Two layers instead of one
The file separates what describes the client stably — activities and markets, accounting setup, group structure, relevant tax areas, disputes, extraordinary transactions — from what belongs to a single financial year. The first layer is built once and corrected only when something really changes; the second records the year’s news.
The separation is not a filing convenience: it is what prevents rewriting the same context every year and losing, in the rewriting, the nuances understood the year before.
The eight sections of the profile
The permanent profile is divided into eight sections: overview, company and business, accounting and reporting, group and corporate structure, tax, disputes and audits, extraordinary transactions, other relevant matters. Each section shows how many entries it holds, so it is immediately visible where the file is still empty.
Inside the tax section, topics are organised by area — corporate income tax, VAT, withholding taxes, transfer pricing, tax consolidation, international taxation, incentives. A conclusion on an area is written only if the documents support it: the absence of findings is not a finding of absence.
The group chart is data, not a drawing
Group entities and the relationships between them are structured information: name, type, economic interest, voting rights, nature of the relationship. The diagram is their representation, not their source.
The distinction matters when economic interest and voting rights diverge, which is precisely the case in which a hand-drawn chart starts to lie.
The year’s changes, and when they become permanent
The financial-year view records only what has changed, classified with the same sections as the profile. If a change is structural — an option exercised, a regime adopted, a transaction with multi-year effects — an explicit action carries it into the permanent profile.
That carry-over stays a decision, not an automatism: it is for the professional to establish whether a fact of the year now describes the client stably.
An incomplete file is an acceptable result
The AI-assisted setup produces a fourteen-sheet workbook, one for each block of the profile. None of those sheets is mandatory: the import reads the ones present, counts the entries found, lists the missing sheets and the rows it could not interpret, and shows everything in a preview before writing anything.
This is deliberate. A template that refuses to start until it is complete pushes people to fill it with assumptions, which is exactly the opposite of what is needed: areas left blank stay “to be verified” and do not become positive conclusions.
What the file produces
The file produces a paginated PDF with long texts flowing across pages and an optional group chart, a structured Excel workbook across several sheets with filters, and a complete JSON backup covering the entire client archive.
Processing happens entirely in the browser: data stays on the device and never reaches a server. The flip side is that the archive follows the device, not the person: periodic backup is not a word of caution, it is part of the method.
Where the tool stops
The file organises, connects and makes verifiable what the professional has already understood. It does not replace professional judgement, does not interpret the rules applicable to the specific case, and does not produce conclusions on behalf of the person using it.
The classifications proposed by the AI at import time are a traceable draft, with the source stated: they must be read and confirmed before any professional use.