Instant-Programming
FRContact

Professional firms · industry · document-heavy organisations

Ask 12,000 pages a question — and get the source

The company’s knowledge is real, but it is locked inside PDFs, procedures, contracts and old meeting notes. Finding it depends on who happens to be in the office that day. When that person leaves, it leaves with them.

Watch the pipeline run
Today
30 min / lookup
Time to production
5 to 8 weeks
What stays human
On sensitive matters — legal, safety, compliance — the answer stays a decision aid. The assistant shows where to read; the expert decides.

Before / after

The same process, run twice

On the left, what it costs today. On the right, the pipeline played step by step, with the data that actually moves from one step to the next.

Today · by hand

30 minutes per lookup

  1. Work out which document holds it12 min
  2. Go through the document, page by page9 min
  3. Check it really is the latest version4 min
  4. Copy out and rewrite the answer5 min

A working assumption, not a measurement taken at your company. You’ll adjust these figures below.

Pipeline · assistant-sur-vos-documents

  1. Your documents come in

    once, at setup

    PDF, Word, scans, intranet, archived email. Badly scanned documents included.

    Document base indexed

    2,184 documents  ·  12,407 pages
Contracts, quality procedures, spec sheets, minutes
Versions and validity dates preserved
Access rights inherited from your directory

  2. They become searchable

    ~40 min for 12,000 pages

    Chunking, vectors (pgvector), keyword index. Search understands meaning, not just words.

    Indexing

    chunks       : 38,942
embeddings   : pgvector · 1024 dim
lexical_index: BM25 (hybrid search)
metadata     : department, version, date, confidentiality

  3. Someone asks a question

    instant

    In plain language, the way you would ask an experienced colleague.

    Question asked

    “What is the withdrawal period under the Legrand
framework contract, and did the 2025 amendment change it?”

  4. The answer is built

    2.8 s

    Hybrid retrieval, then an answer written only from the passages found.

    Answer

    “The period is 30 calendar days (clause 8.2 of the
framework contract). Amendment no. 2 of 04/2025 extends it
to 45 days for orders above €50k only.”

source: framework-contract-legrand-v3.pdf · p. 14
source: amendment-02-2025.pdf · p. 2

  5. Everything is traceable

    —

    Sources cited, questions logged, permissions honoured. An unsourced answer is not returned.

    Guardrails active

    Every statement points to a specific passage
No source found → “I don’t know”, never an invention
Documents outside the user’s rights: invisible
Question log available for review

On sensitive matters — legal, safety, compliance — the answer stays a decision aid. The assistant shows where to read; the expert decides.

Your numbers

What this weighs at your company

Run the numbers on your own figures

Move the sliders: everything recalculates live, from your own volumes.

250lookups/mois
30min
65€/h
75%
What that adds up to

0h/month

Time handed back to your team, every month

0€/year

Value of the time recovered over twelve months

0FTE

Full-time equivalent freed from work that creates no value

0€

What doing nothing costs over three years

Formula: volume × time per item × share automated × loaded hourly cost. The result is only as good as your assumptions — which is exactly what we check together during scoping.

Put these numbers to the test

How we build it

Nothing magic — assembled building blocks

Every block is replaceable and documented. You own the code, the data and the access — that is a condition, not an option.

See the offering Assistants & RAG
  • PDF / Office / scan ingestion (OCR)
  • Embeddings + pgvector
  • Hybrid semantic and keyword retrieval
  • Cited answers, refusal when unsourced
  • Access rights inherited from your directory

Proof

An architecture we already operate: embeddings, pgvector, hybrid semantic and keyword retrieval, sourced answers.

The other scenarios

Often, the real subject is somewhere else

One concrete problem is enough to start

Shall we test these assumptions on your case?

Twenty minutes is enough to know whether this scenario holds at your company, what it assumes about your data, and where to start.

Describe my project