Guides

What we explain every week to equipment manufacturers and operators, put in writing. No theory: numbers, pitfalls and concrete decisions.

Preparation

Preparing your technical documentation for AI: the 5 dimensions that matter

Duplicates, scans, contradictory versions, documentation blind spots — how we objectively measure whether a corpus is ready, and what to fix first.

Costs

How much does digitizing old manuals cost — and how to cap the bill

Text layer or scan? Price per page, re-billed pages, a quote before spending: the questions to ask before paying for OCR.

Compliance

Law 25 and AI assistants: the questions to ask before uploading your documents

Personal contact details in service manuals, data isolation, the right to erasure: what an AI vendor must be able to answer you.

Comparison

An assistant that knows YOUR machines — and cites your pages

What changes with an assistant built on YOUR documents: the answer for your exact equipment, cited to the page, with your data staying with you.

Series · What the science says

We're not "just a RAG." Every engine choice draws on published work. These articles cite the research and explain, honestly, what we do with it. The overview, module by module: Manent draws on science.

Science

What research says about AI that cites its sources

Providing documents isn't enough: the science of RAG requires verifying every claim against its source. How we made it the heart of the engine.

Science

AI distorts documents silently — why "a generalist assistant + a PDF" isn't enough

A Microsoft study (DELEGATE-52) shows that even the best AI drift silently over time. What that implies, and how Manent is designed against this class of error.

Science

The knowledge that retires: what the research quantifies

Tacit knowledge, ~40% of the workforce eligible to retire, billions lost: what the research says, and how Manent captures it before it leaves.

Science

30 years of research on augmented reality in maintenance

From Boeing 1992 to knowledge-management frameworks: what the science establishes, and the exact place of glasses at Manent (hint: a support tool).

Science

The PRISMA lesson: traceability applied to your documentation

The scientific standard for source traceability, transposed to technical documentation: nothing silently ignored, everything reproducible.

Science

Why our approach is built this way

Hybrid search, reranking, citation verification: what the research says, and the principles that guide our choices — oriented toward the verifiable answer.

Science

Answering only when it's certain

Selective prediction, calibration: answering with confidence when it's solid, and pointing to your expert otherwise — trust through verification.

Science

Human-in-the-loop: what research says about human validation

AI proposes, the human decides: why human oversight keeps a system from amplifying its own errors, and how we made it mandatory.

Science

The reliability the field demands

Grounding as the path to verifiable answers: what the research says, and how Manent delivers a reliable answer, backed by the exact page.