Decide what you are actually replacing.
A bookkeeper does many jobs at most SMBs: data entry, bank reconciliation, AP processing, AR follow-up, payroll prep, tax filing support, MIS reports, and the occasional firefighting. AI bookkeeping replaces the first three reliably and assists the next three. It does not replace tax filing or firefighting. So the realistic question is not "fire the bookkeeper" but "what does this role look like when 60 percent of the data entry disappears?"
Most SMB owners we work with keep the person. The role becomes more analytical: working capital management, vendor negotiation, branch P&L review, AR collections strategy. The person moves from typist to analyst. Their value to the business goes up, not down. If the person cannot make that transition, then the conversation is harder, but it is a conversation about role fit, not about cost cutting.
The 90-day transition framework.
Days 1 to 30: parallel run. Keep the bookkeeper doing everything they normally do. Stand up Nonari (or your chosen tool) alongside, with the same chart of accounts and the same opening balances. Both systems run. At month-end you compare the two trial balances. Differences are investigated and fixed in the new system.
Days 31 to 60: switch the source of truth. The new system becomes primary. The bookkeeper now reviews the AI-coded transactions instead of doing the coding. Bank reconciliation moves to the new system. Reports come out of the new system. The old system is kept read-only for historical reference.
Days 61 to 90: redeploy the role. The bookkeeper takes on the new analytical work. New tasks: weekly cash forecast, monthly AR aging review with action plan, branch P&L commentary, vendor payment optimization. The owner gradually steps back from day-to-day finance involvement.
- Days 1-30: parallel run, same COA, reconcile at month-end
- Days 31-60: switch source of truth, bookkeeper reviews not codes
- Days 61-90: redeploy role to analytical work
- Day 90: 1-on-1 to confirm fit and set new metrics
Controls you need from day one.
Three controls are non-negotiable during transition. One: every change to the books leaves an audit log entry, attributed to a named user. Two: bank statement upload is restricted to a controlled list of users, and uploads are timestamped and immutable. Three: a maker-checker rule on payments above a threshold (e.g. $5,000) so no single person can both initiate and approve a payment.
Nonari has these built in: organization-scoped audit log on every change, role-based access control, and a permissions matrix at /settings/permissions that is the single source of truth for who can do what. The setup takes an afternoon. The peace of mind it gives during a transition is worth ten times the setup time.
Worked example: a 3-branch café group.
A 3-branch café group in Sydney with $3.6M annual revenue. The bookkeeper had been with the family for 9 years. The owner did not want to fire her but recognized the data entry was killing her time. They ran the 90-day plan starting January.
By day 30 the trial balances reconciled to within $400 (timing differences on inter-branch transfers). By day 60 the new system was primary and the bookkeeper was reviewing 2,300 auto-coded transactions a month, with about 12 percent flagged for manual decision. By day 90 the bookkeeper had built a weekly cash forecast that surfaced a $48,000 quarterly BAS payment the owner had forgotten was due. The role transition paid for itself in month four with the cash forecast catch alone.
What to do with the historical records.
Keep the old system read-only for at least 7 years (most jurisdictions — IRS in the US, HMRC in the UK, CRA in Canada, ATO in Australia — require 6 to 10 years of retention). Export all transactions, journal entries, and reports as PDFs and CSVs as a backup. Import the closing balances into the new system as opening balances at the cutover date. Do not try to re-import historical transactions: the cleanup cost exceeds the value, and you have the old system for reference if needed.
A common mistake is trying to migrate every historical transaction. Resist. Cut over at a clean date (preferably end of fiscal year or end of quarter), bring in opening balances and active inventory, and start fresh. Comparative periods can be built up over the next 12 months as the new system accumulates history.
When this transition fails.
Three failure modes. One: the owner does not actually want to be involved during the transition, so the bookkeeper goes through the motions and quietly keeps the old workflow alive. Fix: weekly 30-minute check-in with the owner during days 1-90, non-negotiable. Two: the chart of accounts gets re-designed mid-transition, so reconciliation never happens. Fix: freeze the COA before day 1, redesign later. Three: the bookkeeper is not given a clear future role, so they resist the change passively. Fix: have the role-redeployment conversation in week one, not week ten.
Done well, this transition makes the finance function more valuable, gives the owner more visibility, and frees the bookkeeper from the grind of data entry. Done badly, you lose six months of clean books and a long-tenured employee in the same quarter. The 90-day framework is designed to avoid the second outcome.