What changes when the spreadsheet is retired?
A reconciliation spreadsheet is a matching engine built by one person, with rules in formulas and overrides in cells, run at month end from exports, with no record of who changed what. A Finance Controller Digital FTE on ftes.ai matches daily from live connections, proposes each match with the source records attached, follows written rules in its remit, and records every approval with a name. The migration moves the matching and, more importantly, moves the rules from formulas into a remit that a second person can read.
The story of a day with a Finance Controller FTE shows the target state; this guide is about getting there from a sheet.
What should you keep of the spreadsheets?
All of them, read-only, as the baseline: the last two closes' sheets show days to close, the exception list at day one, and the hours the person spent, and they are the comparison the migration will be judged against. Keep the formulas as the primary source for the matching rules: thresholds, date windows, counterparty aliases, fee treatments, and the categories that were always manually overridden. Keep the person who built the sheet as the FTE's first approver, because they know which matches look right. And keep the sheet running in parallel for one close.
The baseline time template in the library gives the fields for the hours the sheet consumed.
What moves to the FTE, and in what order?
First, extract the rules: sit with the sheet's owner for an afternoon and write every formula's intent and every habitual override into the remit's Owns list, the thresholds into the Setup tab, and the facts, fee schedules, counterparty aliases, into the Brand Brain with provenance. Second, record the baseline from the last two closes. Third, connect the bank feed, the ledger, and any processor with dedicated read-scoped accounts, and let the FTE propose while the sheet's owner runs the sheet as usual and approves the FTE's proposals. Fourth, at the first close, compare the FTE's exception list with the sheet's; every difference is a rule that was in someone's head. Fifth, retire the sheet from the second cycle and apply the Two-Cycle Go/No-Go at the second close.
Receivables drafts and supplier matching are later classes; the first move is the bank.
- Rules extracted from formulas and overrides into the remit and memory
- Baseline from the last two closes' sheets
- Parallel run for one close, then the sheet retired
What breaks in the first month?
The hidden logic. Formulas carry rules nobody wrote down, and overrides carry exceptions nobody remembers making; the FTE's send-backs in the first weeks are those rules surfacing, one at a time, as the sheet's owner says that is not how we treat that. Each becomes a remit line or a memory fact with provenance, and the second week's proposals reflect it. Counterparty naming breaks: the sheet had aliases in a lookup tab, and until they are in the Brand Brain, matches to those counterparties are partial. And the month-end habit breaks: the owner has to trust that daily matching happened, which the morning brief shows and the sheet never did.
The myth piece on AI not being trusted with finance describes why the evidence on each proposal, not trust, is the control.
What does the record show after two closes?
Days to close and exceptions open on day one, from the ledger and the FTE's match record, against the sheet's last two closes, labelled first-cycle for the first. The owner's hours from the baseline sheet, redirected to exceptions and judgement. The match approval rate per class with the send-back reasons sorted, and whether exact matches under the extracted threshold met the Promotion Standard. A remit and a Brand Brain that hold every rule the sheet held, readable by a second person, which is the durable benefit. And an audit trail with a name on every entry, which the sheet's version history could not provide.
The metric explainer on days to close describes how the two outcome numbers are read.
What should you refuse during the migration?
Retiring the sheet before the parallel close. Letting the FTE post anything before a class earns it. Extracting the rules from memory rather than from the formulas, because the formulas are the truth of what was done. Skipping the counterparty aliases, which cause most partial matches. And judging on the first close, which contains the surfacing of every hidden rule and is not representative.
The remit template for the Finance Controller FTE gives the Owns, Asks, and Never lists the extracted rules fit into.