Reconciling one bank account is a chore. Reconciling eight — current account, an overdraft facility, two collection accounts, payroll, an FD, and two payment-gateway settlement accounts — is three days of every month that a finance lead at a 40-200 person company does not have. There is enough volume to hurt and rarely a dedicated recon team. Automation solves it, but the migration is where projects fail: teams flip everything on in one month-end and spend the next one cleaning up duplicate postings and unexplained differences.
What does reconciliation automation actually replace?
Automation replaces the line-by-line matching of bank statement entries against ledger entries — not the judgement calls around exceptions.
Most setups pull statements daily (bank API, MT940 over SFTP, or the CSV your net banking exports), normalise them into one format, then apply matching rules: exact amount plus a date window plus UTR or cheque number first, then description-based fuzzy matching. Matched lines post back to your accounting system with a reference; unmatched lines land in an exception queue. What it does not do is decide whether a ₹48,000 UPI credit described as "SWIGGY\*ORDER" is a customer receipt or a refund from a vendor. That still needs a human — but a human looking at 70 exceptions, not 1,400 statement lines.
Why do these projects break?
Most break because the team switches from manual to automated inside a single month-end, with no parallel run to catch mismatches.
The failures are quiet. A rule that matches on amount alone double-posts when a customer pays one invoice in two instalments. Gateway settlements arrive net of MDR fees, so a rule expecting the gross invoice value stops matching every single settlement. Cheques and NEFT cut-offs create timing differences in week one, someone loosens the date window to clear them, and now a genuine duplicate passes through. "Production" here means your books, not your servers — and books are harder to roll back. A bad auto-posting does not crash; it sits in the ledger looking correct until someone finds it by hand three months later.
What does a safe migration path look like?
Run the new matching in shadow mode for two full reconciliation cycles before it is allowed to post anything.
- Weeks 1-2: ingest statements for every account, read-only. No writes to the ledger.
- Cycles 1-2: the automation proposes matches into a comparison sheet; the accountant reconciles manually as always; you compare the two outputs line by line.
- Cycle 3: enable posting for two accounts only — highest volume, lowest ambiguity.
- Cycles 4-6: add one account per cycle.
- One quarter after the last account: retire the manual workbook, but keep the exception report running.
Concrete example: eight accounts, roughly 1,400 statement lines a month. The two collection accounts alone carried 860 lines. Shadow cycle 1 auto-matched 87%, with most exceptions being gateway fee entries. Fixing the fee rule took cycle 2 to 94%. Posting went live on those two accounts in month three. Payroll and the OD account stayed manual until month five — deliberately.
Which accounts should you automate first?
Start with high-volume, low-ambiguity accounts and leave overdraft, forex and intercompany accounts manual the longest.
Payment-gateway settlements are the easiest win: structured files, one counterparty, predictable fee deductions. Customer collections come next, because UPI and NEFT narrations usually carry a UTR you can match back to your GST billing records, where invoice numbers and amounts are already structured. Overdraft and cash-credit accounts are poor early candidates — interest debits, charges and drawing-power entries are irregular and infrequent, so automation saves little and mistakes cost interest. A workable rule of thumb: automate an account once it consistently crosses about 100 statement lines a month and has a stable set of counterparties. Below that, setup and maintenance outweigh the saving.
How do you keep it safe after go-live?
Three controls carry most of the weight: a daily exception report, a hard tolerance on auto-posting, and a named approver for every rule change.
Send the exception report to a person, not a shared folder. A scheduled 9am mail to whoever owns receivables gets read; a dashboard does not. Set a value threshold — anything above, say, ₹1,00,000 goes to manual review regardless of match confidence. Version every rule change with the name of who approved it, and keep a log of what posted automatically and under which rule, so a statutory auditor can follow the trail without reconstructing it from screenshots. Budget for maintenance too: bank formats change, gateways revise fee structures, new accounts get opened. In our experience that ongoing work is better priced as a retainer than treated as a one-off build.
Where to start
Pull the last three months of statement line counts per account and sort them. That list alone tells you which two accounts to automate first and which to leave alone this year. Build the comparison sheet before you build anything else — it is the cheapest insurance you will buy. If you want a second pair of hands on the shadow-mode setup and the exception reporting, Maysan Labs builds exactly this kind of finance workflow; start with two accounts and one clean cycle.
Buying software or a website?
Twelve questions to ask any vendor before you sign: price, your data, fit for an Indian business and support.
Get the free checklistWritten by Maysan Insights at Maysan Labs, a software studio in Gurgaon, India. Follow us on LinkedIn