How it works Demo Pricing Guides The science About Français Request the audit
Series · What the science says

Getting an AI to answer on your documents: the landscape of techniques, and our choices

Manent — July 9, 2026

« We plug an AI into your PDFs » hides a dozen very different techniques, with known strengths and pitfalls. Here is the landscape as the research describes it, then the principles that guide our choices — not the exact recipe, but the why.

What the research says: the landscape

TechniqueWhat it doesIts limit
Keyword search (BM25) Retrieves passages that share the exact words. Fast, exact on codes (« RR42 », « 45 Nm »). Blind to synonyms and to field paraphrases.
Vector search (embeddings) Retrieves by meaning: « it won't build up pressure » finds « pressure loss » even without a common word. May miss an exact code match; depends on the model.
Hybrid search (fusion) Combines the two and merges the rankings. The research widely recommends it as a robust foundation. Requires careful tuning of the fusion.
Reranking Reorders the candidates according to their true relevance to the question. Compute cost; useless if the initial recall is bad.
Citation verification (grounding) Checks that every claim is supported by a source before displaying it. This is what makes hallucinations drop. One extra verification call; requires good sources.
GraphRAG (knowledge graph) Builds a graph of entities to reason over complex links. Heavy to build and maintain; often oversized.
Fine-tuning the model Retrains the model on your data. Costly, freezes the knowledge, does not cite its sources, goes stale.

Our principles of choice (not the recipe)

Our exact sequence — the order, the thresholds, the settings — is our engineering work, and we don't publish it. But the principles that govern it are transparent. They all stem from the same obsession: a wrong answer costs more than no answer.

1. Hybrid, because the field speaks in synonyms AND in codes

A technician writes « it's getting hot » but your manual says « bearing overheating » — and they also type « RR42 » which must match to the character. Neither the keyword alone nor the vector alone is enough. We combine, as the research recommends.

2. Verification is not optional

This is the principle that most distinguishes us from a « chatbot on PDF »: an answer not supported by a source is not displayed. It's exactly the mechanism that the literature associates with the highest rates of citation accuracy.

3. The right to stay silent

When a case falls outside what's documented, the assistant tells you so and turns that moment into a card to be validated. Research on system reliability confirms it: answering where you are sure makes a system more trustworthy.

4. A solid hybrid foundation, ready today

For documentation-based troubleshooting, we favor a robust hybrid foundation and strict verification — effective from day one, simple to operate. Knowledge-graph approaches remain in reserve for the cases that will truly justify them, when the complexity of the links demands it.

5. Your documents read live — the source always on the page

Our approach reads your documents live: every answer cites the page and stays current with every update of your manuals, and your data trains no model. That's what keeps the exact citation and the isolation between clients — where retraining a model would freeze it and make it lose the link to the source (documented trade-off).

The common thread: we don't choose the most impressive technique, we choose the one that makes the answer verifiable. Everything else follows.
Honest transparency: this article describes our principles, not our configuration. The precise assembly (order, thresholds, models) is what makes the quality — we improve it continuously and we don't disclose it. What we commit to showing, however, is the result: where every answer comes from, and its reliability measured on your data.

References

  1. Lewis, P. et al. (2020). Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. NeurIPS.
  2. Robertson, S. & Zaragoza, H. (2009). The Probabilistic Relevance Framework: BM25 and Beyond.
  3. Recent systematic reviews of RAG (hybrid search, reranking, citation verification) and work on GraphRAG (2024–2025).
Discuss my case