Anima Scrinia
The Bureau of Memory.
Scrinia is the corpus layer. It ingests an organisation's entire back catalogue, reports, decks, transcripts, tables and scanned documents, reads and enriches every file on the way in, then answers questions in natural language and hands back the source documents to prove it. Your archive stops being storage and starts being an asset you can ask.
The Largest Asset Nobody Can Search
Every established research organisation is sitting on decades of work. Studies, debriefs, topline decks, verbatim transcripts, tracker waves, competitive reviews. It cost a fortune to produce and it is almost entirely unreachable. The person who knows where the 2019 category study lives is the person who ran it, and increasingly that person has left.
So the same question gets re-answered. A client asks something the organisation already knows, and the honest options are to re-field it or to guess. Both are expensive, and one of them is worse than expensive.
Full-text search does not solve this, because nobody remembers the words. The question is "what did we learn about price sensitivity in this category", not "which file contains the string price sensitivity".
Scrinia turns the archive into something you can interrogate. Ask it a question in plain language and it answers from the corpus, with the source documents attached so you can open the original and check the working.
Ingest, Enrich, Retrieve, Control
Four stages, each one solving a failure the previous generation of document search never handled: scanned files nobody could read, answers nobody could verify, and access nobody could govern.
Ingest Anything You Already Have
Point Scrinia at existing folders or upload directly. Word documents, PowerPoint decks, PDFs, Excel workbooks and scanned paper all go in the same way. There is no migration project and no requirement to restructure the archive first, because the archive's disorder is the problem being solved, not a prerequisite to solving it.
Enrichment on the Way In
Every file is read as it lands. Text is extracted, anything scanned goes through OCR, and keywords and entities are tagged automatically. Personally identifiable information is flagged rather than quietly indexed, so a corpus built from decades of fieldwork does not become a privacy problem the moment it becomes searchable.
A Human Review Gate
Nothing publishes itself. Enriched files sit behind a review step where a person confirms the extraction and tagging before the document enters the searchable corpus. Automated enrichment at volume with a human decision at the boundary is what keeps the index trustworthy instead of merely large.
Answers That Cite Their Sources
Ask a question in natural language and Scrinia answers from the corpus and returns the documents it drew on. You open the original and check it. This is the difference between a research tool and a plausible-sounding guess: an unsourced answer is not an answer, and Scrinia does not produce one.
Control and Export
An administration layer governs who can see what, so a corpus spanning multiple clients and multiple years can be opened to the right people without opening it to everyone. Results export to CSV. Access is a branded web link with nothing to install.
Scrinia Reads. It Does Not Compute.
Searching your reports is reading. Asking what 25 to 34s said about price across ten years of survey data is arithmetic over a dataset. Those are genuinely different jobs, and Scrinia does the first one.
Being explicit about that boundary is deliberate. A tool that quietly blurs the two produces confident answers to questions it never actually computed, which is the exact failure mode this product exists to remove.
For analysis over raw survey data, that is Anima Nima, the natural-language query layer over your research data.
The Product That Improves Every Other One
Scrinia is the only part of the platform whose value compounds across the rest of it. Everything else gets better when the corpus is present.
Personas That Know Your Category
Essentia respondents grounded on a client's real back catalogue stop being generic. The persona is informed by what the organisation has actually learned, not by what a model assumes about the category.
Prediction With a Historical Base
Praesaga gains a decade of prior findings to reason from, instead of projecting from nothing. Foresight built on an organisation's own record is a different proposition to foresight built on a blank sheet.
Briefings With Institutional Memory
Vigilia briefings can reference what the organisation already knows about a signal, rather than reporting it as though it were new. Context is the difference between an alert and an insight.
Calibration Against Prior Findings
Calibra gets a body of historical evidence to validate against, extending calibration beyond the current wave of panel data into everything the organisation has previously established.
Frequently Asked Questions
What file types can it handle?
Word, PowerPoint, PDF, Excel and scanned documents. Scanned material goes through OCR on ingest. One known limit worth stating plainly: clean single-page images OCR well, while long scrolling screenshots degrade badly. If the archive is heavy on the latter, that is worth raising before scoping.
How do we know the answers are right?
Every answer returns the source documents it was drawn from. You open the original and check it. Scrinia is built so that verifying an answer is a click, not an act of faith.
Can we restrict who sees which documents?
Yes. The administration layer controls access per user, which is what makes a multi-client or multi-year corpus safe to open up at all.
Does our archive need organising first?
No. The disorder is the problem Scrinia solves. Point it at the folders as they are.
Can it analyse our raw survey data too?
That is a different product. Scrinia reads documents; Nima queries datasets. They are scoped and quoted separately, deliberately, because folding them together is how this work gets underestimated.