Questions, answered

Two things go wrong when you paste an FMEA into a general chatbot. First, the file itself: an FMEA rarely survives a copy-paste. Tables get flattened or scrambled, cross-references break, and the diagrams and one-line drawings that carry much of the engineering intent are often dropped, so the model is already reasoning over a degraded copy. Second, the answer: a chatbot will respond to almost any question, including the ones it has no basis to answer, so you get a confident, plausible reply with no way to tell which parts are grounded in your document and which it filled in or invented.

GapAnalyzer is built the other way around. An extraction pipeline parses the document first, preserving sections, tables and cross-references and passing embedded diagrams and images to the model as visual context, so the analysis runs on the full content. Each checklist item is then evaluated strictly against your document: the model decides whether that specific requirement is actually addressed and returns a verdict, Yes, Partial, No or N/A, with a citation to the section it relied on. If your document does not cover something, that is a No, not a guess.

The MTS checklist items we cover are deliberately sparse, assuming the reviewer already carries years of context. We enrich each one with that context explicitly, so every item is judged against the same documented criteria, rather than whatever a given reviewer happens to know, or whatever the model would infer from a one-line requirement on its own.

You also get the actual deliverable: the structured MTS Excel workbook with a verdict and a citation on every item, the same artefact an assurance firm would hand you, not a wall of prose in a chat window.

Your documents are stored on a dedicated server in the European Union (Helsinki, Finland) under GDPR, and all supporting services are self-hosted there. You control how long your documents remain.

To perform the analysis we rely on two sub-processors, both named in our Privacy Policy and Data Processing Agreement, and neither uses your content to train its models. Mistral AI (Paris) processes content entirely inside the EU, so there is no cross-border transfer. Anthropic (USA) sits outside the EU, so that transfer operates under Standard Contractual Clauses, and Anthropic retains API request data for up to 30 days before deletion. We will sign a Data Processing Agreement with your legal team so the full processing chain is explicit rather than assumed.

In most cases, yes. The usual concern is that an NDA forbids sharing the documents with a third party, and using GapAnalyzer does not involve disclosing them to a person at all: the analysis is fully automated and no human reviewer reads your files. Combined with the EU storage, no-training and Data Processing Agreement terms covered under data security above, that makes it more straightforward to reconcile with an NDA than the traditional alternative of emailing the same documents to a consultancy for manual review.

If your NDA imposes specific restrictions, send them to us and we will confirm we can work within them before you upload anything.

GapAnalyzer is not itself ISO 27001 certified yet, but the infrastructure and providers it runs on are. Our server provider, Hetzner, holds ISO 27001:2022 covering the Helsinki data centre where your documents are stored, and both language-model sub-processors, Mistral and Anthropic, are ISO 27001 certified. Your documents are therefore handled within an ISO 27001-certified supply chain end to end.

Every verdict carries cross-references to the exact sections of your source document. Click any one and the original PDF opens beside the assessment, scrolled straight to that section, so you can read the model's claim against the source text side by side.

For models that expose their reasoning, you can also open the full reasoning trace for each item to see how the assessment was reached, not just the conclusion. Between the source view and the reasoning panel, your team can confirm every finding is grounded in your documentation rather than taking it on trust.

A full gap analysis across four core DP documents takes a few minutes, rather than the weeks a consultant engagement typically requires.

Ship owners validate their own documentation ahead of an audit or contract negotiation. Consultancies use it as independent third-party verification: because we don't do remediation work, we have no commercial incentive to inflate or downplay the gaps we find.

Operators get an assessment free of the conflict of interest that arises when the firm finding the gaps is also the firm paid to close them.

We cover four MTS DP guidance checklists: O-01 (DP Operations Manual) and O-02 (Annual DP Trials) from the operations series, and D-02 (DP FMEA Proving Trials) and D-05 (DP System FMEA) from the design series. As far as we're aware, we're the only automated tool covering these items; the alternatives are consultants working from a PDF and a spreadsheet.

While your account is active, your uploaded documents and reports stay on our EU servers; there is no automatic expiry on your side. When you close your account or ask us to delete them, we remove them within 30 days.

Separately, the language-model sub-processors that process the content during analysis hold their copies only briefly and never use it for training.