Case study · chartered accountancy · Mumbai
Replacing a shared Excel template with structured records, controlled periods, scoped access and reporting. In daily use since 25 August 2026, on the firm’s own server and under their own domain.
Built for Vatsaraj & Co., Chartered Accountants, who agreed to be named. The video and every screen below are the same software running on invented data. None of the firm’s own data appears on this page.
A read-only copy of the same system. Pick a seat (employee, team lead or administrator) and click through every screen. Nothing you type is saved.
The situation
At a firm of about 50 people across audit, tax and GST, billable time was being recorded against roughly 500 clients in a shared Excel template. The obvious complaint about that arrangement is that chasing timesheets is slow, and it is — but that is not what was actually breaking.
Their template carried a 500-item client dropdown and a second, separate 500-item client group dropdown, with nothing linking the two. The written instructions asked every person to hand-check that the two agreed, on every row they typed. That is job costing built on data nobody can trust, and no amount of chasing fixes it.
The second defect was that a fortnight’s timesheet was only true until somebody edited a closed row, and afterwards nobody could tell that they had.
What the system does about it
Each screenshot is the delivered software, signed in as the role named, running on an invented firm with invented staff and invented clients.
A client belongs to a group as a fact about the record, not as a second thing a person is asked to remember. Choosing the client is choosing the group. Nothing in master data is ever deleted either — archiving hides a row from the dropdowns and leaves every past report exactly as it was, so last quarter’s number does not move because somebody tidied a list this quarter.
Entry for a week closes on a date computed from the calendar, not on an email reminder that somebody has to remember to send. After that an administrator can reopen the period on purpose when a person genuinely needs it — which is a different thing from a closed row quietly changing.
The same screen is the everyday one: a week at a time, entered against a client and an assignment type, with the day’s hours totalling as you type.
A shared file gives everybody everything. Here a person sees their own week; a manager sees the people who report to them, with drill-down, and nobody else; an administrator sees the firm. The navigation itself is shorter for a manager than for an administrator, because the screens they cannot use do not exist for them.
And the reason all of it exists
Time pivots by client, client group, month, assignment type, segment, person, reporting partner or reporting manager, with drill-down from any row to the underlying entries and an Excel export that honours the filters actually on screen. The spreadsheet becomes an output rather than the system of record.
Also part of the engagement
Their existing Excel template was read to recover the real data model — segments, assignment types, the client and group lists, the financial-year structure — rather than inventing one and asking them to adapt.
9,178 historic entries moved across and reconciled against the source, so the new system opened with the firm’s own history in it.
Apply, preview the working days a request will actually consume, approve or reject, withdraw, cancel — with Sundays and declared holidays handled explicitly rather than silently shortening a request.
Their server, their domain, their tenant, their backups, their credentials. The restore was tested before go-live, not merely configured.
Source, runbook and credentials in their hands, with support terms agreed as part of the scope rather than promised open-endedly.
Being precise about it
It proves that I can carry an internal application through real requirements, a data migration, deployment on a client’s own infrastructure, and the changes that keep arriving after go-live.
It does not prove a measured saving. No business KPI or baseline was agreed before the build, and nothing has been measured in production since. Establishing one — time spent consolidating and chasing, or submission completeness by the deadline — is work in progress with the client. Until it lands, I would rather describe delivery accurately than convert clicks saved into rupees saved.
It also does not establish broad business-audit expertise, production AI reliability, or demand for this as a product other firms would buy off the shelf. Those are separate claims and I do not have the evidence for them yet.