In-house first · hosted in four regions
dossier reads an institution's document archive, keeps track of what is owed and what has arrived, and drafts whatever the work needs next: an appraisal file, a grant dossier, a diligence pack. It is built to run end to end on hardware you own, with models you run yourself. No public AI services. No API keys. No telemetry. Want to see it first, or is your material not confidential? Open a hosted workspace.
This page loads no fonts, no analytics and no third-party scripts. It seemed like the least we could do.
The problem
Any one of them rules out the obvious tool. Together they have kept this work manual.
Appraisal files, grant dossiers, diligence packs and counsel correspondence sit under confidentiality obligations that a consumer AI subscription cannot satisfy. For a development bank or a ministry, "where does the prompt go?" is a procurement conversation before it is an engineering one.
Most of the cost is not judgement. It is reading eight hundred files, reconciling a register against the evidence actually received, chasing what is missing, and assembling drafts that a senior reviewer will then correct. Weeks of it, per file.
In a file of record, an unattributable sentence is a liability. Every figure needs the document, the page and the date behind it. A system that can quietly invent one is worse than no system at all.
How it works
You steer it in plain language. It does the assembly and stops where judgement begins.
PDF, DOCX, XLSX, HTML, RTF, email, scans, call recordings. OCR for the image-only pages, local transcription for the audio. Every binary gets a greppable text mirror with a checksum and an honest note on what was lost.
Hybrid keyword and semantic retrieval, fused by rank. Citations resolve to the original document, page and sheet, never to a scraped copy. Real corpora are multilingual, so retrieval is too.
The register is the source of truth and the roll-up is computed, never typed. Rows whose status disagrees with the evidence on file are surfaced the day the evidence arrives, not at the next review.
What is owed, by whom, and since when. New arrivals are matched against outstanding obligations, and everything they make stale is flagged as a sentence you can read, not a diff you must decode.
Deliverables are written section by section from the corpus, versioned and dated, in whatever shape the engagement actually calls for, rather than from a taxonomy fixed in code. Word and PDF come out of the data, natively.
Every change the model proposes is a typed patch carrying its before, its after, its citations and its reasoning. Put a project in review mode and each one waits for you to accept or reject it. Let it act, and it applies only what it can cite, with the same record kept. Nothing is sent anywhere, ever, without you.
See it
Captured from the running application on the public documents of a World Bank schools project, with a terms of reference written for the demonstration. Nothing staged and nothing mocked.
Capabilities
Every claim quotes the document it came from, with the locator a person can check: page, sheet, cell, speaker, timecode. Reliability caveats travel with the citation.
Status roll-ups are derived from the evidence rows beneath them. A number nobody typed is a number nobody can mistype.
Every draft is a version with a lineage. Re-draft when the corpus moves, and see exactly what the new material changed.
Each workspace is a real repo and each mutation is a commit. Audit, diff and undo come free with that. They were never features anyone had to build.
Office formats, scans, email archives, web pages, meeting recordings. OCR and speech-to-text run locally, so confidential audio never leaves the room it was recorded in.
The corpus is asked in every language it contains, so a Cyrillic annex does not vanish because the question was posed in English.
Viewer, contributor, approver and owner, enforced at the route and set per project, with passkeys or your own identity provider. Proposing is cheap. Accepting is not.
Archive a finished engagement and it becomes read-only, still searchable, still renderable, reversible at any time. Destruction requires archiving first, and typing the name back.
Use cases
The shape of the work is the same everywhere: a pile of documents, a list of what is owed, and a draft that someone senior will sign. Only the names change.
Appraisal · monitoring
Read the borrower's file, check disbursement conditions against the evidence actually received, track what is still owed, and draft monitoring reports that cite the page.
Calls · applications · reporting
Read every application against the call's criteria, list the missing annexes per applicant, and draft evaluation summaries and programme reports from what was submitted.
Terms of reference · deliverables
Start from the terms of reference, build the deliverable set it asks for, and redraft when the client's new data arrives, showing exactly what it changed.
Procurement · institutional memory
Ask decades of files in plain language, in every language they hold, and get the decision, the date and the document behind it.
Data rooms · issues lists
Review contracts and correspondence where the material cannot travel, keep an issues register that computes itself, and draft the report with a citation behind every finding.
Evaluation · audit
Reconstruct what was planned, what was reported and what the evidence shows, and trace each finding back to the document it came from.
Claims · progress reports
Keep the obligations register current as contractor letters arrive, see what has gone overdue, and draft the quarterly progress report from the file rather than from memory.
Tenders · proposals
Read the tender, find the matching past projects and references in your own archive, and draft the technical proposal section by section in the shape it asks for.
Submissions · supervision
Match each requirement to the evidence on file, see what a new rule makes stale the day it arrives, and draft the response with its references attached.
Donor reporting
One archive, many donors: draft each report in the format that donor asks for, from the same evidence, without retyping the same figure twice.
Interviews · consultations
Transcribe interviews and consultations on your own machines, search them with the reports, and draft briefs that quote the speaker and the timecode.
Yours
If your work ends in a document someone must be able to check line by line, it probably fits. Tell us about it.
In-house · the main use
This is what dossier is built for. The models run where you put them, and the documents never leave. A single-tenant install is the same application with one organisation in it, not a stripped-down fork.
The floor
A single desk-class AI machine, a few thousand euro, will read a full engagement archive and draft its deliverable set in a working day. A second one makes room for a larger model. No rack, no data centre, no account with anyone. That is the entry point, not the ceiling.
The deployment
Put it in your own data centre and the same application serves whole teams: more capable local models, several engagements at once, concurrent users, your identity provider, your backup regime. It stays just as private, and nothing leaves your perimeter.
WHERE clause
Online · live now
The same application runs on our servers in four regions. Sign-up is open; a person reviews each request, and your organisation gets its own workspace, usually within a day.
Heading for an in-house install
Point a hosted workspace at public documents that look like your real work: published appraisal reports, a programme's guidelines, a past tender. Judge the answers, the registers and the drafts. When it holds up, the same build goes inside your perimeter and the confidential files never touch our servers.
When confidentiality is not the constraint
If your material is public, or you are content to process it in the cloud, the hosted workspace is the product: no hardware and nothing to install. Model usage is prepaid and charged at the model provider's advertised price, with no markup.
Read the terms first: they say where your documents are stored and processed. Already have a workspace and lost its address? Have it emailed to you.
Why it can be trusted with the file
Citations are resolved from the indexed corpus, never authored by the model. A proposal pointing at material this project does not hold is refused outright. Every change that applies names the passage it rests on, so a person can open it and check.
Web access is off by default, warns loudly at every start when relaxed, and can be refused per engagement regardless of the install-wide setting. A search query is composed from your corpus; treating that as a mere toggle would be dishonest.
When a tool is unavailable, the model is told, and told why. A model that is simply not told it lacks web access will fill the gap from memory, with all the confidence of a cited fact. Silence is the dangerous default.
The operator's administrative reach covers organisations, roles and configuration. It does not reach the documents, the chat or the registers. Reading those requires a membership your own administrator grants, sees, and can revoke.
Get started
An in-house install is set up with you, on a desk machine or your own servers, with the models you choose. Tell us about the file you would point it at. We answer personally. Or open a hosted workspace today and try it on public documents.
Plan an in-house install Try it online
Or write directly to file@dossier.technology. We keep the address and nothing else.