Technology
Technology for filings that have to be right
Most filing platforms ask you to move your document into their editor and trust their tagging. BluePrint was built the other way around: your Word document stays the document, every figure is read in its accounting context before any AI touches it, and nothing reaches EDGAR until your reviewer says so.
Three things BluePrint does that most platforms don’t
Each one removes a place where filings go wrong. We describe the result here; the engineering behind it is shared under NDA.
Your Word document is the filing, through every conversion
A filing is converted several times on its way to EDGAR: from Word to EDGAR HTML, from HTML to Inline XBRL, and back through every late draft. In most tools, each conversion is a fresh start, and a tag applied on Tuesday has to be found again on Thursday. The usual fix is to make you abandon Word and draft inside the vendor’s editor.
BluePrint keeps track of every element of your document as the same element across all of those conversions and every re-upload. The practical result: tags survive late changes, any two versions compare word for word, and the Inline XBRL you file is the document your team drafted, not a re-keyed copy of it.
What this means for you: finance drafts in Word until the end, and nobody re-tags a 10-K because the auditor moved a paragraph.
Statements are read as statements, not as tables
To a web browser, a balance sheet is a grid of cells. To an accountant, it has a shape: line items that roll into subtotals, columns that are periods, parentheticals that flip a sign, a bold row that is a total and an indented row that belongs to it. Tagging tools that work from the grid alone make the mistakes you would expect: a subtotal tagged as a component, a parenthetical read as a positive, a prior-period column mistaken for the current one.
BluePrint recovers that shape first, from the way the statement is actually laid out in your document, and only then asks what each figure is. The AI never sees a bare number; it sees the number in its row, its column, its section and its sign.
What this means for you: fewer wrong-sign and wrong-period errors in the first draft of tags, and reviewers who correct judgment calls rather than clerical ones.
The model adjudicates among plausible elements; it doesn’t search for them
The US GAAP taxonomy offers thousands of elements, many with near-identical definitions. The right one for a line item is usually settled less by the definition than by practice: how comparable filers have tagged comparable captions in filings the SEC has already accepted.
Rather than asking a model to search the whole taxonomy, BluePrint gives it a short, ranked shortlist for each line item, built from the elements comparable filers have actually used for comparable captions in public EDGAR filings, including the near-substitutes a reviewer would weigh. The model then adjudicates: it weighs each candidate’s definition against the caption, the section, the sign and the period, and states its reasons. Rules test the result. No single step is trusted on its own, and the only model we train ourselves learns from public filings, never from your drafts.
Every element on the shortlist exists in the taxonomy version your filing declares, and deprecated elements are never proposed.
What this means for you: suggestions that match what your auditor and the SEC staff expect to see, with the unusual cases raised to a person instead of guessed.
BluePrint was designed and written by a practicing securities lawyer. The order of operations above, structure first, precedent second, a person last, is how a filing is reviewed, built into software.
The path every number takes
From your draft to EDGAR, each figure passes through the same five steps.
-
01
In your draft
Your team drafts in Word. BluePrint prepares it for EDGAR with tables, underlines and page breaks intact.
-
02
Read, then suggested
BluePrint works out the statement’s structure, then AI proposes the XBRL concept, period and scale for the figure, starting from how comparable filers tag it.
-
03
Checked by rules
Rules drawn from SEC requirements test the suggestion. What can’t be confirmed goes to a person.
-
04
Approved by a reviewer
Your reviewer approves, corrects or flags it. Nothing enters the filing without this step.
-
05
Validated, then filed
The finished filing is checked against EDGAR’s rules, and you can test-file it before it goes live.
The machine steps are fast so that the time in a filing cycle goes to review, where it belongs.
Five commitments built into the platform
People decide what is filed
AI proposes; your reviewers decide. No suggestion becomes part of a filing until someone approves it, and each approval is recorded with the reviewer’s name and the time. Live submission to EDGAR is a separate permission, so preparing a filing and transmitting it can sit with different people.
Suggestions are checked before anyone sees them
Every AI suggestion is tested against rules drawn from the SEC’s published EDGAR requirements and the standard taxonomies. When a suggestion can’t be confirmed, BluePrint raises it to a reviewer as an open question instead of filling the gap with a guess.
Every number keeps its history
Each tag stays attached to the printed number it describes. BluePrint keeps a running record of how each one was proposed, changed, carried forward or approved, so the answer is there when a question comes up after filing.
Your document stays the source
Teams keep drafting in Word to the end. Every upload becomes a saved version you can compare with any other, word by word or side by side, and because BluePrint tracks each element across conversions, existing tags stay in place when late changes arrive.
The rules you’re held to are the rules we check
Before you file, BluePrint validates the assembled filing against the EDGAR Filer Manual, keeps errors that would stop EDGAR separate from warnings, and lets you run an EDGAR test filing from the workspace before the live one.
What the AI does, and what it doesn’t
BluePrint’s AI works inside the filing you are preparing, with its statements, notes, prior filings and validation results in view.
It does
- Propose XBRL tags for the financial statements, the notes and the cover page
- Carry last period’s approved tags forward and point out what is new or changed
- Answer questions about the filing in front of you and explain validation messages
- Prepare drafts, tables and proposed edits when you ask for them
It doesn’t
- Approve a tag, change a filing or transmit to EDGAR on its own
- Learn from your drafts before they are public. After filing, de-identified tag corrections may refine our tagging unless you opt out.
- Ask you to take its word for it. AI suggestions can be wrong, so each one is shown in context for a reviewer to confirm.
The AI providers we use, and what each may retain and for how long, are listed in Trust & Security.
Engineered around deadlines and confidential drafts
No maintenance near due dates
No planned downtime on a Form 10-K, 10-Q or 20-F due date or in the two business days before it, or between 4:00 and 10:00 p.m. Eastern on business days.
Credentials kept apart
EDGAR credentials get an additional layer of encryption, are never written to system logs, and are used only to send submissions a user starts.
Export at any stage
Download a complete, standard EDGAR package whenever you need it, for sign-off, archiving or filing through another agent.
No tracking where drafts live
No advertising, tracking or analytics scripts run on the signed-in pages of the product.
Hosted in the United States
Customer data is encrypted in transit and at rest.
72-hour incident notice
If we confirm an incident affecting your data, your workspace administrators hear from us within 72 hours, to support your own Form 8-K Item 1.05 assessment.
What we publish, and what we share under NDA
This page describes what BluePrint does and the results it delivers, not the engineering behind them. If your security, IT or risk team needs more, we provide a written description of our controls on request, answer one security questionnaire a year for paid plans, and walk qualified teams through the technology under a nondisclosure agreement.
See it on your own last 10-K
Every free workspace includes a one-time AI test-tagging run and validation report on your most recently filed 10-K, using only public EDGAR data. No confidential draft needed.