What a cutover wizard should ask (and what it should refuse)

Search demand for a cutover wizard is a polite way of saying the last implementation was a person on a screen-share. I write the impolite version because the polite version never asks which systems you are leaving.

Sevak Girard here. I built Ficary so the mapper is the wizard. I said the growth-engine line on the record in Accounting Today, July 2025. The desk version is simpler: automate the obvious, hold the leftover. A wizard that hides the leftover is a second product.

A plain definition

A cutover wizard should ask enough to map a native export and refuse enough to keep the books honest. It is not a tour. It is not an AI that remaps in secret. It is not a workbook we repair after you fill it wrong.

The questions are boring on purpose. Track: owner books, firm, both, or a roll-up. Source: bank, named export, or both. History: none, open WIP and AR, or dated history. Rooms this week: clients, work, time, books. Mapping mode: manual, assisted, or auto on exact matches only. Named signer. First proof: invoice, job, or report. And the question firms skip: **which systems are you leaving, and you may pick more than one.**

If that list feels long, compare it to a six-month implementation that still cannot age unbilled work on a date. The wizard is short because the physics is short. Open how to leave a five-tool practice stack if the leftover is the pile. Open data migration for practice software if the leftover is opening numbers. Open what to export before you switch if the leftover is the file. Money page: practice software. Talk to a person: early access.

The stack question is the engine

One SKU is a file. A firm is a stack. If the wizard only lets you pick the product that produced this CSV, you will land clients from the CRM and forget the time app, the portal, and the bank. Multi-select does not mean we import five files at once. It means the leftover list of systems is visible. This file maps one pack. The rest wait with a reason.

Standard CRM tables have a default map because those names are public: client-shaped records, people, prospects, jobs. Custom columns on a practice layer are not a public cookbook. They are leftovers. They become fields on the record you already have. They do not become a table we invent for your firm. One schema is how the second office still closes.

Mailboxes are not a pack. Calendar is not a pack. Staff users are names on time rows, not logins we mint from an export. If the wizard offers to import email, close it. Reconnect mail from the suite you already run.

What the wizard must refuse

It must refuse a vendor family as an export path. “The tax suite” is not a file. Pick the product that actually emits clients or work. It must refuse a File menu we invented because the surface is missing. Recovery is a named report, an API, or omit work. It must refuse cutover without a count acknowledgement. Mapped 412, skipped 18 is a sentence a signer can read. “Looks good” is not.

It must refuse auto-mapping alias columns without a human when the mode is assisted. Exact matches may pre-confirm in auto. Everything else waits. It must refuse day-two utilization as go-live. Optimization is a later statement of work. Go-live is one proof job on honest opening numbers.

It must refuse starting history at zero because import was hard. Opening WIP keeps the date it was incurred. If you cannot keep the date, you do not have WIP. You have a guess booked as this month’s revenue.

See practice software implementation timeline if someone sold you a quarter of workshops instead of a week of maps.

Rooms this week are leftovers with a calendar

You will not land every room on Tuesday. That is allowed. Unchecked rooms are leftovers, not failures. If time is leftover this week, the wizard blocks a time file. That is fail-closed, not rude. The inbox should show “time is leftover this week” the same way it shows an unmatched bank line.

Books as a room means the ledger, not a dashboard. If the firm keeps another GL as habit, say so. Ficary can be the practice money while their file remains the GL book. Running both stories in one head is how the close splits. Pick a posture. Write it down. Sign it.

The first proof is small. Send an invoice, open a job, or open a report that matches the books. If the proof needs a custom dashboard, you are not live. You are decorating.

The metric

Discovery complete versus files imported versus leftover systems is the Monday number. Then count ack. Then one proof URL with real rows. If sales needs a screen-share to finish the map, the wizard failed and we hired a desk. I will not staff that desk. The mapper is the product.

Assign a signer who is allowed to say the counts are true. If the signer is “whoever is free,” the cutover will move when that person is on leave. Named human. Dated answers. Same as a close.