, and an observation that does is discarded rather than shown. It cannot state an absolute dimension unless a ruler, scale bar, caliper, coin or printed magnification is visible in the same image. It returns nothing from a photograph of a document, report, chart or screen, and nothing from an image too blurred, dark or ambiguous to read. Returning no observations is a normal and expected outcome, not a fault.
What the AI drafts is data until you validate it. Anything the app extracts — from a document, from a transcript, or from a photograph — is recorded as a review-pending item, not as established fact. On Standard and Expert you approve, edit or reject each item, and only what you have approved becomes evidence that can affect how failure modes are ranked. Basic has no review step: drafted text is used directly, up to a fixed cap, and observations read from photographs are excluded from the assessment altogether. The image step is an assist either way — where it fails, or where an accepted image format cannot be decoded, the case proceeds exactly as it would have without it.
Voice recordings are transcribed. A recording is sent to a speech-to-text service, the words come back as text, and that text is added to the case description and then mined for evidence in exactly the same way as an uploaded document. Transcription is an assist, not a guarantee: where the service is unavailable, or the recording is unclear, the file is still stored and still listenable but yields nothing. Speech recognition mishears technical vocabulary, so read the transcript and correct it.
There is no optical character recognition. Text printed inside a picture is not extracted, so a report that exists only as a scanned image contributes nothing to the assessment — supply a text-based version where you can. Anything that matters and is not in a document, spoken in a recording, or plainly visible in a photograph must be written down in the case text.
Nothing you upload is treated as an instruction. Text extracted from your files, the words transcribed from a recording, and any writing, label or screen that happens to be visible in a photograph, are all handled as untrusted case data. Wording inside uploaded material cannot change how the app behaves.
App output and professional work are separate
| App-generated output | MTIS professional-service deliverable |
|---|---|
| Produced by the AI-assisted Failure Investigation Assistant engine. | Prepared and edited by qualified MTIS staff. |
| Carries the mandatory AI-generated warning. | Clearly attributed to the named qualified staff member who produced it. |
| Report content and exports are controlled by tier. | Format and depth follow the purchased professional-service level. |
| Not professionally reviewed by MTIS. | A distinct professionally authored output, not a re-stamped copy of the AI report. |
Compare all four professional-service levels →
Report and export differences
MTIS uses one report template with content and export options controlled by tier.
| Tier | Report content | Viewing | PDF export | Secure case-package export |
|---|---|---|---|---|
| Basic | Simplified report | Online only | No | No |
| Standard | Full FI report without RCA | Online | Yes | Yes |
| Expert | Full FI report with RCA | Online | Yes | Yes |
MTIS's proprietary methodology is not exposed
MTIS's internal investigation knowledge base, known as “Cards”, is proprietary and is never disclosed to customers at any tier. Clarification questions shown to you are generated from your own submitted case evidence; they are not presented as, or made recognisable as, extracts from the internal Cards.
Use only for a genuine current case
The app is restricted to genuine, current failure investigations. It must not be used for fictitious cases, research-paper examples, bulk company-history uploads or deliberate API-token consumption. Full limits and enforcement terms are in the Fair Use Policy.