Skip to content
Every figure traces to a ledger row, or the deliverable does not render

The console

Where the judgement is applied, and where the signature happens

The engine computes and the console displays. Nothing on these screens does arithmetic of its own, which is a constraint rather than a limitation: a figure calculated in a browser is a figure the deliverable cannot trace, and this product will not render one of those.

The screens

Eight surfaces, each built around one decision

The question above each screen is the one the person opening it is actually holding. If a screen cannot name the decision it serves, it is a table with a heading.

Engagements

What have we been engaged to do, and for which year?

The engagement is a record rather than an email. Client, kind of work, financial year and assessment year, and the member responsible — captured once, because the assessment year goes on to decide which rate set every computation under it is allowed to use.

  • The assessment year is required and never defaulted — a computation without one is refused at the door
  • One client, many engagements; one engagement, many deliverables and reconciliations
  • The member responsible is recorded, and attribution outlives the account of whoever left the firm
  • Everything under the engagement is scoped to the firm underneath, not filtered in the interface

Imports and the rows that were refused

Did all of it actually arrive?

Every import shows what was parsed, what was rejected and what was a duplicate, and those three have to add up to the file you gave it. The rejected rows are on this screen exactly as the file wrote them, each with its reason — not summarised, not counted, and not in a log somewhere.

  • Parsed plus rejected plus duplicate equals the rows submitted, enforced underneath rather than displayed
  • Each rejected row shown as the file actually wrote it, with the reason it was refused
  • An unrecognised format is an error rather than an empty ledger that balances perfectly
  • A truncated export names where it stopped, instead of silently importing the part it could read

The reconciliation grid

What still does not match, and how old is it?

The two sides side by side, with the difference at the top walking down to zero as items are cleared. Ageing is real: an item carried forward keeps the period it was first raised in, so a stale item looks stale rather than looking new every month.

  • The difference recomputed by the engine on every change, never by the page
  • Carry-forward on the row fingerprint, keeping the period an item was first raised in
  • Clearing an item requires a narration, because a cleared item with no reason is the first thing a reviewer checks
  • Bank, GST and TDS have real workflows; a kind that does not is refused rather than shown as reconciled

Adjustments and their approval

Who proposed this, and who is allowed to approve it?

Proposed adjustments with the reason, the account and the amount, routed for approval. The person who proposed one cannot be the person who approves it, and that is enforced in the database rather than in a screen — role and identity are two separate controls.

  • The proposer cannot approve their own, as a constraint rather than as a convention
  • An articled assistant may not approve at all; a manager may approve, but not their own
  • Every state change attributable to a named person with a time against it
  • Approved adjustments flow into the computation, so the schedule and the adjustment cannot disagree

Working papers and tickmarks

Does this schedule support the figure it is under?

The schedules a reviewer actually reads, with tickmarks and cross-references between them. Each figure on a paper carries the computation lines behind it, so a cross-reference is a link rather than a note in a margin saying where to look.

  • Tickmarks and cross-references held as structure rather than typed into a cell
  • Every figure on a paper traceable to the computation lines that produced it
  • The cross-foot the arithmetic gate performed shown on the paper, rather than asserted
  • Exports keep the working and the cross-references intact rather than flattening to values

Review notes

What did the reviewer object to, and has it been answered?

A note raised against a specific section or a specific figure, assigned to somebody, answered, and closed. This is the mechanism the sign-off gate reads: a deliverable with an open note cannot be signed, so a note is a block rather than a comment.

  • Notes anchored to a section or a figure, not to the document in general
  • Assigned, responded to and closed, with each step attributable to a person
  • An assistant prepares, a manager reviews, a partner signs
  • Open notes are counted where the signature happens, not filed somewhere else

The deliverable and its provenance

Where did this number come from?

The finished document, and the ability to click any figure in it and walk down to the computation line, the working paper and the ledger row. This is the screen that makes the rest of the product checkable rather than merely careful.

  • Every figure traceable to its line, its paper and the row it was imported from
  • The rate set a tax figure was computed under identified by its checksum
  • Exports as a document, a spreadsheet with the working intact, or structured data
  • Unsigned exports carry a draft watermark, and the watermark fails closed

The firm dashboard and the calendar

What is due, and what is at risk?

What is outstanding across the firm, and the statutory calendar underneath it: advance tax, the tax audit report, the return, GST and TDS, with overdue and at-risk flagged rather than left to be noticed. A client-role account sees its own engagement and nothing else.

  • Real due dates rather than a checklist somebody maps onto a diary
  • Overdue and at-risk surfaced without being asked for
  • Work in progress by client, by member and by kind of engagement
  • A client sees their own engagement only, scoped underneath rather than hidden in the interface

Who sees what

Five roles, and the scope is enforced underneath all of them

A practice is not one user. Scoping sits in the layer that builds the query rather than in the handlers, so a new screen inherits it instead of remembering it.

  • Partner

    Everything, including signing. The only role that can sign a deliverable, and only with no review note left open.

  • Manager

    Prepares and reviews, and approves adjustments — but never one they proposed themselves, which is a constraint in the database rather than a rule in a screen.

  • Articled assistant

    Creates and edits records and assembles working papers. Cannot approve an adjustment and cannot sign; reaching the signing endpoint is refused on role before anything else is looked at.

  • Read-only

    Reads the firm’s work and writes nothing. The role most firms actually want for a peer reviewer.

  • Client

    One engagement, and nothing outside it. The role that lets a client see their own working papers without seeing that the firm has any others.

A read, list, search or download that crosses a firm boundary answers as though the record does not exist. A refusal would confirm that it does, and a real engagement id is a fact about another firm’s client list.

The screens are the part worth seeing rather than reading about

A walkthrough against one of your own engagements answers more than this page can: your ledgers, your reconciliations, your reviewer’s objections.

Get in touch

Talk to the people building it

No chatbot and no ticket queue. Tell us what your practice actually looks like — how many members sign, how many prepare, which of audit, tax and compliance you do most of — and someone who works on the software will reply.

info@legosphere.com

Please keep client names, PANs, GSTINs, account numbers and ledger extracts out of this box — it is an ordinary enquiry form, not a channel for a live engagement.